Picture yourself at a massive tech conference. There are 6,000 startup booths spread across the convention center floor. Founders are pitching, investors are circling, and checks are being written. By the end of the quarter, $300 billion has changed hands.
Now imagine walking past booth after booth after booth—5,976 of them, to be exact—and discovering they collectively received $122 billion. Not bad, right? But then you reach a small VIP section in the corner. Just 24 booths. And they walked away with $178 billion.
This actually happened in Q1 2026. And it tells us something important about where AI investment has gone completely sideways.
The Math That Doesn’t Add Up
Let’s break this down in human terms. The average company outside that VIP section raised about $20 million. Meanwhile, each of those 24 foundational AI companies averaged $7.4 billion. That’s not a typo. We’re talking about 370 times more money per deal.
If you’re wondering what “foundational AI” means, think of it as the companies building the base layer—the big language models, the core infrastructure, the fundamental technology that other AI tools sit on top of. Companies like OpenAI, Anthropic, and their competitors.
Why This Should Make You Nervous
Here’s what worries me about this concentration of capital. When this much money flows to this few companies, we’re not really investing in AI anymore. We’re placing massive bets on a handful of horses and ignoring the rest of the race.
Consider what $178 billion could fund instead:
- 8,900 companies at $20 million each
- Thousands of specialized AI applications for healthcare, education, or climate
- Regional AI development outside Silicon Valley and a few other tech hubs
- Smaller teams working on AI safety and alignment problems
But that’s not what’s happening. Investors are convinced that building bigger models requires bigger checks, and that only a few players can win this race. So they’re writing checks with more zeros than ever before.
What Happens When The Music Stops
Every investment boom eventually faces a reckoning. The question isn’t if, but when—and what it looks like when $178 billion worth of expectations meets reality.
Some of these 24 companies will succeed spectacularly. But statistically, most won’t deliver returns that justify their valuations. And when investors realize they’ve concentrated too much capital in too few hands, the correction won’t be pretty.
The real tragedy? All those other applications of AI that didn’t get funded. The medical diagnostic tool that needed $5 million. The educational AI that wanted $10 million. The climate modeling system seeking $15 million. They’re still out there, still needed, but they’re competing for scraps while foundational AI feasts.
A Different Way Forward
I’m not arguing against funding foundational AI research. These companies are doing important work, and yes, training large models costs real money.
But $178 billion in one quarter for 24 companies isn’t a sustainable investment strategy. It’s a mania. And manias always end badly for someone—usually the people who arrive late to the party.
What we need is a more balanced approach. Fund the foundation, sure. But also fund the applications. Fund the safety research. Fund the alternatives. Fund the teams working in languages other than English, in countries other than the United States, on problems that don’t involve making chatbots slightly better at writing marketing copy.
Because right now, we’re building an AI future that’s top-heavy, concentrated, and fragile. And that should concern anyone who cares about where this technology is headed.
đź•’ Published: