YWR: The Final Bottleneck

May 15, 2026

∙ Paid

Can you believe AMD is +130% in the last 6 weeks?

Well I know you do, because we have been calling for markets to make new highs since the Hormuz lows (4 reasons we make new highs, Insanity Trades), but I’m talking about the rest of the non-YWR world.

This upside crash in the NASDAQ is the last thing the bears expected.

I know it’s painful but take a moment to remember all the bear calls from the last 2 months. While events are fresh in our minds let’s remind ourselves of the key market lessons.

Lesson 1: At the bottom of the market the downside is always clear and logical. It’s why bears always sound smart and bulls seem naive (Stupidly Bullish). It’s because the problems are obvious while the solutions aren’t. The bulls might not sound smart but they make more money.

Lesson 2: Always take sentiment and positioning into account. The future is unknown, but the risk/reward of unexpected outcomes is highly dependent on how investors are already positioned.

Remember these lessons. We’ll come back to them.

Accelerating AI Growth

AMD gave a presentation this week on the growth of AI.

I attached a copy at the bottom of the post.

Their message is we’ve never seen anything like this.

AI usage reached 1 billion users in 3 years, while it took the internet 10 years. In the next 5 years we will add 4 billion new users. Take that in. Everything so far has been 1 billion users. We are about to add the rest of the world.

Compute demand is going exponential. Computation is growing so quickly, +100X, they had to invent a new word for it. ‘YottaFlops’. We go from 100 ZettaFlops in 2025 to 10 YottaFlops by 2030. It’s more users (+4 billion) all using AI more.

Customer growth is huge, but the real power trend is Enterprise adoption. Yes, lots of companies have people using chatbots, but AMD estimates only 5% of companies have moved to AI workflows. AMD has worked with AI-first companies and seen their usage, which makes them excited that enterprise AI usage is a big opportunity.

Enterprise spend on AI is only 4% of global enterprise software spend. $38bn out of $1 trillion.

Think about that.

Biggest tech trend ever and companies are still spending 96% of their IT budget on legacy things like Microsoft, software licenses, or whatever.

The gasoline on the AI trend is that it keeps getting smarter. AI is coding itself. Every 4 months we get new models able to do things previously thought unimaginable (images, coding, video, financial modelling).

And then around the corner is robotics.

So it’s super bullish and exciting.

And AMD is trading at 35x.

But it’s not just AMD.

This 100X growth to 10 YottaFlops drives everything.

It’s the story of the whole market.

Playing Bottlenecks

There’s a $700bn AI gold rush going on and everyone is scrambling to play it. Everyone is hunting for the bottlenecks.

Initially the bottleneck was GPU’s (NVDA, AMD, TSMC).

Then the bottleneck was memory (SK Hynix, Samsung, Micron).

As datacenter construction trend got underway the bottlenecks moved to the physical world. Utility grid connections, distributed power (CAT, GE Vernova, Siemens Energy, Mitsubishi Heavy), electrical equipment (Schneider Electric, Eaton), optical connectors (Amphenol, Coherent, Lumentum) and even human capital (electrical engineers).

GE Vernova made an illustrative comment that in the last 18 months there was a ‘run on the bank’ for turbines. Everyone had to have turbines.

These AI bottleneck stocks are going vertical.

But we should be cautious about this trend.

We are in the heat of the moment.

The big hyperscalers came out with 2025 earnings results and instead of guiding down 2026 Capex (which the bears had been expecting), they did the opposite. Capex budgets went even higher. Unbelievably, $500bn in spending went to $700bn.

General investor consensus is the hyperscalers are in an existential AI arms race and will continue to plough every dollar of free cash flow into new data centers. Spending could reach $1 trillion by 2030.

But the risk to all this bullishness is we have had a huge implosion in hyperscaler free cash flows from $224bn in 2023 to just $65bn in 2026. And the delta in investment from here could be a lot more moderate. Maybe spending grows much slowly, or even flatlines. Who knows the exact number, but it’s hard to repeat the explosive move in capex expectations from $200bn to $700bn.

So that’s a risk to be mindful of.

But let’s think ahead.

There is a construction surge underway.