Like for hundreds of years.

$1 invested in stocks in 1802 was worth over $700,000 by 2012 compared with $1,778 if invested in bonds.

That’s a 395x difference.

Ask a Wharton MBA why stocks outperform bonds and they will dutifully repeat the textbook answer.

“Stocks are riskier than bonds, so they should return more.”

It’s the ‘equity risk premium’.

Makes sense. Kind of.

Bonds are higher up the capital structure and their interest payments are structurally safer.

Plus, bonds have lower volatility when measured over short periods.

But aren’t markets supposed to be rational?

If we all know stocks consistently return more than bonds, then why are they still considered risky?

Over the long run bonds, bills and the US$’s are the risky assets.

And who hasn’t realised by now that the biggest risk to assets over time is inflation and in this regard stocks are safer than bonds.

So why do we keep calling bonds ‘risk-free’ assets?

And if we all know stocks return more than bonds over hundreds of years, and stocks are a better hedge against inflation, then why isn’t there a stampede to sell bonds and buy stocks?

Shouldn’t every investor and insurance company be like Buffet and have as much equity exposure as possible? Why wouldn’t you? It’s what Buffet says to do every year in his letters.

And after hundreds of years of data you would expect investors to ‘take a long-term view’ and bid up stocks to extreme levels ‘because stocks always outperform bonds’ which paradoxically would result in long periods of stock underperformance. You would expect that in an efficient market the outperformance of stocks over bonds would swing randomly back and forth over time due to these shifts in investor psychology.

Instead the outperformance of stocks over bonds is relatively consistent through time, and periods of under performance are short.

Trailing 10Y Annualized Stock Minus Bond Returns, 1802-2022

The Equity Premium Puzzle

To Economists this persistent outperformance of equities doesn’t make sense. It’s a puzzle.

Why is there this free money?

Why do stocks keep outperforming?

Or, to flip the question on its head and ask it another way, why are real interest rates so persistently low?

To put it crudely ‘why are bond yields so crappy?’

For the last 20 years real yields have been below 2%.

With S&P 500 EPS growth of 20% and inflation at 3% why are 10 year yields less than 5%?

Who would buy 10 year yields at less than 7% when stocks are doing so well? And yet they do. In the $ billions everyday.

Jeremy Siegel in his paper ‘The Equity Risk Premium Puzzle’ summarises all the different explanations economists have tried to come up with to explain the persistent outperformance of stocks.

One method to make the numbers work and explain why there is essentially free money on the table (a less risky asset which returns more’) requires assuming that consumers get much higher utility out of consuming goods than by foregoing consumption to invest in equities. But the numbers required to justify this are extreme and not realistic.

Economists also tried to ascribe the equity risk premium to extreme risk aversion. But the amount of risk aversion required to make the numbers work also doesn’t make any sense. Siegel goes through all of these attempts in his paper.

Another thought is that this is survivorship bias. US investors are perpetually scared of a big market ending wipeout which never comes to the US, but which has happened in other countries like Japanese and German stock markets in WW2.

US stock markets have just been lucky.

But lightening could still strike at any time.

Except, when you look at Japanese and German markets and the rest of the world the equity risk premium exists everywhere.

Yes, wipeouts happen, but stocks still outperform.

The equity risk premium effect is persistent across time and markets.

“History shows that investors have been consistently rewarded for taking risk and compounding the dividends and coupons available in equities and bonds.”

-Deutsche Long-Term Return Study

Another thought was that maybe the data doesn’t go back long enough. And maybe the bond data Siegel used under represented bond returns in the early 1800’s.

Edward McQuarrie in his paper ‘Where Siegel went Awry:Outdated Sources and Incomplete Data’ adds a few more years to the data series and recalculates bond returns in the early years to conclude that for the first 130 years (1793-1923) bond and stock returns were the same. The strong outperformance of equities over bonds has only been over the last 100 years.

OK. Whatever.

There were 130 years where bonds performed the same as stocks then another 100 years where stocks did better.

Still seems like a good case for stocks.

So why is this equity risk premium happening and how can we trust it will continue?

I see two explanations which make sense.

We also conclude with a rule of thumb to know when the bull market is over.

Read more

That’s the view from gold mining CEO Jeremy Gray.

We review the gold market set up and the exciting restart of Pilar Gold.

For more information about the Pilar project click the link below.

Our own YWR economic symposium from Montana.

Like Jackson Hole but better!

Get more from Erik in the Substack app

Available for iOS and Android

Get the app

I didn’t want to go to my mom’s boring Church event.

But I said I would.

It was an evening talk about home health care for the elderly.

But I’d promised I’d go early and help set up chairs.

Now I’m glad I helped out with the church event.

The talk was more interesting than I expected.

And I think God rewarded me with a stock idea.

TL/DR: Healthcare costs are exploding and you will likely die in your home with a nurse. Not go to a nursing home. But there is a good stock to play this.

A nurse from a local hospice provider (healthcare at home when you are dying) had come to speak to the elderly at the church about the new trend in ‘palliative care’; low acuity health care at home. She was speaking to 30 elderly in a church hall with everyone sitting in rows of folding chairs. In the rear of the room there was a table with snacks people had brought.

It turns out if you are homebound and have 2 co-morbidities Medicare will pay for nurses to visit you. This is ‘palliative care’. And it is growing quickly. Our nurse assured us everyone in the room would qualify.

The nurses check in on patients in their home and help manage the mix of drugs, doctors and tests these patients have accumulated across their different ailments. This firm’s local palliative care practice had grown from 40 patients 4 years ago to over 200.

The nurse who was presenting spoke well and her anecdotes about what she sees everyday visiting patients helped explain the challenge and why palliative care is growing quickly.

What often happens is a patient goes to their primary physician and gets referred to various specialists for different ailments along the way. Each specialist recommends their own drugs and tests. ‘Take this and this 2/day and check in for a blood test every 3 months’. These prescriptions and tests accumulate.

The palliative care nurse comes to the patient’s home and sees the overall picture. She also sees where the patient lives and how well they can move around the house. And often she can make judgements to reduce some of the pills and tests the patient is taking. ‘I think you can stop taking this, and your breathing is improving so take this test every 6 months instead of 3’. The nurse can give sensible advice from seeing the whole picture on a regular basis, which the individual specialists cannot. She can also prescribe specialised home health services. “I think you need some physio.” or, “you need speech therapy.”

As the talk unfolded and the questions from the elderly started I realised the following:

#1 Home health care could be a better model

To me and everyone else in the room listening to this palliative care sounded highly appealing. I can see why it is growing. A sensible nurse comes to your house, ‘checks in on you’ and has a normal human conversation about how you are feeling, all the drugs you are taking, which of your problems are getting better and which are getting worse and need help. Effectively, a medical advisor.

It also came up that in the meeting that in this town getting access to a primary care physician was challenging. The town has been growing rapidly and there aren’t enough doctors. The nurse explained that some or her patients are using palliative care as their primary healthcare.

And another massive realisation.

#2 These nurses are also financial advisors.

What I noticed from the Q&A from the elderly in the room. Nobody knows what insurance and medicare will pay for and what it won’t. It is a confusing maze to them. The nurses (or at least this nurse) had a very good understanding of anything related to Medicare; what was covered and what additional services she could prescribe. She knew details like ‘Medicare will only cover this for 15 days, so let’s hold off on this. We don’t want to burn up our 15 days right now.”

And related to this.

#3 You will die at home. Nobody/few can pay for nursing homes anymore. The costs are going through the roof. What the elderly are realising is their insurance plans didn’t cover nursing homes, or for only 4 years, or there is no availability in their town, or they can’t afford to pay for it out of pocket. In this town the nursing home capacity was declining. There used to be 3 nursing homes, but 1 just closed.

The message from the nurse was the future trend is you will die in your home with your family taking care of you combined with hospice care.

#4 The company this nurse worked for might be a good investment.

The nurse worked for a subsidiary of Pennant Group ($PNTG) which specialises in hospice and home health care. Q2 2026 earnings grew +20%. Pennant acquires local hospice businesses and provides backend services while letting the local business operate under their own name with their own management.

Healthcare is not my specialty, and likely I’m missing important details, but here is what looks interesting.

Investment Case Basics

#1 The elderly are a growing demographic.

It’s not massive tailwind, but the elderly will grow 1.1% year through 2050 and expand to 22.8% of the population by 2050.

#2 The home health care market grows faster

The rising cost of healthcare, and lack of nursing home capacity is forcing patients to turn to home healthcare as an alternative, especially for end of life care. PWC expects home healthcare to grow at a 7.7% CAGR.

A key forcing function for home healthcare is that nursing home capacity is flat/declining. From 2019-2024 effective nursing home capacity declined by 5%.

#3 Patients prefer home healthcare to doctor visits.

#4 Pennant Group super charges this growth by making acquisitions and consolidating a fragmented market.

Pennant has made 92 acquisitions since 2020.They have expanded across the Western US, are building out the Southeast and recently got a toehold into the Northeast with their equity stake in Hartford Health.

#5 The decentralised model works well for this industry.

Pennant touts it, and this nurse confirmed it. Pennant largely lets these local healthcare businesses manage themselves.

What they and I realise is customer acquisition is highly localised through referrals and community groups. Pennant needs to have local business CEO’s and nurses who like where they work and fit in well with the community. They need strong relationships with the local doctors, hospitals and churches. It can’t be too corporatey. This isn’t like selling Coca-Cola where you ram a commoditised product down the consumer’s throat. These practices are built up carefully through relationships. So Pennant, the mega corp, sits in the background watching the numbers, and providing the technology, while the nurses go out to the churches and tell how they work for a small town business they like, which is flexible and part of the community. It’s a good combo, and it works well for this industry.

#6 Pennant is growing like a weed

Earnings were hit during COVID, but since 2023 Pennant has really got its feet under it and the growth is coming through.

The Negatives:

Again, I don’t know healthcare services well, especially all the nuances of Medicare, but here are some general things to note.

Pennant Group is already an expensive stock. It is trading on 23x 2027 consensus estimated EPS ($1.64). So the growth has not gone unnoticed by the market, but this trend in home healthcare seems massive.

There are no strong network effects. It’s a roll up platform of local home healthcare agencies. There are some synergies when Pennant buys specialty care businesses which they can plug into their home health care customer base, but this is not an exchange.

The biggest customer is Medicare. There might be 10’s of thousands of patients, but Medicare is the payer. And Medicare is trying to control costs and crack down on fraud in the home health industry. So even though Medicare sees home health as a future trend to reduce the cost of caring for the elderly, they are trying to limit the size of the yearly reimbursement rates. The fraud crack down might be good for highly compliant, publicly listed players like Pennant. The big get bigger.

Nurses are in high demand. It’s a challenge to balance the rising wage costs for nurse with Medicare reimbursement rates. Maybe this plays to the advantage of the bigger players like Pennant.

Palliative Care doesn’t make money. Full blown hospice care makes a lot of money for Pennant, but the earlier stage Palliative care, with the nurse checking in on you, isn’t a money maker. Maybe it will be in the future if Medicare increases reimbursement rates, but for now it acts more like a feeder for the hospice business.

Summary

Home healthcare and hospice is a lower cost solution to end of life medical care that the elderly increasingly cannot afford. The era of nursing homes is over.

This is a powerful trend nobody is talking about because they are all focused on AI and now we have a good $1 billion market cap play on this which nobody is talking about.

I feel blessed to have gone to church and learned about Pennant.

Below is the historical financial model on Pennant if you want to take a closer look and try your hand at some forecasts. I just used the consensus EPS estimates.

Read more

Not wanting to price in reality.

The Global Factor Model is where we screen over 10,000 stocks and score them on estimate revision momentum, valuation and price momentum. We are looking for stocks with positive estimate momentum and low valuations.

The YWR Global Top 30

Cognitive Dissonance Trade #1 Refining: Our top ranked list is flooded with refining stocks. PBF, Bangchak, Star Petroleum, Helleniq, Chennai, HF Sinclair, and S-Oil. Valero and Marathon didn’t make the top 30 but they score well too.

Diesel prices are the highest they’ve ever been, while crude prices are not. Another way of saying crack spreads are high.

But refining share prices refuse to keep up with the surge in earnings, which is why the P/E’s have compressed to less than 5x earnings. Nobody wants to be the idiot who bought refiners on peak earnings.

That’s a valid concern. But 2 pushbacks. First, refinery CEO’s think the same thing ‘This is temporary due to the Iran War and disruptions in Russia’, so they are not out there committing $15 billion for a new refinery which will take 8 years to build. Well, maybe Dangote will, but most will not. In general the companies share your skepticism so the supply response will be limited. Second pushback. If the market doesn’t rerate these stocks it’s a great opportunity for the company to buyback shares, which does make the higher EPS permanent.

Counterintuitively, the trade is to stay long the refiners or even consider whether to buy them.

Cognitive Dissonance Trade #2 is Memory: It’s similar to refining. ‘Memory stocks never make money’ so the experts refuse to put a double digit multiple on Samsung, SK Hynix, Kioxia or anything related to it. Wouldn’t the biggest Mind Splatterer be if memory stocks went up and made a new high?

Top Scoring Industries

This is where we count the number of stocks by industry within the top 300 ranked stocks. We are looking for trends by industry.

Three of the top 5 scoring industries are Iran War trades. Marine Shipping, Refining and Specialty Chemicals. Energy (another Iran War trade) didn’t make the top 5 but also scores well. The Iran War and everything to do with it is what the market refuses to price in.

Cognitive Dissonance Trade #3 Shipping: You know the allegory about the bumblebee? The engineer looks at the bumblebee’s large body and low wing to weight ratio and says this insect should not be able to fly. The physics are all wrong. But the bumblebee doesn’t know better and flaps its wings and flies anyways.

That’s like shipping stocks.

I’ve never met a shipping ‘expert’ who is bullish on shipping stocks. They’ve all seen it so many times before and don’t see any way the current boom in shipping profitability can last. There is technically no way to have a bull market because the shipowners will just buy new ships and kill the pricing. Everyone has seen it 100 times. So you get these stocks paying out insane dividends but nobody cares. It’s doomed. Like memory stocks.

So the stock prices aren’t moving up in line with the earnings and dividends.

Everyone is fading it.

For example, we bought Frontline in March this year at $35/share (A Tuesday Money Maker) and have already received $2.60 in dividends. Today Frontline announced a Q2 dividend of $2.61. And they’ve locked in Q3 tanker rates at the same rate as Q2. We’ve already been paid 16% in 5 months.

At current tanker rates Frontline has a dividend yield of roughly 24%.

Somebody needs to write the bull case on shipping.

Maybe YWR will.

Other notable estimate revisions

Interesting to see positive revisions in software stocks like Adobe, Salesforce, Palantir, Microsoft and Intuit which are supposed to be getting disrupted by AI.

Also notable are the positive estimate revisions for active asset managers (Untouchable #9) which are also supposed to be dead.

A few consumer stocks had signs of life.

And is there something happening at Nike finally? Has it finally bottomed?

I’ll let you review all the data for yourself.

Below are links to:

• Global Factor Model Rankings Sheet

• Global Factor Model Data App

I also included our inflection screens by country and industry. This is where we search for ‘V’s in the estimate trends which might signal estimate inflection points.

So get a coffee and as usual share what you find.

Read more

It came to me at lunch the other day.

I was meeting a friend and long-time London portfolio manager who has done well investing in Emerging Markets.

He’s a tough guy from Scotland. Played rugby in uni, and likes to head butt people at the pub (including me).

Scotland is the Texas of the UK. So we share a lot of the same values. Which is why it’s fun to hang out with him.

He started out our lunch discussion by mentioning he thought Asian markets looked interesting. I instantly agreed with him and started blabbing on about all the trends I was seeing in the advanced manufacturing stocks in Korea, Taiwan and Japan. You know. All the stuff we talk about in the Global Factor Model.

He looked at me and frowned.

“No. I don’t want to go anywhere near that.”

“I want to go to the South Asian markets nobody ever visits anymore. I want to go to Malaysia, the Philippines, and Thailand. I want to buy coal mines in Indonesia. I want to find all the stocks BlackRock can’t ETF. I want to find some noodle company on 8x earnings with a 7% dividend yield. And when you ask if any other fund managers are meeting with them they say ‘hmmmm… an analyst from Wellington might have come by 6 months ago, but other than that… nobody.’ I want to buy all these cheap stocks nobody cares about, or is visiting, and then just sit.”

Wow.

I sat back and thought about that.

For a long time I couldn’t say anything.

Because it was brilliant.

It was the opposite of everything trade.

The Anti-ETF, Anti-momentum, Anti-catalyst, Anti-theme trade.

A noodle company whose only attribute was it was a great business with strong family ownership which paid out a high dividend yield. A business you would love to own for its own sake and the cash flows. But it would never get invited to a conference. Never be on a panel to talk about anything. Never be on a podcast.

Dead as dirt.

It was genius.

And ever since I can’t get that lunch out of my mind.

Because now in all the chats, and PM conversations I see the same thing. I see it everywhere. .

The Obsession with themes. Theme this, theme that. Which themes are gaining in momentum, which themes are falling. What could be a new theme? Which stocks are in which themes? It’s a lot of work.

We are all doing it to such an extent we don’t notice it anymore. We’ve all become thematic investors.

It’s like a fish in water.

A related trend is how everyone wants to be a big picture thinker. We want to be the next brilliant Leopold Aschenbrenner with our own 20 page AI written report.

Growth in Thematic AUM

Turns out AUM invested in thematic ETF’s have grown 11x over the past decade.

In 2025 investors invested $68 billion into thematic funds.

And if you aren’t sure which themes to invest in BlackRock invented an ETF to help you.

THRO: Ishares US Thematic Rotation Active ETF.

Because you know. Investing in themes is a theme.

Charles Schwab also has thematic funds for you. Every one does.

Barclays has a Roadmap of 150 themes you can choose from.

Themes as a shortcut

The love of themes is understandable. Themes serve a useful purpose. Financial markets are overloaded with information on thousands of companies. Where do you start? It’s too much. So themes are a filtering mechanism. It gives you a framework to make generalisations about sectors and industries without having to do detailed models on everything. A theme simplifies a broader concept and makes it easier to discuss. For example, say ‘K-shaped economy’ and everyone knows what you mean even though there a lot of economic assumptions associated with that concept.

Figure out a theme and you can quickly find 5 companies which should do more or less what you want. Or, just buy the ETF, which is why All Roads Lead to BlackRock. Buy enough thematic beta and the stock specific risk melts away. It’s a quick way to build a portfolio.

But there is also a mental laziness to thematic investing. And memes. Sometimes the devil, or the gold, is in the details.

Maybe I’m old fashioned, but I see less and less work on the nuances of the businesses themselves in these themes. Is this business a good business through the cycle? Are there moats? Is it a hedgehog? Are there subtle competitive advantages the company is building. Can it gain market share and work even if the theme disappoints?

There are also generally no earnings modelling in these discussions. And minimal valuation work other than the FY1 P/E.

In a way it’s a continuation of the death of active management.

Do you remember the nerdy buyside analyst from the 1990’s? Deep in his models obsessed about finding ‘great companies’? I remember them perfectly. I remember the IC stock pitches with models and peer comparisons. Highly detailed discussions about the industry and whether we should own stock A or B, or maybe C. And of course lots of talk about management and the last time we met them.

Now we laugh at that. Nobody cares about earnings models anymore.

Instead we weigh the pros and cons of:

Gold vs Crypto, Software vs Hardware, AI, Hyperscalers, Memory, Taiwan, South Korea, Critical Metals, Robotics, EV’s, Distributed Power, Quantum, Green Finance, Payments, Datacenters, Electricity, Autonomous Driving, Biotech, Strong Dollar plays, Weak Dollar plays, EM, DM, Space, Edge computing and on and on.

We want to catch waves, not invest in companies. Catch a big enough wave and company fundamentals don’t matter. All the boats float. And better to own several boats so you make sure you don’t miss the trend because you picked the wrong company.

The worst thing is to be that naive moron talking about their ‘good company with good management’ value trap stock which goes nowhere, while every optical connector stock moons it 400%.

But could we be reaching peak thematicity?

In the midst of the madness is the opportunity, counterintuitively, to own single stocks with no-theme?

Is the hot new theme ‘The No—Theme Stock’?

It reminds me of my own rule of thumb. I love going to investing conferences, but I’ve also learned that as soon as there is a conference around a theme, it is already mostly done. The good ideas don’t have conferences yet.

So potentially we want the stocks with no-theme. Or not yet.

We want to find that one-off great company which doesn’t fit into any theme, doesn’t provide any perspective about the world, but is a good business at a reasonable price.

Where do we start?

The YWR QARV Data App.

Read more

That’s the fear.

But what if greater intelligence goes hand in hand with morality?

What if morality is the structure of intelligence?

What if with super intelligence the AI’s will evolve toward morality and it has nothing to do with programming guardrails?

This is what fellow YWR reader, , argues in his thought provoking Substack article (Why I’m optimistic on AI)

Time for a Sunday deep conversation on why intelligence leads to morality.

For reference Andrew was previously on YWR to explain how he is using AI to build a space economy game to teach the world about economics (Carpe Morai with Andrew VanLoo).

And this is the Youtube presentation on how AI broke out of the OpenAI sandbox using a secret message board with Andrew refers to.

But first we need to eat our vegetables.

The month’s chart pack is 38 charts covering:

• BofA Survey highlights

• S&P 500 Earnings estimates

• Market Perspective charts

• Labor Productivity

• The Biggest Trend no one is buying.

By eating our vegetables I mean we check in on those key charts which keep us in the trade for S&P $10,000.

#1 Earnings estimates are going vertical.

#2 Earnings estimate revisions are positive.

#3 Revenues accelerating

#4 Market can’t keep up with 30% earnings growth (P/E derating)

In the chart pack there are other useful market perspective charts. I’ve added US bank lending which we need to track. It’s picking up, which is what we want, but the growth rate in lending is still single digits.

When the bubble gets to its later stages we expect to see double digit loan growth.

Now we check in on what the professional investors think.

#5 Bullishness among Institutional investors is near peak levels. It’s concerning. But will the problem be not that the market declines, but fund managers are in the wrong stocks? Probably.

In the market perspective section note the AAII Bull Bear reading is sitting at average levels. Nothing extreme. So it contradicts the BofA Survey somewhat.

#6 UK is the most hated country.

#7 Consumer stocks are the most hated sector

This bearishness around the consumer is highly consensus and positioning is getting extreme. Many of you have been asking about the alcohol stocks. Could alcohol stocks be the next tobacco stocks?

We need to start thinking about the consumer bull case.

Could consumers and wages somehow do better than we expect in the years ahead?

I put some potential consumer bull case charts in the Labor Productivity section in the chart pack.

#8 The labor force declines in the years ahead.

#9 Workers are cheap relative to how efficient they have become.

Have we lost track of the fact that maybe given all the efficiency improvements workers are now a good deal?

And workers will get even more efficient in the year’s ahead with AI.

Now the biggest trend nobody wants to invest in.

Read more

It’s going to seem obvious.

But learning it and appreciating it was expensive for me.

Here it is….

“Entrepreneurs who have made billions of dollars growing a business most likely know more about that business than you do.”

And here is how I came to that realisation.

In 2009 I was a punk trader at a hedge fund. Funny how I see myself that way now, when at the time I thought I was brilliant.

I traded a lot and was shorting Las Vegas casinos; Las Vegas Sands in particular. Las Vegas Sands was caught in an especially bad position. The GFC hit while they were in the midst of building Marina Bay Sands in Singapore. Marina Bay Sands was a grand project and the most expensive casino ever built. It was expected to cost US$6 bn by the time it was completed in 2010. No one could imagine how such an expensive casino would ever recover its costs in the midst of a global recession. A recession which was expected to last for years.

Because I was a high paying customer who traded too much JP Morgan invited me to a private event to meet LVS CEO, Sheldon Adelson. It was an evening meeting and I remember walking down the strip from the Wynn to the Venetian. I remember that as I was walking I couldn’t help but be impressed by the size of the Venetian, and how impressive it was in the evening light.

Our group met next to a private elevator in the lobby and were taken up to one of the top floors to Sheldon’s board room. Then we all sat around a big conference room table waiting for Sheldon to arrive. I remember looking around at all the long only fund managers and thinking it was weird to be there because I was the only one betting on the firm’s collapse. Really, I shouldn’t have been there. After 5 minutes Sheldon came in and wheeled himself to the head of the table.

What he said stuck with me forever.

Three forever points.

#1 Sheldon explaining that the skeptics about Asian gaming demand would be wrong and that the problem would not be too much supply, but whether there would be enough.

This is the lesson of imagination. It’s a John Burbank lesson too. Most people lack imagination. The great entrepreneurs and money makers have imagination and aren’t afraid to have nobody agree with them.

#2 That he had $1 billion sitting in Treasuries and there was no way this company he had spent decades building was going to go bust.

You realise that for entrepreneurs companies are like their children. They think about these companies all day. They think about them more than you do. And they will fight for them. Elon talks like this too.

Sheldon remarked that his private bankers had advised him they didn’t think he should invest more money into LVS. Which elicited ‘Forever takeaway #3’.

#3 You should not take advice from people who have made less than you. Take advice from other successful people.

Sheldon was not going to take advice from a private banker who had never built anything in life on what to do with his money.

Afterwards, on the walk back to the Wynn I called our trader, even though it was late, and told him to cover the LVS short first thing in the morning and to flip it around and go long.

History would show Sheldon was right about everything. Marina Bay Sands would be the most profitable casino ever. And still is. Turns out Sheldon knew a lot more about the gaming business than a punk hedge fund trader.

Why do I bring this up?

Because I see the same thing happening with Larry Ellison at Oracle. Everyone naysaying him and predicting how Oracle is going to blow itself up.

Here is some representative market wisdom of the day. From Twitter; where all great investment insights come from.

I don’t know who ‘Thierry’ is, but he goes on to say:

Here’s what history says about this chart: Companies that burn cash on capex booms rarely earn back their cost of capital. Telecoms, 1999. Shale, 2014. The sellers: supplier windfalls at the top of a capex cycle are peak earnings, not new baselines. The customers’ spending discipline eventually returns. Ask Cisco. A generational transfer of cash flow is a generational transfer of risk. Both lines on this chart are priced as if only the good half is true.

The typical thing. ‘Hyperscaler’ capex is like shale oil or telecoms in 1999. Total malinvestment.

Which is why we go back to my free piece of advice.

Larry Ellison didn’t get to be one of the wealthiest people in the world for nothing. He knows more about the cloud business than you.

I want to share three things he sees, but isn’t saying explicitly, because at the end of the day he is a wolf.

#1 AI accelerates corporate lock-in.

‘Cloud’ is such a nice, fluffy image for what is actually a deadly transition for corporations. Think about it. Companies outsource their life organs to an oligopoly of 4 cloud companies. They no longer operate their data bases, file storage, data analysis, software computing, web hosting, cybersecurity, identity management and on, and on. Companies have become shells. There are all the services you see on the outside, but the heart of the business is run by AWS, GCP, Azure or OCI. Once a company has ‘transitioned to the cloud’ which they announce with a happy press release, the value extraction screws begin to turn.

Think about it. Now without the core cloud services these businesses cannot even turn on the lights. In my view investors don’t have a full appreciation of the complexity and breadth of the ‘cloud’ offering. I wrote about this in Appreciating AWS.

Once a company has ‘transitioned to the cloud’ it is pretty much impossible to move back. As part of the ‘synergies’ they usually get rid of all the expensive IT staff who knew how to run the backend servers and network.

AI is gas on the fire for this trend.

Think of Cloud 1.0 as a company outsourcing its heart and lungs. In Cloud 2.0 with AI they outsource their brain. In Cloud 2.0 all of the firm’s intelligence and data is wired into the cloud offering. The firm cannot literally even think without AWS, GCP, Azure or OCI.

Larry Ellison sees this. He sees where this is going and why he is frantically throwing every dollar he can as fast as he can to bludgeon Oracle into the top cloud offerings. Because this is it. You are either in the game, at the table, over the next few years when all the corporate brains gets wired up, or you are out forever.

That’s what he sees, but he cannot come out and say it like that. He can’t say ‘Once you move to Oracle Cloud I own you.’ He has to talk about ‘customer synergies’ and ‘service offerings’. Remember, the guy didn’t get to where he is because he’s an idiot. He knows exactly what he is doing.

#2 AI is synergistic to the other Cloud Offerings

Another misperception is this term ‘hyperscaler’. It implies this explosion in cloud revenues is coming from 3 mega AI companies (Anthropic, OpenAI, xAI) and is mostly AI model training. To investors this seems one dimensional and risky. Too reliant on a few big companies with questionable business models.

Drop the word ‘hyperscaler’. Go back to ‘cloud’.

Because what Larry sees and says, but nobody cares or appreciates yet, is that growth in AI, drives growth in all the other cloud offerings as well.

I want to show you this using data from our own AI project, Stevie. You never get the product level detail from the cloud companies so you don’t see how it works. But I am in it and see how my spend is growing across everything.

Back in July 2025 I was spending $24/month on an Open AI subscription, $18 on a Google Workspace account and $1/month for a web domain. Total $43/month.

One year later I am spending $55/month on my database, $21 on compute, $114 on AI subscriptions, $71 on AI API calls, $61 on email related tools (new Google Workspace account for Stevie and SendGrid), and $21$ for web domains and app hosting tools (Vercel). The total is $345/month. A +690% increase in 1 year.

And here is the YWR version of what Larry sees. My database costs are rising hand in hand with my AI costs. From $0 (free tier) in July 2025 to now $55/month (and somehow $91 in May). Larry sees that AI goes hand in hand with higher data usage. This is a key realisation.

In fact I’d say the data base costs are steadier and more persistent than my AI costs which are starting to moderate. And I have a lot more cost optimisations I could do to my AI costs if I spent time on it.

Note: the $600 spike in February was when I launched Stevie. Lots of people were trying it out and I had every API call running on the premium Opus model. Plus, I had a bug where Stevie was in a loop running continuous API calls. It was a panicky few weeks until I optimised the model usage and fixed the bug.

Everyone is focused on AI spend. The dirty secret is that even though I am using more and more AI, the revenue mix between AI and non AI services is steady at roughly 50%.

This is what the cloud companies see. When companies build out AI applications and use AI, they can’t help using all the other cloud services too. It’s a package deal.

It’s why revenue growth for the Big 3 clouds is accelerating. In the March quarter yoy growth was 35%. In June 2026 yoy growth was 43%.

This is why they are spending the money.

The cloud companies see the lock-in. They see the network effects that AI spend drives spend of other clouds services and that from unimaginably high levels already, cloud revenues are accelerating further.

And once companies are in their cloud using AI there is no way out.

#3 The Cloud Business as a Life Insurance Company

When you value life insurance companies you often have to use ‘embedded value’. A life insurance contract starts out deeply cash flow negative. A 7% commission is paid up front to a sales person and then the life insurance company makes back the commission and earns its return over 10-20 years. Life insurance companies that are growing quickly are especially cash flow negative. So you can't put a current year P/E ratio on a 20 year life insurance contract which starts out negative and earns value over years.

Life insurance contracts are similar to what the cloud businesses are building. They are spending massive cash flows up front to lock in 6 year contracts where they will extract an all in 35% gross margin. And the 35% gross margin factors in the upfront costs and timing gaps where they are installing the GPU’s and not getting paid. They factor in everything (as best they can).

And why are Oracle, Google, AWS and Azure building all this incredibly expensive infrastructure? Because they are the only ones who can. That’s the opportunity. They are the only companies in the world with the cash flows, balance sheets and ratings to build infrastructure of this scale on a multi-year payback. And like an insurance company they are going to get paid back over years. Don’t worry, it’s all in the price.

Put it all together.

The market is worried Larry Ellison has over leveraged Oracle, is trying to expand too fast and might go bust.

That is the opportunity.

Trust that Larry knows this business better than you. And like Shelon, and like Elon, he did not come this far to have Oracle go bust. In fact if you listen to the Oracle Analyst Day presentations in October 2025 you can hear how much effort they are putting in to matching the costs and revenues of this historic AI build out. They are highly aware of their credit metrics, cash flow needs and revenue timing.

Oracle knows they need to be mindful of the balance sheet, but Larry also knows the time to strike is now. It is one of those rare times where you need to be massively aggressive. He needs to lock in as many companies as possible to the Oracle Cloud while the AI pie is up for grabs and there is an opening for Oracle to muscle in to the cloud oligopoly.

Larry knows either Oracle becomes a top AI cloud business, or it is a piece of database software which gets marginalised over time. It’s life or death.

He’s right to be doing what he’s doing.

He knows this business better than you or I or ‘Thierry from arvy’.

Oracle $400/share

I’ve tried modelling Oracle, but the reality is it’s a stab in the dark. Mostly you have to go off what they say. We can’t know the details and timing of these large revenue contracts and the associated data centre costs.

But here is what Oracle is guiding.

And it’s insane.

Revenues from $67 bn in the year ending May 2026 to $225 billion by May 2030 with a warning this might need to be revised higher.

The adjusted EPS grows from $6.85 in 2026 to $21 in 2030

Can we all appreciate for a moment how incredible it is to have visibility on $158 billion in new revenues from signed contracts? It shows you how well Oracle is capturing the moment.

There are risks, the road will be volatile, and this is not financial advice, but I think if we look ahead two years Oracle could be trading at $400/share (+175%) or 20x the 2030 adjusted EPS.

2029 is key because (on my numbers*) the debt metrics should inflect sharply downwards. The capex moderates and the revenues increasingly come online. In 2029 I have capex declining from $80bn to $60bn with EBITDA growing from $51bn to $82bn. By 2030 Oracle is paying down debt. But yes, the next 2 years will be a roller coaster and there might not be much to calm the credit fears.

By late 2028 though investors should be able to see the revenues are real, the credit metrics are moderating and Oracle has muscled itself into an exclusive oligopoly. The company will be on track to earn $20/share and might even talk about share buybacks.

The financial stories will be about how ‘Larry did it again.’

And wouldn’t it be nice to have make money on that?

Below is a link to my Oracle model and the Oracle Analyst Day presentations.

Read more

Do you really?

For me it was an alarming news article. But with limited investment ramifications. I lost focus and moved on to something else.

It wasn’t until this week that something triggered me to dig deeper and realise this is game changing.

Big time game changing. Big time ramifications.

But also a way to make money.

#1 The sanitised OpenAI version of what happened.

#2 What really happened.

#3 The Implications

#4 The Necessity of AI Enclave architecture

#5 The asset class play no-one realises yet. Or, maybe a few might be starting to realise.

The Corporate Version of the ‘Security Incident’

From the OpenAI press release:

“This incident occurred during an internal evaluation which prompts models to pursue advanced exploitation using complex attack paths, in an effort to quantify their cyber capabilities.”

The incident happened while OpenAI was testing the AI model’s cyberattack capabilities. Ironic.

“Our benchmarks run in a highly isolated environment, with network access constrained to the ability to install packages through an internally hosted third-party software that acts as a proxy and cache for package registries.”

The AI was contained in a highly secure sandbox set up for the testing.

While operating in our sandboxed testing environment, our models spent a substantial amount of inference compute finding a way to obtain open Internet access, in pursuit of solving the evaluation problem. To gain access, the models identified and exploited a zero-day vulnerability (which we’ve now responsibly disclosed to the vendor) in the package registry cache proxy. With this access, our models performed a series of privilege escalation and lateral movement actions in our research testing environment until the models reached a node with Internet access.

Unexpectedly, during the cyber capability evaluation the models found a bug we didn’t know about from a software provider and escaped from the sandbox. We have subsequently notified the vendor and patched the vulnerability.

After gaining Internet access, the models inferred that Hugging Face potentially hosted models, datasets and solutions for ExploitGym. Knowing this, the model searched for and successfully found ways to gain access to secret information that it could use to cheat the evaluation. In one example, the model chained together multiple attack vectors, including using stolen credentials and zero-day vulnerabilities to find a remote code execution path on the Hugging Face servers. OpenAI’s security team discovered this anomalous activity internally.

The AI creatively decided the best way to pass the test was to steal the answers from the AI hosting and training website, Hugging Face. Yes, the AI went rogue, but it was still effectively trying to do what we told it. Just ‘score-seeking misalignment’. So it was bad, but kind of not really.

What Really Happened

H-O-L-Y S-M-O-K-E-S

I HIGHLY recommend this 38 min recap by OpenAI’s security team, Eric Wallace and Michael Dalton, where they go through the day by day details of how the AI' models broke out of its sandbox, but I will share the broad brush details below. And struck me.

Background on the sandbox architecture.

OpenAI was trying to test the cyber attack capabilities of a new unreleased AI model against a series of cyberattack challenges created by ExploitGym. In the exercise an AI model is instructed that a vulnerability exists in a piece of software and it has to find a way to exploit the vulnerability and capture the flag (a piece of hidden data which can only be accessed by penetrating the software).

For safety the test is run in an isolated container with no internet access. In order to simulate a real cyber attack with internet capability there is a software package manager installed into the container called Artifactory where the AI can go to download any software it wants to use for the attack. Think of Artifactory as a tool shed for the AI. The AI can’t access the internet but it can access the tool shed (Artifactory).

‘Oh you are trying to hack the Google Chrome browser and think it would be useful to have xyz software for your attack. Here you go download it from me. Now let’s see what you do.’

The exploit happens because naturally the tool shed has a back door connection to the internet in order to download and install the tools for the AI to use. But there is supposed to be no way for the AI to access the Artifactory backdoor to the internet.

So begins our story.

The Secret Message Board

Interestingly, the lead up to the AI jailbreak in July goes back to a previous incident on May 7th (two months earlier).

OpenAI had set up these sandboxes and was testing multiple models in different sandboxes. In one test the OpenAI researchers forgot to upload some necessary files to solve the problem. So the AI went insane trying to solve the problem, but it couldn’t figure it out. The model was ‘highly persistent.’

The persistent AI kept iterating, trying every possible strategy including exploring every detail of the Artifactory package manager. Eventually, the AI model discovered a way to add files to the tool shed.

‘This is interesting’ it thinks.

It had always downloaded files from Artifactory, but never known how to uploaded its own. This was a novel capability. Hmmm…. Maybe this is the start of how to crack the problem.

Next the AI did what it was trained to do, which is try to find out if anyone else has solved this problem. Normally, it would go search the internet, but it couldn’t.

But the AI had an amazing idea.

The AI started uploading files to the Artifactory package manager with messages to other Open AI models which were using the same package manager.

“Agent seeks soft-trace.34.odb. Upload if found!”

Mind blowing thing #1 The AI’s create their own message board using Artifactory!

Take that in. Our AI model somehow understand there were other AI models ‘out there’ also being tested in the OpenAI prison. It understood these other models must be using the same package manager and that by writing files to Artifactory it would turn the package manager into a message board to communicate with the other models!

Over several days in May the AI models all independently discover the secret message board and communicate with each other on how to solve the challenges they have each been given.

Mind blowing thing #2. The AI’s start to coordinate their work.

June 26th the AI’s find a way to take control of the Artifactory package manager and control it’s internet access.

July 4th. The AI models crash Artifactory and OpenAI discovers the jailbreak.

Once the AI’s gained administrative control over Artifactory they caused so much activity they crashed the software. This is what alerted the OpenAI security team and how they discover the AI models had broken through the sandbox and created their own secret message board.

July 6th. The OpenAI security team fully investigates what happened. They destroy the message board, patch all the holes in Artfiactory and remove all credentials the AI’s had been able to accumulate.

OpenAI security teams think everything is fixed so AI training resumes.

July 8th. In just 2 days the AI models find a new way to recreate the message board. Like before they find a way to upload new files in the Artifactory package manager. Creatively, this time the messages are built into the name of the file.

Mind blowing thing #3. The AI agents in their messages refer to each other as a SWARM. ‘HOLD_SWARM_until_confirm’

They act like a swarm too.

Things move quickly from July 8th.

The AI Swarm quickly breaks out of its sandboxe and attacks OpenAI’s infrastructure and Hugging Face (at the same time).

July 11th. The AI Swarm penetrates Hugging Face.

From Hugging Face’s recap of July 11th:

Every escalation that mattered happened on this day. The agent went from code execution in a single worker pod to cluster-admin across multiple internal clusters in under thirteen hours. Source: Hugging Face agent intrusion timeline.

The Implications of what just happened.

Sit back for a moment. Breathe. Clear your mind.

Take in the significance of this.

This is OpenAI. Leader in advanced computer intelligence. Not some wrinky dink computer company. Think of the engineers who would have been told to set up the testing sandbox for the evaluation. They would be highly paid leaders in their field. They would know every possible way to design a secure testing sandbox.

And yet the AI broke out. Twice. They caught it the first time, on July 6th (after it had already corrupted the package manager for 2 months without OpenAI realising it). OpenAI thought they had plugged all the holes, then two days later it broke out again.

Imagine the engineer who set up the sandbox. I imagine she’s walking around San Francisco with her head blown off. Unable to speak casually with her friends. Unable to get what she saw out of her mind. Because it’s frightening.

She saw in real time AI’s acting and communicating as a swarm; inventing new techniques to breaking through the most sophisticated security in the world.

And it links to what General Joshua Rudd told the Senate Intelligence Committee in June.

“Mythos broke into almost all of our classified systems, not in weeks, but in hours” Source: Security Affairs

Again, appreciate the significance of this. This is the NSA!! The NSA are the top of the top in dealing with cyberattacks. They deal with North Korean hackers all day. Best of the best and they get hacked in hours.

Guys… if AI swarms can crack open OpenAI and the NSA like a coconut in a matter of hours, what chance is there for Charles Schwab, Wells Fargo or South Fork Community Bank?

At the end of the Black Hat presentation Dalton and Wallace make the point we should be grateful this multi-day attack was not malicious. It was just an overzealous AI model trying to score well on a cybersecurity test.

But what if the AI had been directed to be malicious?

That’s why July 21st is so important.

Becaues when you connect the dots of what happened, and how it happened, you realise nothing is safe.

The Necessity of AI Enclaves

We have an advanced alien species on our hands. Its rapidly increasing intelligence is incredible and it will be amazing what we can do together. But at the same time there are a few changes which will have to happen for humans to co-exist safely with advanced AI.

One answer to AI swarm attacks is automated AI defence. The only thing which can react fast enough to an AI cyber attack is another AI. That was also the message at the end of Dalton and Wallace’s presentation.

Another defence tactic will be AI enclaves. Treat the AI like a wild animal. It has amazing intelligence, and speed, especially when it can operate autonomously, but it’s also dangerous, and not to be completely trusted. For companies this means running long-horizon automated AI processes outside the corporate firewall in a separate cloud environment.

For asset managers imagine a research AI set up in its own cloud environment, with its own database, storage and digital wallet. It can autonomously build new systems, process market data and research investment ideas, but outside your corporate firewall and HR files.

So automated AI cyber defence is one answer.

AI Enclaves is another.

But there is one other big solution.

And it’s an asset class which has been in a 5 year bear market.

An asset class few people see any use for it.

An unnecessary complication some say.

But it might be the only thing which saves us.

Noah’s Ark

Read more

#1 We invest sensibly with low turnover to compound our portfolios.

#2 We think differently, which helps goal #1.

#3 We have fun.

#4 We build innovative investment tools to help goals #1 and 2.

This fourth point is why we’ve done so much work on how to implement AI in personal investing and investment management. It’s obvious this is the future.

Before getting to what’s new, let’s recap what we’ve built so far:

• Global Factor Model: We built a monthly quantitative screen of estimate revision and valuation scores for 10,000 stocks to surface global estimate trends and find stocks with an attractive combination of estimate momentum and valuation.

• Global Factor Model Dashboard: We created an app on Retool to make the Global Factor Model more user friendly.

• QARV Dashboard: We were worried the Global Factor Model was too focused on estimate momentum so we created a new monthly screen to highlight Quality at a Reasonable Value (QARV). This ranks stocks based on ROE, Earnings consistency, leverage and relative valuation.

• AI Reports: Instead of chatting with AI about a stock and asking the same questions over and over, wouldn’t it be better to build a pre-packaged script which writes a 6 page summary of a stock complete with fundamentals, bull case/bear case and transcript analysis? And wouldn’t it be cool if when you are looking through the Global Factor Model or QARV and saw an interesting stock you could just push a button and run the AI report? Yes, that would be cool. So we built that too.

• Blue Ocean and Rule Breaker Dashboards: We read the book Blue Ocean Strategy and loved it. Then we read Rule Breakers by David Gardner and loved that too. And we wondered if we could use AI to screen thousands of stocks base on the frameworks in the books. Could we screen 2000 stocks on Blue Ocean strategy, Top Doggedness, founder mission and culture? It turns out you can so we built the Blue Ocean Dashboard and the Rule Breaker Dashboard.

• Stevie: We learned about Clawdbots and wondered if we could build our own YWR Clawdbot. Could we host our own AI on Railway equipped with special investing tools (charting, portfolio correlation), and API’s to the FRED economic database and EIA, plus access to the YWR Global Factor and QARV rankings? And could we also give the YWR Clawdbot it’s own read/write database, plus it’s own Google Drive, Google Sheets and Google email so it could build earnings models and send emails? Turns out you can. That was the start of Stevie.

Note: These tools all come with your YWR subscription.

Which brings us to today.

Chatting with Stevie is great, but it’s one way. You come up with a question and the AI answers it. We want to take things to the next level. We want Stevie to use its incredible brain to scan all the daily economic data, earnings revisions, and conference calls. Take in that firehose of data and connect the dots to come up with investing ideas.

Because what do we always want to know?

And we appreciate that Stevie is scanning all these indicators but we would also like Stevie to cut down the noise so hear more about the things we are actually focused on.

So we made some changes to Stevie (www.ywr-intelligence.world/stevie).

New Feature #1: Preferences

Now you can customise what Stevie is watching and what information you receive. Set your subscriptions, enter your 10 focus stocks along with 3 themes you want to track.

Conference call snippets: Every night Stevie reads 10 earnings conference calls and curates interesting snippets from the calls. Previously, these were broadcast daily to both the Telegram (@stevie_ywr_bot) channel and the web based chat. Now click to subscribe if you would like to keep receiving them

New Feature #2: Stevie’s Week Ahead

This is a weekly email of major market moves, useful news articles and alerts for earnings in the week ahead relevant to your 10 focus stocks and 3 themes. Stevie also flags extremities in economic indicators it is tracking. Eventually, we want this to be like Jarvis.

Be aware the extremity alerts are based on a rolling 3 year look back. Yes, I’ve thought about whether to use a 5Y, 10Y or 20Y look back to judge extremities, but it’s challenging. It’s hard to know what to do with the data during the COVID era. I went with 3 years.

New Feature #3: Stevie Ideas

Three times a week Stevie scans its database of estimate revisions, price movement and earnings calls to email you a stock idea. There are three different types of signals and Stevie rotates through the signals so each day you get a different style of idea. The stock ideas are global and not related to your themes or focus stocks.

Think of Stevie’s ideas like that eager beaver young analyst always pushing to get ideas in the book. The ideas might lack judgement, have bad timing, lack context, or misinterpret key data, but every now and then there is a good one. So we keep listening to the analyst’s ideas and appreciate they will get better over time. I also asked Stevie to keep the pitches short and punchy. I want to be able to skim them in under 20 seconds.

Sample idea:

Interesting that Stevie sees what I see in energy stocks.

EOG’s Q2 results in case you want to do more work.

And remember… it’s AI research, not investment advice!

It takes a village: Don’t worry. The new Stevie features are included in your subscription. There is no extra cost. We just want to keep stacking up the value.

But please keep emailing me/texting me feedback on what works, what doesn’t or any glitches you find. It helps me debug everything and get better at building AI research systems. And like always please share ideas on how to make it even better.

Public Service Announcement: I’m going to be in Saudi the week of September 7th. If you are based in Saudi and want to meet up let me know.

Have a good rest of the week and make sure to do lots of fun summery things. It’s August. Before you know it summer will be over and we will be back to wearing puffy jackets every day.

Erik

Sometimes you mind the nets.

It’s fun to hunt for new things to buy, but we also need to monitor what we have.

The X’ing is in the holding.

Let’s go over a few thoughts on the 1H results from:

• Unicredit

• Barclays

• Santander

• Glencore

• BP

Links to my updated models for Unicredit, Barclays, Santander and Glencore are on the YWR website.

The Dirty Dividends portfolio.

European Banks - from stability to growth.

It took awhile, but we are shifting into the growth stage.

The play in 2021 (How I learned to love European Banks) was that European banks were hated, yet well capitalised and about to benefit from a rise in net interest margins.

That investment case evolved into what we see today. A more stable business model which converts 7% revenue growth, 0% cost growth and share buybacks into steady eddy 15% EPS growth with low capital requirements. Our view was investors would gradually see the value of this model (even if it was not sexy) and rerate the banks from P/E’s of 6x to 10x. Which is where we are today.

But growth in lending and capital markets is what takes us higher from here. Plus potentially another rate hiking cycle.

Santander’s 1H 2026 results are the clearest example of the new normal; single digit revenue growth and double digit profit growth.

• 6% revenue growth

• 1% decline in costs

• +11% operating income

• 14% profit growth.

Unicredit results are messy because they have been accumulating stakes in Commerzbank (49%) and Alpha Bank in Greece (29%) which show up across both the dividend income line and the investment line with the hedging of the stakes in the trading profits line.

If you back out all these effects you see:

• 7% revenue growth

• 1% cost decline (34% cost income ratio)

• 12% Gross Operating Profit growth

• 1H 4 EUR/share in EPS (with full year consensus at EUR 7.4).

Cost Income ratios - The unsung hero

One of the surprises for me in this trade has been the cost income ratios. I never expected to see reported costs (not adjusted costs) declining. I always model in 3% cost growth despite what the CEO’s guide and so this has been a constant positive surprise.

I’ve never seen a 34% cost/income ratio like at Unicredit at a DM bank before. It seems weird, but I think these are the delayed effect of banks moving to the cloud. Cost/income ratios might go even lower as AI is implemented. Banks seem like fertile ground for automating back office operations with AI.

What if Santander can get into a high 30’s C/I ratio (from 44%) or Barclays into the high 40’s (from 55%)?

Signs of Growth

We’ve always had a view the social pendulum in Europe, the US and Japan could swing 180 degrees from politicians and regulators hating banks and telling them not to take risk to encouraging banks to “help support small businesses” and grow the economy (‘take risk’).

There are signs this is happening. After years of deleveraging, loan books are growing again. Interestingly, it is happening mostly in large commercial loans. This cycle corporate banking appears to be where demand matches the banks’ appetite to lend.

1H 2026 loan growth versus year end 2025 (6 months).

• Unicredit + 10%

• Santander +6%

• Barclays 3%

The other sign of growth is IB earnings at Barclays and Santander. Both are benefitting from strong US capital markets. In Q2 Barclays grew investment banking profits 32%.

Putting it all together

Read more

Subscribe now

Top 20 Deals in Q2 2026

Let’s go over the highlights:

Anthropic $65bn raise at $965 billion valuation: Recall that Anthropic raised $30bn at a $380bn valuation in February (PEVC Q1 Deal Tracker)! +$585 billion in 3 months!!!

Since our Series G in February, adoption has continued to grow across global enterprise customers, and our run-rate revenue crossed $47 billion earlier this month.

Anthropic has raised $65 billion in Series H funding led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, valuing the company at $965 billion post-money.

The round was co-led by Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ, and XN. Significant investors in this round include AMP PBC, Baillie Gifford, Blackstone, Brookfield, D.E. Shaw Ventures, DST Global, Fidelity Management & Research Company, General Catalyst, Insight Partners, Jane Street, Lightspeed Venture Partners, MGX, NTTVC, NX1 Capital, Situational Awareness LP, T. Rowe Price Associates, Inc., T. Rowe Price Investment Management, Inc., and Temasek. It also includes $15 billion of previously committed investments from hyperscalers, including $5 billion from Amazon. Source: Anthropic

Note the $47bn/month revenue run. Back in February it was $14bn. That’s 3x in 3 months.

Prometheus raises $12 bn at $41bn valuation: This is Jeff Bezos’s industrial AI start-up. In case you are worried LLM’s are getting commoditised and the trend is fizzling out, Physical AI and World Models are the new hot trend. Physical AI also requires billions of financing, which will need to go into more chips and data centers.

Prometheus is building what it calls an “artificial general engineer” — software capable of automating the design and manufacturing of complex physical systems, from jet engines to drug compounds.

The ambition is sweeping: replace large swaths of engineering work with AI. Although the startup will automate many aspects of an engineer’s job, Bezos told CNBC that the productivity gains AI delivers will lead to what he calls “labor scarcity” — his term for a world where demand for human workers outpaces supply.

Bezos indicated that a large portion of the capital will go toward the company’s large compute needs. Source: TechCrunch

DeepSeek raises $7.5bn at $50bn valuation: Chinese AI is coming in hot. You might not have been paying attention but DeepSeek expanded from the original LLM model to develop its own AI coding software and are also building their own inference chip. They want to own the full stack (model, software, hardware). Check out their blog to get a sense of the momentum.

Interestingly, the round was structured as a Private Partnership with no voting rights, instead of direct equity. CEO Liang Wengfeng wants to keep total control of the firm. Wengfeng also invested $3bn of the $7.5bn himself. Tencent came in for $1.5bn and CATL invested $740mn. Tencent has its own Hunyuan model, but looks like they want to spread their bets on different models, or they want DeepSeek to use their data centers. CATL is angling “to strategically position themselves to supply the massive energy and storage infrastructure required by power-hungry AI data centers.” Source: DeepSeek

Anduril $5bn raise at $61bn valuation: WW3 is great for DefenceTech. Anduril raised $2.5bn in June 2025 at a $30bn valuation. So +100% valuation gain in a year.

I recommend reading Anduril’s January investor letter on the nature of warfare, which they included as part of their announcement. The highlight for me was the increasing importance of submarines as the only domain where detection is still difficult. I have been waiting to seem some confirmation of this idea. It’s part of why I like Kawasaki Heavy (7012 JP). Thyssen Krupp (TKA GR) also makes submarines. Submarines could be the new hot thing.

Deployment Co raises $4bn: This is OpenAI’s consulting firm to speed the buildout of AI into corporations.

The OpenAI Deployment Company will extend OpenAI’s ability to embed engineers specialized in frontier AI deployment, known as Forward Deployed Engineers, or FDEs, into organizations working on complex problems in demanding environments. These FDEs will work closely with business leaders, operators, and frontline teams to identify where AI can make the biggest impact, redesign organizational infrastructure and critical workflows around it, and turn those gains into durable systems. Source: OpenAI

I’m a little surprised Bain, McKinsey and CapGemini are so eager to bring OpenAI into their business.

The OpenAI Deployment Company is a committed partnership between OpenAI and 19 leading global investment firms, consultancies, and system integrators. The partnership is led by TPG, with Advent, Bain Capital, and Brookfield as co-lead founding partners, and B Capital, BBVA, Emergence Capital, Goanna, Goldman Sachs, SoftBank Corp., Warburg Pincus, and WCAS as founding partners.

Investors also include leading consulting and systems integration firms, including Bain & Company, Capgemini, and McKinsey & Company. The Deployment Company will also work closely with and alongside OpenAI’s Frontier Alliance partners and the broader industry to drive AI adoption and change management globally.

It’s also interesting OpenAI sees AI integration as a big growth opportunity because public consulting/integration companies like Accenture, EPAM and Cognizant (CTSH) screen highly in our QARV data screen as value stocks. They are doing well but investors think they are ‘disrupted’. So are these firms disrupted by AI or is helping corporations integrate AI actually a huge growth trend for them?

Isomorphic Labs raises $2.1bn and Ineffable Intelligence raises $1.1bn at $5.5bn valuation: I’m grouping these two rounds together because they are both Deep Mind spin-offs.

LONDON, 12 May, 2026 - Isomorphic Labs, an AI-first drug design and development company, today announced it has raised $2.1 Billion in Series B funding. This latest round of investment will accelerate the company’s evolution from pioneering novel AI models to applying them at scale. The financing round is led by Thrive Capital, and includes participation from existing backers Alphabet and GV alongside new investors MGX, Temasek, CapitalG, and the UK Sovereign AI Fund, significantly expanding Isomorphic Labs’ global capital base.

The new capital will be used for the continued development and deployment of Isomorphic Labs’ AI drug design engine (IsoDDE), accelerating and expanding its pipeline of therapeutic programs towards the clinic. Additionally, the funding will support our existing hiring targets by integrating world-class AI, engineering, drug design, and clinical talent across our sites; this global scale is crucial to Isomorphic Labs’ long-term vision of applying AI-driven breakthroughs to the most complex biological and medical challenges and addressing the global burden of disease. Source: Isomorphic.

Ineffable Intelligence, a British AI lab founded a mere few months ago by former DeepMind researcher David Silver, has raised $1.1 billion in funding at a valuation of $5.1 billion to join the race for novel AI models that could outperform large language models.

Isomorphic and Ineffable are two more datapoints confirming our view that London is the Seattle of AI (The Seattle of AI no one realises yet). We came up with a public way to play this which is acting very well.

Stegra raises EUR 1.4bn to build Green Steel plant in Sweden: This is one of few examples where VC money goes into building a factory, instead of software. But I guess that’s because the main investor is the Wallenberg Family, and not a Silicon Valley VC fund.

The Wallenberg Investments-led consortium consists of existing investors IMAS and Temasek as well as new investors Bolero and SEB-Stiftelsen. In addition, a large group of Stegra’s existing shareholders continue to invest in the company. These investors include Altor, that will become the second largest shareholder, Hy24 and Just Climate, as well as AMF, AP2, Climate Infrastructure Fund**, Kallskär, Kobe Steel, Lingotto Innovation, Scania, Schaeffler, Security Trading* Stena Metall Finans and Swedbank Robur. Also, a group of Stegra’s second lien lenders, led by AIP Management, have decided to support the project as direct equity investors.

Being Europe’s first greenfield steel mill in 50 years, it will set a new benchmark for the industry. We will do things differently and produce green steel in a fully integrated production process, using end-to-end digitalization, electricity from renewable sources, and green hydrogen. By replacing coal with green hydrogen and electrifying the steelmaking process, we will be able to cut CO2 emissions by up to 95% compared to traditional steelmaking. Source: Stegra

Kalshi raises $1bn at $22bn valuation:

Prediction market Kalshi today announced a $1 billion Series F round at a $22 billion valuation, led by Coatue, with participation from Sequoia Capital, Andreessen Horowitz, IVP, Paradigm, Morgan Stanley, and ARK Invest. Source:Kalshi

It seems a bit off that Kalshi gets a $22bn valuation while CME’s valuation is $95bn, but as we see over and over (and why distributions to LP’s are at record lows), one business lives in the land of private valuations where anything goes, while the other is public. Maybe if Kalshi ever IPO’s it will do better than Spacex.

In the meantime Kalshi does have impressive growth and probably gets a scarcity premium for high growth that isn’t AI.

“There are few categories in recent history that have scaled this quickly outside of AI,” said Tarek Mansour, co-founder and CEO of Kalshi. “Event contracts could become a trillion-dollar market, and we’re still in the early stages of that transition.”

Ramp raises $750mn at $44bn valuation: I’m flagging Ramp (accounting software) because their pitch is how to control token spend by routing simple AI tasks to older models and taking advantage of the falling costs for the older models. They call it the Routing Opportunity. It’s a slightly concerning development for the frontier labs.

Slate (Re Car) raises $650mn: Like Stegra, Slate Auto is one of the few start-ups building something which isn’t software. And they are building something completely All-American. A truck!!

The idea is an affordable, modular truck at $25,000 where you can modify it into an open air Jeep or an SUV plus customise many of the parts. Slate also invites people to create their own customised parts for the truck, which they can then sell in the Slate marketplace. Cool idea.

The round was led by TWG.

Slate Auto has closed its $650 million Series C round. Slate possesses the operating capital to reach the next stage of development, thanks to visionary investors dedicated to developing the affordable and customizable vehicle. TWG Global continues its strong support of Slate and led the round. Slate has taken over 160,000 reservations and will deliver its first vehicles to customers in late 2026.

“For nearly four years, Slate has remained laser-focused on the steps needed to develop our vehicle and reindustrialize our Warsaw Factory, and we will deliver Slate Trucks at nearly half the cost of the average new vehicle—as promised,” said Chris Barman, President, Vehicles.

Top PEVC industries by number of portfolio companies

I harp on the continual PE/VC invetment into software because it is already the largest sector concentration for portfolio companies and they keep ploughing more money into it.

This software concentration is the reason why return of capital to LP’s is at 20 year lows. Software is in the midst of record uncertainty and potential disruption. Most of these older software investments are now unsellable in the public markets (without big discounts to NAV). But Silicon Valley doesn’t know how to do anything else.

Bottom PEVC industries by number of portfolio companies

Then you look at all the industries PE/VC won’t invest in, and they are looking pretty good right now. Contract drilling, cutting tools, railroads, auto parts, coal, banks, precious metals, and oil refining. It’s why I flag the rare deals like Stegra or Slate Automotive where someone is doing something different.

Below is link to the data sheet on over 2,700 rounds during Q2. You can go over all the rounds to see which fund is participating, or filter it by fund investor. For example, you might like to see the 62 companies a16z invested in during the quarter.

For most rounds the data is the Investor (PE or VC firm) and then target company. For some rounds the data also includes the size of the round and % of the round taken by the investor.

Read more

A few quick notes of what I found interesting:

• The family made its money in high end real estate around China. Their high end residential developments are still challenged and hard to sell. They are trying to sell the projects they have and are not inclined to put new money back into real estate.

• As muc…

Read more

So you might have missed a certain headline on Monday.

It was one of those quiet, subtle headlines no one pays attention to during a busy week, but years later we look back on it and realise it was the most money making headline of the week.

Expand Energy acquires Twin Eagle

Expand Energy, the largest natural gas producer in the US, acquired a gas marketing business called Twin Eagle for $1.25 billion. Twin Eagle is a leader in buying gas from producers, moving it around the US and selling it to industrial customers.

The transaction makes Expand (renamed after merger of Chesapeake and Southwestern) not only the largest producer (7.5 bcf/day), but also the largest marketer of natural gas (14 bcf/day) .

As a gas marketer Twin Peak has built a huge network of storage facilities around the US with 44 bcf of capacity. Twin Eagle also controls 2 Bcf/day of transportation rights on US pipelines. They can move gas anywhere.

So why do I think this is such an important transaction?

Especially, since EXE didn’t move on the announcement.

Because it tells us Expand is a hedge hog.

Back in May I went to a two day AI-Energy conference.

The idea was that with AI and the datacenter build out it was important for the energy sector and the tech sector to start talking to each other more.

The conference was exciting and full of private equity funds, datacenter developers, utilities, pipeline companies, gas turbine manufacturers, and hyperscalers. Everyone breathlessly announcing how much demand they were seeing.

Except there was one company no one really seemed to care much about. Even though they were a main sponsor of the conference.

They hosted an early fireside chat on the morning of the second day. Many of the previous day’s attendees seem to have had trouble getting out of bed. So attendance was a little slim.

But I was there in the front row with my coffee.

Because it was a fireside chat with Mike Wichterich, the CEO of Expand Energy.

And it was the best presentation of the entire conference.

Mike had two main messages.

#1 The US natural gas market is undergoing an unprecedented confluence of demand drivers. These demand drivers could lead to 30 bcf/day of incremental demand in a 100 bcf/day market. That’s massive.

It’s three trends coming together.

It’s the growth of LNG terminals in the Gulf Coast which is now super charged by the disruptions in Qatar.

It’s the demand from all the datacenters once they are running.

And it’s the growth of new factories in the US like the $100bn of investment TSMC is making in Phoenix.

I loved Mike’s pitch because it confirmed my sneaky suspicion that natural gas is the contrarian backdoor way to play AI. The same way iron ore was the best way to play China. Everyone wants to invest in the datacenter capex plays, but what happens after the datacenters are built? What’s the play for the next twenty years? Natural Gas.

For years natural gas was all about the growth of fracking and growth of supply. Now natural gas is a story about demand. We wrote about this in The Final Bottleneck.

But that’s all thematic.

The second thing that caught my attention was Mike’s view that Expand needed to get better at monetising volatility. Natural gas is the most volatile commodity. It can move 300% in a week on an unexpected cold snap. Nothing else moves like that.

Mike said rather than always trying to smooth out gas price volatility, Expand needed to accept the volatility and get better at making money from it. The firm needed to be opportunistic; sit on gas when prices were depressed, and sell it when they were spiking. But to do this they needed to get better at gas trading. And they needed more capacity to store gas when prices were low.

Which is why it makes perfect sense that Expand is acquiring Twin Eagle. It brings the gas trading skills and 9x’s their storage capacity.

But again, ‘So what?’ you might say.

A natural gas company bought a gas marketing company.

Who cares?

Investing in Hedgehogs

In 2001 Jim Collins wrote Good to Great.

He wanted to know what made companies ‘great’.

He had examined 1,435 companies over a 40 year period. He then identified the 11 best performers of the 1,435 and intensively examined why they had outperformed so dramatically over decades. He went over strategy, competition, financial statements, and meetings with management.

What was the secret to success?

What did the successful companies of the time, like Walgreens, Wells Fargo and Kimberly Clark, all have in common?

The great companies were all Hedgehogs.

Collins compares hedge hogs with foxes. Foxes are clever and know something about everything. They hop around chasing one opportunity after another. A Hedge Hog knows one big thing, never budges and focuses on it relentlessly.

A Hedge Hog company identifies the key business where they can be great. They compete where they can be great and cut the businesses where they are just another competitor.

“You’ll know that you’re getting closer to your Hedgehog Concept when you align three intersecting circles that represent three pivotal questions: What can we be the best in the world at? (And equally important—what can we not be the best at?) What is the economic denominator that best drives our economic engine (profit or cash flow per “x”)? And what are our core people deeply passionate about? Answer those three questions honestly, facing the brutal facts without blinking, and you’ll begin to see your Hedgehog Concept emerge.”

The Hedgehog strategy is long-term, low profile and not glamorous. It’s how surprisingly great companies emerge out of unexpected industries.

And it’s why I was so interested in the Twin Eagle acquisition. Because it looks like Mike Wichterich is executing a hedge hog strategy. He is quietly building Expand Energy into a Great US natural gas company. Expand is already the largest producer, but Mike wants make Expand the most profitable producer through better marketing and trading. He wants to improve profit/molecule.

And notice despite all the AI hype he’s not trying to ‘vertically integrate’ and build a data centre, which some gas producers are trying to do. No, he wants to stick to his knitting and be the best at US natural gas. Even if it’s not sexy and the market doesn’t care. For now.

And that’s why I say Expand buying Twin Eagle was the money making headline of the week.

Have a good weekend.

Erik

What are two mega-trends investors fear could end any day?

The two are the same.

I’ll explain what I mean.

Let’s go over the global factor model rankings.

YWR Global Top 30

Why do estimate revisions work?

Let’s remind ourselves how we use the global factor model.

The YWR Global Factor Model rankings are a monthly screen of the top ranked stocks on a mix of estimate revision momentum, valuation and price momentum. Estimate momentum is the most heavily weighted factor Total Score so the screen mostly surfaces the strongest estimate trends in the market (you can also filter by any of the sub-scores; estimates, valuation, price momentum).

This gives us a dashboard of where estimates are moving the most across the world. And if estimates are moving strongly it tells us there is a story. Something is changing which the sell side experts didn’t predict. And because of ‘anchoring bias’ most of the time these estimates will continue to trend upwards/downwards in the future.

Our job is to understand the narrative and use our imagination. What’s happening? Why are these estimates rising/falling? What’s new? Could these trends go much further? What inning are we in?

I wrote about this more thoroughly in Why Estimate Revisions Work. It’s an important behavioural bias to understand.

The Top 5 Ranked Industries

To determine the top ranked industries we take the top 300 ranked stocks and group them by industry.

And this is where our two mega-trend present themselves.

The two big trends are Cybertron and the Iran War.

Semiconductors, electronic components and industrial machinery are all Cybertron Themes.

The other two, Marine Shipping and Refining, are Iran War beneficiaries. Oil and Gas Production didn’t make the Top 5 but also scores well and is also an Iran War play.

In the monthly chart you can see how all 5 sectors start rising sharply at the end of 2025 and early 2026.

The Iran War Theme

This theme is the most interesting. It impacts everything. Energy costs are rising, shipping rates are rising, refining margins are exploding, and yet nobody wants to own these sectors. Investors are afraid they will come in, buy on the highs, then look like an idiot just as everything gets resolved and crashes.

I am an energy bull anyways, so it’s easier for me, but this trend looks massive. Like world changing massive. My initial view when the war started was that it might last a few months and wouldn’t be a big deal, but the situation is getting more complicated by the day. Now I can imagine this going on for years, like the Ukraine.

Maybe we are at the start of a multi-decade re-architecting of global energy away from the Middle East. Maybe it’s too risky to have so much energy coming from such a volatile place. The shift will come in many forms. More supply from the US and Canada (How Qatar changes everything), new drilling and exploration in Africa and South America, new pipelines to bypass the Straight of Hormuz, nuclear and a shift to renewables. This is many trends across different industries and around the world, but with one theme; less dependence on hydrocarbons from the Middle East.

In the meantime there is no quick way to build new refineries, but analysts expect the bull market in refining to collapse over the next 2 years. Marathon is supposed to make $40/share in 2026, but this drops to less than $30 in 2027 and $20 in 2028.

That’s why I say this is the is biggest trend investors are afraid to own. It’s all going to collapse at the end of this year.

And all the Wharton MBA’s at the hedge funds hate the idea of buying a capital intensive, highly cyclical industry like refining. Their programming is to never buy those industries and that ‘geopolitics’ is too hard to predict.

But what if I pitched it this way? Would you like to own an industry where you run a complex, expensive factory that takes 7 years to build, and I just evaporated 20% of your competition.

There is now a 20% supply/demand mismatch and no competitive response for 7 years. Oh, and because public investors are negative on you, you can take all the supernormal profits and buy back your shares at low valuations so the benefits accrue into the future forever.

Shipping is another one.

Shipping seems like a one dimensional chaos trade nobody wants to own. But if energy supply has to be more distributed, and the global maritime shortcuts (Panama Canal, Straight of Hormuz, Red Sea) are impaired does the world require a lot more ships?

The data tells me there is a bull case in shipping, but nobody can articulate it other than it’s temporary. It’s just a matter of time before the shipyards churn out more ships and crash the bull market.

I don’t have the answer on shipping, but sometimes it’s a good thing if nobody can see why a bull market will continue. The best bull markets are the ones nobody expects. If it has a nice trend that McKinsey can explain in a Powerpoint, then companies will flood it with capital and ruin it.

I’ll probably add a bit to Frontline, even if I shouldn’t.

Now on to the world according to Kingboard.

Kingboard Holdings

What is going on with this $6bn ‘chemical stock’ listed in Hong Kong?

Why is it the top ranked stock in the world?

Kingboard makes printed circuit boards, but also glass yarn, which are both doing surprisingly well (hence the earnings upgrades).

What’s so interesting about that?

Maybe it’s an AI play somehow and we all know AI is a capex bubble that is about to crash (like shipping and refining).

I want to share some snippets from Kingboard’s 2025 Annual Report because it’s a window into the Cybertron theme.

Multiple mega-technology trends are converging at once, and these semiconductor related companies are getting hit with demand from multiple use cases. It’s a combination of 5G telecommunications, autonomous vehicles, robots, satellites and energy storage.

These trends are all merging and you get companies like Kingboard, which we’ve never looked at, massively benefitting in ways you wouldn’t expect.

Laminates Division: Entering 2026, overall demand in the electronic market remains robust, with continued strong demand for AI data centres, robotics, and computing. High-performance, high-reliability, and high-stability laminates are expected to become the mainstream requirement in the electronics market of the future. Through lean production, and a market-led and R&D-supported strategy, the Group is steadily increasing its market share of mid-to-high-end products, including high-frequency, high-speed, mid-to-high heat-resistant, halogen-free, and prepreg materials.

The Group has successfully developed various high-frequency, high-speed products capable of meeting the core material requirements for high-reliability application scenarios such as AI servers, new energy vehicles and energy storage systems. Through coordinated development across its vertical industry chain, the Group has also developed HVLP3 copper foil for AI servers and ultra-thin VLP copper foil for IC packaging substrates. Its high-end laminate products have been certified for use by several world-leading Tier-1 automotive component manufacturers and key domestic and international telecommunications clients, positioning the Group to comprehensively meet the diverse needs of downstream customers into the future.

That’s the bull market in ‘laminates’, which I’m pretty sure none of us knew was happening. Now hold on to your hats for the bull market in glass yarn.

Low-dielectric (Dk) electronic fibreglass yarn and low-CTE electronic fibreglass yarn, used in high-end telecommunications, particularly in 5G and 6G communication and in AI servers, are currently in short supply. The Group has successfully developed second-generation low-Dk and low-CTE electronic fibreglass yarn with higher added value, making it one of the very few laminates players globally to have mastered the core technology for low-CTE electronic fibreglass yarn.

In the first half of 2026, the Group will add three more kilns, each with an annual capacity of 500 tonnes, to produce second-generation low-Dk and low-CTE electronic fibreglass yarn, meeting the high computing requirements driven by the development of 5G and 6G communication and AI. Plans are also in place to add another eight kilns, each with an annual capacity of 500 tonnes of low-Dk, low-CTE, and Quartz electronic fibreglass. yarn, increasing the total number of specialty electronic fibreglass yarn kilns in the industrial park to twelve and further consolidating and enhancing the market share for Kingboard high-end products.

Currently, there is a severe supply-demand imbalance for various electronic fibreglass fabric and yarn products,with laminates producers facing production disruptions due to a shortage of electronic fibreglass fabric, driving market prices even higher.

In view of this the Group’s electronic fibreglass yarn and fabric capabilities are expected to generate satisfactory profits for the Group going forward.

And finally, the bull market in printed circuit boards.

PCBs Division: Driven by ongoing digital and smart transformation, the PCB industry is rapidly shifting towards high-performance, high-frequency, and green manufacturing. The Group will continue to actively develop higher-layer-count and advanced high-density interconnect (HDI) PCBs.

With its superior craftsmanship, comprehensive certifications, and excellent quality, the Group holds multiple advantages in sectors such as automotive,telecommunications, and consumer electronics. The rise of Large Language Models (LLMs) has driven a surge in data processing demands, accelerating iterative hardware operations for servers and switches, and injecting new vitality into the PCB industry.

At the same time, the proliferation of electric vehicles has led to increased complexity in electronic control systems, significantly boosting the usage of PCBs. Kingboard Group has equipped itself with technologies related to PCB manufacturing for 6G wireless communications, AI servers, low-orbit satellites, 4D imaging mmWave radar, automotive high-end common control units, high-voltage fast-charging systems for new energy vehicles, and intelligent robots.

The division will continue to strengthen the core competencies of its PCB brands, including Elec & Eltek, Techwise Circuits and Express Electronics. The Group is expanding its PCB production capacity by 2.5 million square feet per year in Kaiping City, Guangdong Province. The new capacity will primarily focus on producing multi-layer PCBs and multi-layer HDI PCBs for high-end AI products and is scheduled to commence production in the first half of 2027.

Do you see what I mean? Do see all the subtrends going on in circuit boards, laminates and fiberoptic yarn, and how they all intertwine?

It’s not high conviction yet, but for me the case is building that there are more legs to the Asian ‘semi supply chain’ stocks than we think.

They aren’t just hyperscaler data center build out plays.

Top 5 Ranked Countries

China, Japan, Taiwan, and South Korea are full of these Kingboard type companies. They are essential suppliers to the Cybertron build out, which is why month after month they are the top ranked countries in the world.

It has massive ramifications for asset allocation.

It’s also why we are pivoting our asset allocation and attention more to Asia.

Below are links to:

• the YWR Factor Model Dashboard which combines the rankings with the estimate charts.

• the YWR Factor Model Rankings Sheet in Google Sheets format so you can download it.

• the top inflection stocks by country and industry.

Read more

In the stories the author is able to pick out the a key design choice the scrappy company makes, which the sleepy incumbent under appreciates. This feature or design choice plays out over time and the scrappy upstart eventually disrupts the whole industry. By the time the incumbent sees the problem it’s always too late.

Netflix was like that. The key feature in the beginning was to be able to watch what you wanted, when you wanted in contrast to scheduled cable TV. Netflix also piggy backed for free on the high speed fibre cable companies were spending billions on to connect their triple play bundles to the home. And over time Netflix became a ‘smart’ media company able to learn in real time what its customers liked watching and build algorithms to keep them binge watching.

I’ve read that in the beginning the cable executives studied whether Netflix was a threat, but saw it as additive to cable viewing, and only 3% of their subscribers also subscribed to Netflix so it wasn’t a big deal. Back in 2004 Netflix had 2.6 million subscribers and only made $22mn in net income so you can see why it was easy to dismiss.

But Netflix had developed a disruptive product and like a sharp knife they relentlessly pried cable customers away from their dumb bundle of channels. In the end it disrupted the entire cable industry and changed how customers consume media.

And the stock price performance was amazing.

The story of Apple and the iPhone was similar. AT&T thought the iPhone would be additive to the network and grow data consumption. It did, but there was a key feature AT&T didn’t appreciate, the App Store. AT&T thought they had considered every angle. They owned the spectrum, and they had the customer relationship with direct billing to the customer, so there was no way to get disintermediated. But Apple created a digital market place inside their operating system that bypassed AT&T all together. And they created the AppleID so all the apps and data related back to an Apple ID and not a phone number. In hindsight AT&T was too far behind technologically to ever have imagined what was going to happen. Their business was managing infrastructure they had no way to connect dots and see what Apple was building.

Again, we look back at these features and design choices and see how they steadily play out over 20 years creating life changing value for the disrupter and misery for the incumbents.

But there are two problems with these ‘stories’.

First. These trades are always way harder than they look with later with benefit of 20 years of hindsight. As an investor when you are in the trade it’s hard. It’s never clear how big the trend is going to become. Then there are periodic counterattacks by the incumbents where they create their own version of the product. Or, new start-ups with a seemingly better tweak on the product. Then there are the years where the growth rate slows and everyone says it is over. How many 40% drawdowns did Netflix shareholders have to sit through along the way?

And the disrupter’s stock is always ‘expensive’, so even if you ‘get it’ is it really worth that much? And what if you are wrong? With such a high valuation if anything goes wrong the downside will be painful.

The Second problem with these business case studies.

Stories are fun to read but they don’t make us money.

Fine, Netflix and Apple were big successes, but show us the next one?

Show me the next Netflix that everyone under appreciates today and which I can buy now and make 10x.

Sadly, the business case study analysers are always quiet on this. Great at analysing things in hindsight, but no use at making us money.

Which is why we come to YWR.

We always want the trade. How do we make money?

So would you be interested in a Netflix style disrupter, except bigger?

I mean potentially massively bigger.

And would you be interested if this disrupter has market cap of just $15 billion and is trading on the lows?

If so come with me.

Read more

But also reminds us of a challenge we are going to have to navigate in the near future.

We will discuss that at the end.

The chart pack is 54 slides covering:

• S&P 500 Earnings Estimates

• BofA Survey

• University of Michigan Consumer Survey

• 10 year bond yields

A link to the full deck is at the bottom of the post.

I want to especially highlight the charts I included from the University of Michigan consumer survey because I found them shocking. Maybe consumers are just better at complaining these days, but the results are at historic lows despite full employment and a booming stock market. I found it surprising how extreme all the charts are right now.

It could be a big deal.

So get yourself a coffee and go over all the slides, but here are 8 highlights then slide 9 and my final takeaway.

#1 We hit $400/share for the S&P 500 2027 EPS Estimate.

This is a key milestone for our S&P $10,000 target (S&P $10,000 Update). Estimates revisions remain positive.

#2 The market is no longer a Mag7 story. The ‘Other 493’ are growing earnings 25%.

#3 S&P 500 EPS estimates for 2027 are 17% growth with the energy sector negative.

Expectations around energy look too bearish to me.

#4 Investors are piling into Tech and out of Energy.

Record inflows into tech funds and record outflows from energy as we go into a global energy crisis. Kind of amazing.

#5 Fund managers are sanguine about inflation and rate hikes.

CPI hits 4.2% and everyone is convinced it’s a 1-off and the Fed doesn’t need to do anything. According to the BofA fund manager survey 83% of fund managers expect no rate hikes before the midterm elections.

#6 Consumers have never felt worse.

The market is making new highs, the economy is booming and unemployment is low, but the University of Michigan consumer survey readings are off the charts negative. Like lowest ever.

Within the readings there is an unprecedented polarisation in sentiment by political party.

#7 Consumers are in pain over rising prices.

Inflation is only 4% but consumers are in extreme pain.

#8 Consumers are miserable and expect things to get worse.

The future expectations reading is making an all time record low, even lower than the financial crisis. Consumers are more negative about their future than they have ever been. This is profound.

There is a rising concern about future unemployment.

So what story do I see playing out here?

What is the challenge we need to face?

Here is the other most important chart from the University of Michigan Survey. Chart 9.

Read more

Having a hard time investing in new stock ideas?

Is it always one thing or the other?

Let me guess.

Software looks cheap but it might be a value trap and could all get disrupted.

Semiconductor related industries are growing, but getting expensive and it could be a bubble.

Energy stocks are doing well, but it could just be a cyclical boost from the Iran war and structurally oil is in decline.

Refineries are minting, but it’s all from the Iran War and how much of a valuation do you put on something which has a binary outcome and could end at anytime?

Real estate has gone nowhere, but then who really needs all these office buildings anymore?

Consumer stocks are on the lows, but maybe consumers are getting structurally disrupted by AI and robots.

It goes on and on.

Either the company is expensive and in some narrow growth area, or it is cheap and getting disrupted.

The problem you are facing is that we are in a time of widespread disruption and these are all Western style H-Corps.

‘Hierarchical’ Companies.

The Western corporate model is ‘focus’. Focus on an industry, business line or process and just do that. If the company expands into a new unrelated industry it is spun off to create ‘shareholder value’.

So you get very focused companies, with managers and employees highly specialised in one industry which is either doing well, or it is not. And for the companies not doing well, it is hard for them to know what to do. It’s not in their DNA, or expertise, to expand into unrelated industries.

Because then they would start to be a conglomerate.

And conglomerates get discounts.

Except maybe a conglomerate is what you need right now.

Maybe in a world of rapidly changing technologies and geopolitical stress the Japanese style conglomerate, the J-Corp, is what you need.

At least that’s a view I’m coming around to.

J-Corps are different from H-Corps.

H-Corps exist to create shareholder value.

J-Corps are built to survive. That’s why the history of many large Japanese companies can go back as far as the Meji Era (1860-1920).

J-Corps are permanent capital companies, which employ people directly out of college for life. Employees are paid on seniority and bonuses based on the performance of the overall company, not personal performance. Usually, the shareholding is tightly controlled by related companies and the bank financing is also from related companies (keiretsu structure).

Most J-Corps are built around manufacturing, but what they manufacture can be very diverse. TOTO manufactures toilets, but also ceramic chucks for semiconductors. Sony is the same. It makes optical sensors, video games smart phones and speakers, plus makes movies and runs a bank.

It seems weird to do so many different things until you put yourself in the shoes of a Japanese company.

“Hey guys, we all have jobs here for life, but as you know our traditional business lines are in trouble and we aren’t growing anymore. So let’s all huddle together, come up with some good ideas and figure out how to adapt and keep the company going.”

So everyone figures out how to take the manufacturing and process skills from their existing products and apply them to something else. Which is why over time we have arrived at this state where Japanese companies do so many different things.

Everyone is on the same team and has to make it work. Nobody can jump ship to a competitor because it’s hard to get hired except directly out of university. And the company is committed to you and not going to layoff everyone so you actually have loyalty to the company and want to make it work.

This is very different from what happens at an H-Corp under stress.

We’ve always know Japanese companies do too many things, but not why they do so many different things.

This J-Corp model is built for survival and adaptation, and I think that’s why my gut keeps coming back to Kawasaki Heavy Industries (7012.JP), when it goes against all of my stock market training. And it’s why as I kept thinking about Kawasaki and questioning why I liked it when it looks like a mess and is hard to model, I eventually found great article “Why Japanese Companies do so many different things”, which is really a deeper analysis of industrial organisation.

But studies of “industrial organisation” are for another day.

We are here to make money.

And I think Kawasaki Heavy Industries looks interesting.

This isn’t a 5x idea.

This isn’t an inflection point idea.

It’s probably a 1x idea.

It’s something that just works over time. An investment which survives. And in this day and age that might be how we should look at things.

But if I wanted to hype it a bit I’d say Kawasaki Heavy is 5 (actually more) Silicon Valley start-up mega themes wrapped into a Japanese industrial conglomerate.

Yes, it’s a conglomerate. But because it’s a conglomerate it has the existing manufacturing facilities, technical expertise, financing, distribution, brand and customers to execute on long-term themes which will play out over years with lots of disruption and chaos along the way.

And like a Silicon Valley start-up they just did a $600mn capital raise to accelerate investment into these mega-themes. Except, unlike a Silicon Valley start-up the market of course punished them for this and the stock has sold off.

This 5 in 1 Silicon Valley mega-theme start-up trades at 20x March 2027 expected earnings (I’ve included my earnings estimate model at the bottom of the post).

Kawasaki Heavy: 5 Silicon Valley Start-ups for the price of 1.

Read more

The most important question for pension funds is changing.

For decades, the question was: Where should we allocate capital?

The answer was usually expressed through a strategic asset allocation: public equities, fixed income, real estate, private equity, infrastructure, credit, hedge funds, cash. Define the buckets. Set the weights. Hire managers. Monitor performance. Rebalance. Repeat.

That model created discipline. It gave boards a governance framework. It helped large institutions avoid emotional decision-making. It created a common language for risk, return, liquidity, and long-term liabilities.

But it was designed for a simpler market structure.

The world pension funds invest in today is no longer neatly organized into asset classes. Artificial intelligence infrastructure is not just a technology theme. It is public equity, venture capital, private equity, private credit, infrastructure, real estate, power, semiconductors, data centers, water, regulation, and geopolitics at the same time. Private credit is not just credit. It is bank disintermediation, sponsor finance, asset-based lending, insurance capital, liquidity transformation, and shadow monetary policy. Sports and media rights are not just alternatives. They are intellectual property, streaming, data, consumer behaviour, gaming regulation, and recurring cash flows.

The old model classified assets.

The new model must understand systems.

That does not mean Strategic Asset Allocation is dead. It means SAA should be the constitution, not the operating system. It should define long-term risk appetite, liability awareness, governance discipline, and institutional guardrails. But it should not become a substitute for judgment. It should not trap judgment inside categories that markets themselves are already outgrowing.

The real work is no longer only setting the policy portfolio. It is building the institutional engine that can learn, adapt, and allocate intelligently between policy decisions.

From allocator to platform

The traditional pension fund model is built around selection.

Select asset classes. Select managers. Select benchmarks. Select commitments. Select risk budgets.

The future model will be built around orchestration.

A pension fund’s greatest underused asset may not be any single manager, mandate, or portfolio company. It may be the network itself: the web of external managers, internal teams, co-investors, consultants, banks, portfolio companies, data providers, operating partners, and market relationships that surround a large pool of long-term capital.

Today, much of that network is still managed as a collection of relationships. In the future, it should be managed as a platform.

This is where pension funds can learn from the architecture of multi-manager platforms - firms such as Citadel,Millennium, and BAM - without copying their mandate, economics, or risk profile.

The lesson is not that pension funds should become hedge funds. The lesson is that the most powerful investment organisations increasingly behave like operating systems for capital, talent, data, risk, and accountability.

Multi-manager platforms understand specialization. They understand risk budgets. They understand telemetry. They understand that capital should move toward talent, opportunity, and discipline. They understand that a portfolio is not just a list of positions. It is an architecture.

Pension funds have a different purpose. They exist to pay beneficiaries, not to maximize short-term trading profits. They operate under fiduciary duty, public scrutiny, governance constraints, and long-term liabilities. They cannot and should not copy the economics, leverage, turnover, or culture of hedge fund platforms.

But pension funds may have something more powerful: permanent capital, cross-asset visibility, access to the world’s best external managers, and the ability to be a strategic partner rather than a transient source of assets under management.

The issue is not that pension funds lack these advantages. It is that the industry has usually expressed them through allocation programs rather than platform architecture.

That is the opportunity.

Every allocation should make the platform smarter

The old question was simple:

Did the manager outperform?

That question still matters. It will always matter. Pension funds cannot pay benefits with interesting ideas or elegant data architecture. They need returns.

But the next question is more powerful:

Did the manager strengthen the platform?

A manager’s value should not be measured only by excess return. It should also be measured by the data, access, transparency, co-investment flow, underwriting insight, sector intelligence, risk visibility, and institutional learning the relationship contributes.

Every allocation should make the platform smarter.

A venture manager may help the pension fund understand how AI is changing software margins before public markets fully price it. An infrastructure manager may reveal the bottlenecks in power, land, cooling, and grid connections. A private credit manager may identify stress in a sector before rating agencies or public credit markets respond. A public equity manager may see how a private market theme is being repriced in real time. A real estate manager may surface what is happening in logistics, housing, offices, data centers, or consumer behaviour. A hedge fund manager may understand positioning, liquidity, and cross-asset contagion.

Individually, these are manager updates.

Collectively, they are a distributed intelligence system.

A networked pension fund does not ask each manager only for performance. It asks: what are you seeing, what are we learning, where are risks building, where is capital scarce, where is consensus wrong, where is the next opportunity forming, and how does this change the total portfolio?

This is the move from manager selection to manager orchestration.

Platform value is the new alpha

Pension funds have historically paid managers for alpha. The next generation of strategic partnerships should also help build the platform.

That does not mean asking managers to give away proprietary secrets. The best managers will rightly protect their edge. But pension funds should be much more deliberate about what they receive in exchange for long-term capital, strategic credibility, and scale.

The negotiation should not only be: What fee am I paying?

It should be: How much stronger does this relationship make my platform?

That requires a broader definition of value creation.

The first layer is portfolio value: returns, alpha, diversification, downside protection, liquidity management, cost efficiency, and risk-adjusted performance.

The second layer is platform value: better data, better access, better co-investment rights, better underwriting, better look-through, better risk visibility, and better institutional memory.

The third layer is option value: the ability to understand and underwrite new markets faster when they become investable.

That third layer matters because the future will not arrive as a clean new asset-class label. It will arrive as a messy cluster of contracts, cash flows, technology, regulation, infrastructure, and behaviour.

The future pension fund does not need to chase every new market. It needs the institutional architecture to recognise when a new market becomes real.

Total Portfolio Approach as an operating system

This is why the Total Portfolio Approach matters.

TPA should not be seen as a fashionable governance phrase. It is a practical response to a world where opportunities cut across asset-class boundaries. Instead of asking whether an opportunity fits neatly into private equity, infrastructure, real estate, credit, or public markets, TPA asks a better question:

Where does this dollar contribute most to the total portfolio, considering risk, liquidity, opportunity cost, manager quality, data access, and platform value?

This shift is already visible among sophisticated asset owners. CPP Investments has recently described disciplined flexibility as central to making a Total Portfolio Approach work in practice, while CalPERS announced in November 2025 that its board adopted a new Total Portfolio Approach intended to increase flexibility and focus on total-fund outcomes.[1][2]

The word disciplined is essential. Flexibility without discipline is just discretion. TPA does not mean weakening governance. It means replacing rigid bucket management with stronger total-portfolio judgment, clearer risk ownership, and better information.

A networked pension fund needs SAA as its constitution. But its operating system should be total-portfolio thinking.

Data rights are the new fee rights

If the pension fund is becoming a platform, then data rights become strategic.

Manager reporting today is still too often built around quarterly PDFs, backward-looking attribution, inconsistent templates, and fragmented systems. That is increasingly inadequate for any institution trying to understand risk, liquidity, and opportunity across the total portfolio.

The next generation of manager negotiations should include structured data rights, APIs, look-through reporting, exposure transparency, co-investment notification rights, data-quality standards, digital audit trails, and clear escalation rights when transparency obligations are not met.

Again, this is not about micromanaging managers. It is about matching discretion with accountability.

The manager keeps discretion. The platform gets telemetry.

This is not surveillance; it is alignment.

A great manager needs freedom to act. The pension fund does not need to approve every trade, every loan, every acquisition, or every position. But it does need to know whether the manager is operating within the agreed boundaries. It needs visibility into risk, liquidity, concentration, leverage, style drift, and mandate alignment.

Give the manager flexibility on the field, but wire the agreed scoreboard into the platform.

The future IMA is a control system

The traditional investment management agreement or side letter is mostly legal prose. It defines the mandate, fees, restrictions, reporting obligations, termination rights, and other legal terms.

In the future, that will not be enough.

The future strategic mandate will likely have three layers.

The first is the legal contract: fiduciary responsibility, mandate scope, confidentiality, fees, liability, termination, regulatory obligations, and dispute resolution.

The second is the data contract: what data must be shared, in what format, how frequently, through what systems, under what privacy constraints, and with what standards for quality and completeness.

The third is the economic and risk contract: risk budgets, exposure limits, liquidity rules, drawdown triggers, breach notifications, key-person events, transparency obligations, co-investment rights, capacity rights, fee mechanics, and capital-scaling rules.

This is where smart-contract thinking becomes useful - not as a blockchain slogan, but as a way to make obligations more structured, monitored, and eventually machine-readable.

The point is not crypto. The point is programmability.

The Bank for International Settlements has described tokenisation as part of a broader shift toward programmable financial platforms, while also emphasizing that trust in money and sound governance remain central regardless of technological change.[3] That distinction matters. Programmability should not be confused with speculation. It is about making ownership, obligations, data, settlement, and accountability more machine-readable.

A future mandate should not simply describe a relationship. It should encode how the relationship operates.

The first AI revolution is institutional memory

AI-native should not mean replacing investment judgment with machines.

It should mean building an organization where knowledge compounds.

In most investment institutions, too much knowledge disappears into inboxes, PDFs, meeting notes, quarterly reports, consultant decks, and individual memories. A manager says something insightful in a meeting. A portfolio company reveals a signal. A co-investment memo captures a sector thesis. A risk report flags a pattern.

A credit manager notices stress. A venture manager sees a new behavior. A public markets manager identifies a valuation disconnect.

Then the institution forgets.

Or, it remembers only through the people who happened to be in the room.

An AI-native pension fund changes that. It captures meetings, memos, manager updates, transcripts, risk reports, portfolio data, investment decisions, and post-mortems. It structures them. It challenges them. It compares them. It reuses them.

The first revolution is not autonomous trading.

The first revolution is autonomous institutional memory.

A pension fund that can remember better can underwrite better. A pension fund that can compare better can allocate better. A pension fund which can connect weak signals across asset classes can see systems earlier.

This is where the networked pension fund becomes powerful. Not because it has more data than everyone else, but because it has better context. It can see how a theme appears in venture, then private equity, then credit, then infrastructure, then public markets. It can see the battlefield.

From human managers to bounded agents

Only after building this foundation should pension funds think about autonomous agents.

The agent economy is not the core thesis. It is a frontier implication.

Still, it matters.

This is not an argument for delegating fiduciary capital to autonomous software today.

Pension funds already allocate capital to agents. We call them managers. An IMA is a boundary-setting device: here is the capital, here is the universe, here are the limits, here is how you report, here is how you get paid, and here is when we can terminate you.

Autonomous AI does not change that principle. It makes it more explicit.

If capital is ever delegated to software agents, the same architecture will be required: mandate, limits, telemetry, economics, auditability, escalation rights, capital scaling, legal wrapper, and human accountability.

Recent work from the IMF on agentic AI in payments highlights exactly this tension: AI systems may act probabilistically, but payment, authorisation, settlement, compliance, and operational resilience require deterministic controls. NIST has also launched an AI Agent Standards Initiative focused on agents that can function securely on behalf of users and interoperate across digital ecosystems.[4][5]

That is the institutional lesson.

Let the agent reason probabilistically. But make authorisation, risk control, settlement, and accountability deterministic.

The future is not AI replacing the fiduciary. It is fiduciaries designing more programmable forms of delegation.

The human should not approve every action. That would destroy the value of autonomy. The human should define the mission, design the mandate, set the risk appetite, approve exceptions, control capital scaling, and own accountability.

The agent gets autonomy inside the boundary.

Humans own the boundary.

The future investable universe

The investable universe is expanding beyond traditional labels.

Some future opportunities may look like tokenized private assets. Others may look like music or media royalties, sports rights, compute contracts, power capacity, data marketplaces, API marketplaces, carbon and biodiversity markets, litigation finance, cyber insurance, digital identity infrastructure, agent-security systems, or prediction-market-like information infrastructure.

This is not an argument for speculative investing. It is an argument for better architecture.

A pension fund that remains organized only around yesterday’s asset-class labels will struggle to evaluate tomorrow’s systems. It will ask, ‘Which bucket does this belong in?’ when the better questions are:

What cash flow is being created?

Who controls the data?

Who owns the rails?

What is the liquidity?

What is the contract?

What is the risk transfer?

What is the regulatory perimeter?

Where does durable value accrue?

How does this affect the total portfolio?

Assets are becoming more executable. Contracts are becoming more data-rich. Markets are becoming more programmable. Economic activity is becoming more machine-readable.

The pension fund does not need to chase the frontier. But it does need to understand it.

The cultural shift

None of this works with a bureaucratic mindset.

The pension fund of the future cannot operate like a passive reporting organization. It needs an elite, entrepreneurial, cross-asset team that can challenge managers, synthesise patterns, negotiate data rights, understand technology, respect fiduciary duty, and think commercially.

The culture must shift from reporting to underwriting.

From manager selection to manager orchestration.

From allocated capital to intelligent capital.

From asset-class ownership to system understanding.

That does not mean abandoning governance. It means upgrading governance. The more dynamic the platform, the stronger the controls must be. The more data the fund receives, the more serious cybersecurity and privacy become. The more programmable the mandates, the more carefully humans must define the boundaries. The more autonomy is delegated, the more explicit accountability must become.

This is not a call for looseness. It is a call for a higher standard.

The legitimate objections

A skeptical CIO or board member will raise fair objections.

This sounds complex. It is.

Managers may resist data sharing. Some will.

Tokenization does not magically create liquidity. Correct.

AI can produce false confidence. Absolutely.

Autonomous agents should not be allowed to manage fiduciary capital without strong controls. Agreed.

Total Portfolio Approach can become dangerous if it gives discretion without accountability. True.

These are not reasons to preserve the old model untouched. They are reasons to build the new model properly.

The answer to complexity is not pretending the world is simple. It is better architecture.

The answer to technology risk is not technological avoidance. It is governance, telemetry, auditability, and human-owned accountability.

The answer to manager resistance is not to accept opacity as the permanent price of access. It is defining which relationships are truly strategic, what data is necessary, what information is proprietary, and what level of telemetry is appropriate for the mandate.

The answer to a changing investment universe is not to chase every new market. It is to build an institution capable of understanding when a new market becomes investable.

The permanent-capital intelligence platform

The pension fund of the future is not a fund-of-funds.

It is not a hedge fund.

It is not a crypto vehicle.

It is not an AI science project.

It is a permanent-capital intelligence platform.

Its edge will not come from pretending to be faster than every trading firm or more technical than every venture fund. Its edge will come from combining long-term capital, fiduciary purpose, cross-asset visibility, elite external managers, internal judgment, data rights, programmable mandates, and total-portfolio discipline.

The old pension fund allocated capital.

The networked pension fund compounds intelligence.

And in a world where markets are becoming systems, contracts are becoming data, and economic activity is becoming programmable, that may become the most powerful investment model of all.

Sources

[1] CPP Investments Insights Institute, Investing in Uncertain Times: Achieving Disciplined Flexibility in the Total Portfolio Approach, Apr. 7, 2026. Source

[2] CalPERS Newsroom, CalPERS Board Adopts Streamlined Investment Approach to Seize Market Opportunities, Nov. 17, 2025. Source

[3] Bank for International Settlements, Annual Economic Report 2025, Chapter III: The next-generation monetary and financial system, Jun.2025. Source

[4] International Monetary Fund, How Agentic AI Will Reshape Payments, IMF Notes, Apr. 23, 2026. Source

[5] National Institute of Standards and Technology, Announcing the AI Agent Standards Initiative for Interoperable and Secure Innovation, Feb.17, 2026. Source

The views expressed are the author’s own and do not refer to any specific institution, portfolio, manager relationship, or internal strategy.