Neeraj Bhargava

Chamath just pointed out one of the biggest problems with the entire AI industry. According to @chamath, his back of the envelope estimates suggest that just 1% of companies account for approximately 90% of all AI spending, and roughly 80% of that spending comes from unprofitable, venture-backed startups. We're constantly hearing about how AI adoption is exploding, how companies like OpenAI and Anthropic are generating enormous amounts of revenue and how demand for compute continues to exceed supply. But Chamath is questioning where a significant portion of that demand is actually coming from. Here's how he explains it. Venture capital firms raise billions of dollars from pension funds, family offices, and other institutional investors. They invest that money into AI startups, which then spend enormous amounts of capital purchasing access to AI models and computing infrastructure. The companies providing those services report impressive revenue growth, while the startups themselves use that technology to attract customers, grow their businesses, and raise even more money from investors. Essentially, you have a relatively small group of companies continuously raising money and spending it within the same ecosystem. The revenue being generated is real, but the question is whether that spending would continue at the same pace if venture capital funding suddenly slowed down. And this is where I think his argument gets particularly interesting. We're seeing frontier AI labs command valuations in the hundreds of billions, and in some cases potentially trillions, despite many of their customers being businesses that aren't profitable themselves. If those customers eventually struggle to raise additional capital, they could be forced to reduce spending, which would directly impact the revenue growth of the AI companies supplying them. Chamath compares this to the early days of cloud computing, when AWS initially relied heavily on startups purchasing cloud infrastructure. The difference is that Amazon already had a massive retail business supporting the broader company. Eventually, cloud adoption spread to profitable enterprises, governments, banks, and practically every other industry. AI needs to undergo a similar transition if today's growth expectations are going to be sustainable. But I've been saying for a while that I think some of these frontier AI labs are being valued way too aggressively. There's a big difference between believing AI will transform the global economy and believing every company developing AI models deserves a trillion dollar valuation. And the implications extend beyond the model developers themselves. Companies like $CRWV and $NBIS are investing enormous amounts of capital into infrastructure based on expectations of continued demand. If some of their customers eventually struggle financially, that could create problems even if the long term demand for AI remains strong. Personally, I think AI infrastructure will continue to be one of the biggest investment opportunities of the next decade. But I also think we're going to see a much bigger separation between companies generating sustainable revenue and those whose growth depends heavily on continually raising outside capital.

Today's "DeepSeek selloff" in the stock market -- attributed to DeepSeek V3/R1 disrupting the tech ecosystem -- is another sign that the application layer is a great place to be. The foundation model layer being hyper-competitive is great for people building applications.