Five architects of the AI economy explain where the wheels are coming off

So imagine five people who build, fund, and run the actual machinery behind AI — not the chatbot front end you use, but the chips, the data centers, the electrical infrastructure — all sitting in a room admitting that things are getting complicated. That’s basically what happened at a recent conference in Beverly Hills. The honest takeaway? The people closest to the engine are worried about some pretty fundamental stuff: not enough chips to go around, power grids that can’t keep up with demand, and a quiet but growing concern that the basic technical approach powering today’s AI might have some serious ceilings on it. Think of it like building a massive highway system and realizing halfway through that you’re running out of concrete, the electrical grid can’t power the traffic lights, and maybe the cars themselves were designed wrong.

The chip shortage issue is the most immediate headache. Right now, the specialized processors that AI runs on are genuinely scarce, which means the companies with the deepest pockets are grabbing them first. Smaller players get squeezed. The data center problem is even wilder — the industry is now seriously discussing putting servers in orbit because we’re running out of places on earth with cheap enough land, water for cooling, and reliable power. And that last concern, about the underlying architecture being wrong, is the kind of thing that doesn’t make headlines but matters enormously. It’s the equivalent of someone in 1999 whispering that maybe the whole internet business model doesn’t actually work.

So what does any of this mean for your wallet? First, if you run a small business that uses AI tools, lock in your pricing now. Services like ChatGPT, Claude, or Midjourney are still subsidized and cheap. That won’t last forever as infrastructure costs rise — use them heavily while they’re practically free. Second, keep an eye on smaller regional cloud providers. As the big players fight over chips, scrappy competitors are offering serious discounts to grab customers. You can save real money on storage and computing by shopping around instead of defaulting to Amazon or Google. Third, if you have any investment budget, even modest amounts, the boring picks here are infrastructure-adjacent — electrical utilities near data center hubs, cooling technology companies, and grid modernization plays. The AI gold rush rewards the people selling shovels, not the ones panning for gold.

The people actually building this thing are nervous, and that’s probably the most useful information you’ll get all week.