America has bet everything on winning artificial intelligence, but it’s China who has cornered the four technologies that are necessary to convert electricity into profitable outcomes.
These include lithium-ion batteries, magnets and electric motors, power electronics, and embedded compute.
It’s the whole electric stack, if you will. And the cost of this stack has fallen 99% since 1990. China makes 75% of the world’s lithium-ion batteries and 90% of its neodymium magnets, which means it controls the means of production for EVs and robotics.
Why does that matter for an AI investor in May 2026?
Because it explains why the AI trade keeps grinding higher even as Main Street bleeds out.
In , he quotes economist Joel Spolsky’s old line: “Smart companies try to commoditize their products’ complements.” China is happy to commoditize intelligence because they own action. Translated to Wall Street: whoever owns the complement to the hot new thing captures the profits.
Right now, the hot new thing is AI compute. And the complements are in short supply.
Memory is in shortage. Optical interconnects are in shortage. Power delivery is in shortage. Networking equipment is in shortage. Every hyperscaler capex announcement makes the shortages worse. And the market has finally noticed, which is why Home Depot Inc. (HD) and Lululemon Athletica Inc. (LULU) are sitting at 52-week lows while SanDisk Corp. (SNDK), Applied Optoelectronics Inc. (AAOI), and Bloom Energy Corp. (BE) are up more than 400% year to date, and even higher over the past year:
The bottlenecks are not random, despite what you may hear. They are tradeable, they are predictable, and we now have a line on where the next one will be.
Let me show you where the chokepoints are, and how to position before the rest of the Street figures it out.
Click the video below to watch now:
The Consumer Is Bleeding Out. Why Are AI Stocks Ripping?
Because of the bottleneck.
Bloomberg Intelligence now projects AI capital expenditures growing at a 10% compounded annual rate through 2030, with peak annual spend hitting roughly $1.1 trillion.
Within that wave, the highest-growth slice, networking equipment and memory, compounds at 24%. The dot-com comparisons are real, by the way. There will be a bust eventually. But the dot-com boom was hopes and dreams. This one is hopes and dreams plus real revenue, real earnings, and a buildout that hasn’t peaked. We’re in the eighth inning, not the ninth.
And the smart money is doing exactly what Rockefeller did. It’s hunting bottlenecks.
That’s the entire investment thesis right now. Let me break down where they are.
Memory: The hate trade. Doubters have called the top five times and been wrong five times. They’ll eventually be right, but only when supply exceeds demand, and based on the capex curve above, that’s years away. SanDisk is my favorite single name. Micron Technology Inc. (MU) is my second. For diversified exposure, the VanEck Fabless Semiconductor ETF (SMHX) and the dedicated memory-focused DRAM-themed ETFs give you the basket without single-stock risk. Strategy here is simple: do not chase. These names go up 100% and pull back 40%. Wait for the flush.
Optics: This is where Nvidia Corp. (NVDA) just tipped its hand. Nvidia’s deal with Corning Inc. (GLW) tells you exactly where the next supply chokepoint is. Optical interconnects are the new tank cars. Nvidia isn’t building its own custom silicon to fight off the hyperscalers. It’s deploying its cash hoard to lock up the parts of the AI stack everyone else needs. That’s why I’m not bearish on Nvidia. I’m just more bullish on Corning, Coherent Corp. (COHR), Lumentum Holdings Inc. (LITE), Broadcom Inc. (AVGO), Marvell Technology Inc. (MRVL), and Fabrinet (FN). For the high-torque small cap, I like Applied Optoelectronics. Real orders, real hyperscaler scaling, no execution fairy dust.
The IPO signal: Cerebras Systems Inc. (CBRS) just priced its initial public offering at $185, well above its marketed range, after an order book oversubscribed by more than 20 times. A year ago, the same company pulled its filing. Two fears killed it: peak AI capex and Nvidia eating the entire chip market. Both fears are now demonstrably wrong. That’s why the IPO came back upsized rather than downsized. It’s a sentiment thermometer, and it’s reading hot.
The setup right now: technicals are stretched, earnings season is behind us, and a short-term pullback is the most likely next move. That pullback is your entry point, not something to chase.
This week’s full Being Exponential episode goes deeper on the memory rotation, the specific optics names worth front-running, and how to position for the back half of 2026 without getting flattened by a normal 10% correction. Watch the full episode of Being Exponential on or wherever you get your podcasts. Also, be sure to (formerly Twitter) for more exclusive content.
