YWR: The New-New Pt. 1 of 2.
I’m excited about a new stock idea.
Well, not exactly new.
We’ve discussed it before (the next Amazon), but I never pulled the trigger.
Now it’s even better than I thought.
And on the lows.
But first let me set the table.
Explain why it’s a big deal.
Why it’s the New-New.
And how we find other New-New’s.
First, I want you to stop thinking about ‘AI’.
Or, at least think differently.
The AI Bear Case
There are so many bear cases going around on AI. Why the $500bn/year in datacenters is mal-investment and going to take down the market.
Or, why OpenAI will run out of money and leave Oracle high and dry.
Here’s my view.
Maybe OpenAI makes money, maybe it doesn’t. Maybe it’s super overvalued and gets written down 80%.
Maybe Nvidia buys OpenAI for 20cts on the $ and runs the datacenters themselves. Maybe the US government buys it.
Who knows? But it’s a problem for the VC funds.
Not me and you.
As for Microsoft, Amazon, Oracle, Meta and Google. Is this $500bn/year in datacenters a good investment? Again, who knows. But I agree it doesn’t look great. And that’s what the market is sniffing out. Running compute for large LLM’s like OpenAI and Anthropic is not the high free cash flow network effect businesses which got them to where they are.
The market is sniffing out that this looks like it’s going end up a capital intensive low ROE business. Similar to telcos investing in ‘5G’. Remember how great ‘5G’ was supposed to be? Tons of money invested and it hasn’t changed the business. Yes, the amount of data we consume on our phones grew exponentially, but AT&T revenues didn’t. That’s the bear case of where we are going with the AI datacenter business.
But I don't think it’s a train smash.
Not for another 2 years at least.
My guess from the projects I’ve worked on is that Fortune 500 AI adoption is highly nascent. ‘IT departments’ in these firms are still caught up with questions of data security, how LLM providers might use their information, and how AI fits within the traditional architecture of users and roles. To make matters worse, most corporate ‘IT Departments’ these days are glorified Microsoft Office tech support. When back-end servers got outsourced to Azure it really dumbed down what it meant to be the ‘IT department’. And so building AI agentic workflows is not something they take to quickly.
Over the next 2 years (at least) we are going to have heavy corporate adoption of AI. Maybe, like Ed Zitron, you think AI is useless. I totally disagree, but it doesn’t matter. The corporate borg mind is going to implement it anyways. The train has left the station.
So we have a lot of adoption still to go. Which is why earnings earnings revisions for MSFT, AMZN, GOOG, META and ORCL are still positive.
Think about that.
Every analyst in the world knows about the $500 billion capex. Every analyst in the world knows costs are going up and that maybe ‘AI is a bubble’ and yet they are still raising estimates.
Which means the AI hardware plays can keep working too. Nvidia, AMD, the memory plays (SK Hynix, Samsung, Micron), all of Taiwan, likely still have more to go. Another leg.
But ‘another leg’ and ‘not a train smash’ is not super exciting.
It’s not the New-New.
Which is why we let everyone else watch the 100th podcast about AI.
We need to find a new path.
Appreciating what got us here.
Think back to what made Amazon, Apple, Meta, and Google great.
The internet in 2000 was a new thing. It created new business models. New monopolies. New network effects.
Monopolies and network effects are what make us rich.
We hunt for them. It’s all we think about.
Think about Apple. The iPhone is amazing. But why do you buy iPhone 17 for $1,500 even though it looks the same as iPhone 12 and costs 3x a Chinese phone? Because Apple created a network effect at the device level. For the first time it mattered if your phone, laptop, tablet and watch are all made by the same company.
Jobs said it best. ‘It just works’.
Network effects and a great product. It’s why you never buy the cheaper phones and Apple has 63% gross margins.
Paid Search. Another amazing never-would-have-thought-of-it business which came out of the internet. In the future companies will pay to be the name that comes up when you type their topic in a search bar..
Who would have thought that would be the biggest monopoly in the world?
In the future people will post their own content on the line and there will be things called ‘social networks’. And everyone in the world will be on one network which will be owned by a company called Facebook.
The internet was a new world with new business models we didn’t expect.
Yes, ‘connecting one computer’ to another was the technology but finding the new monopolies was what made you rich.
Which is how we are going to look at AI.
Where are the future network effects?
Hint. Is isn’t datacenters.