OpenAI is Playing Race To Becoming Too Big To Fail
Sam Lessin
Context:
Let’s be clear: the AI headlines have been absolutely relentless, and it’s no surprise there’s a bit more caution creeping into the market. From my seat as a GP at Slow, I see a sector that’s been running at full tilt—expectations through the roof, capital pouring in, and everyone trying to capture a piece of the narrative that AI is going to be the engine of American growth.
Market Signal:
We’ve seen this before, right? The “too big to fail” vibes are back—OpenAI’s recent comments about a government backstop were a real tell. It’s not just Silicon Valley anymore; Wall Street, sovereign wealth funds, the whole global capital stack is now in the game. When you see those players all crowding in, you know the narrative has gone mainstream—maybe too mainstream. And when high-risk investments suddenly seem safe because everyone’s doing them, that’s when you need to watch your step.
Takeaways:
The AI trade is real, but the hype is even more real. There’s substance here, but also a ton of narrative engineering—almost a national project to will higher GDP growth into existence via LLMs and AI.
We’re flirting with the same dynamics that led to the 2008 financial crisis. The government backstop talk, the “we’re all in this together” language—it’s all very reminiscent of the GFC, and that should make everyone nervous.
Social risk is as real as financial risk. The public remembers bailouts and inequality. If it looks like a handful of technologists are going to get insanely rich while everyone else foots the bill, expect a backlash.
As investors, we need to be more cautious, not less. Just because AI is now “obvious” doesn’t mean it’s de-risked. In fact, the crowding in makes it riskier. Writing bigger and bigger checks because everyone else is, that’s a recipe for disaster.
If you’re building and not using AI, you’re missing out. But let’s not confuse “AI is useful” with “AI will deliver 7% GDP growth.” There’s a gap between utility and macroeconomic transformation.