Why the US Is Not in an AI Bubble, Compared to ‘01 or ‘08
For all the talk about an “AI bubble,” the data tells a different story. Demand across consumer, enterprise, and sovereign levels is astounding—and it’s not speculative; it’s tangible, measured, and accelerating.
Follow the usage
CoreWeave’s CEO recently called AI demand “overwhelming,” a sentiment echoed by nearly every major player in the ecosystem. OpenAI’s GPT and now Sora usage have exploded to historic adoption levels, with hundreds of millions of consumers and creators actively engaging with these tools. This isn’t Pets.com; it’s people and businesses adopting AI at scale.
At Google, the numbers are staggering. The company reported nearly a quadrillion monthly tokens processed by Gemini models—up from 480 trillion just months prior. AI accelerator consumption has grown 15X in two years, and Google now handles over 5 billion AI-powered retail searches per month. This is real, sustained utilization, not speculative exuberance.
TSMC—the literal foundation of the AI economy—just reported record profits, up 39%, fueled by surging demand for AI chips. CEO C.C. Wei said it plainly: “Our conviction in the AI megatrend is strengthening.” Dell’s leadership echoed the same point: “AI demand is very solid.”
So yes, there’s an enormous amount of capital rushing into the space. That alone should make anyone cautious. History reminds us that “this time it’s different” is usually a red flag. But remember that during Britain’s railway mania, there were more rail lines in development than passengers. In the dot-com boom, companies went public with zero revenue. In 2008, home loans went to borrowers with no income.
The difference today? Utilization.
AI isn’t being built and ignored—it’s being consumed faster than it can be supplied. Until we have too many data centers, too much compute, or too much electricity to power them, it’s not a bubble. It’s a capacity race, or as Nvidia's Jensen Huang refers to as the next industrial revolution.
Venture vs infrastructure
Don’t mistake venture as a signal. Will some AI startups fail spectacularly? Of course. Right or wrong, the venture capital business model is designed for that: a few massive winning bets must offset mostly losers. Some companies will be absurdly overvalued; other overnight unicorns could be acquired just for their teams, see Meta's recent spending for elite AI talent. But beneath the froth lies a durable core: massive, compounding demand from every sector of the global economy.
If history rhymes, we’re not in the crash verse yet. Will there be corrections and downright atrocious investments along the way? Absolutely. But, we’re still in the infrastructure movement—the part where the U.S. rapidly rebuilds itself to keep up.
Disclaimer: I do not hold positions in any of the aforementioned companies but do hold positions in various publicly-traded companies related to AI, and I serve as an advisor to Salt AI.
