The AI Oligarchy Is Already Here
Anthropic is spending $1.25 billion per month on compute just to stay in the game. Five hyperscalers will spend $690 billion on AI infrastructure this year. The companies that own the physical layer will control AI. Everyone else will rent access.
Anthropic filed for an IPO this week at a valuation approaching one trillion dollars. The headline sounds like a triumph of innovation. And frankly, It is. But buried inside the story is a detail that tells you where the AI economy is actually heading.
Anthropic's outlined why the company is going public. Not to reward early investors. Not to fund new research. To pay for computing power. The company is spending roughly $1.25 Bn per month on computing infrastructure alone. That number is not a rounding error. It is the cost of staying in the game and remaining relevant.
Apollo's Jim Zelter said it plainly this week. The AI capex boom is "phenomenal and unprecedented," and it is going to cause serious problems for smaller companies trying to compete. The five largest hyperscalers are on track to spend between $635 and $690 Bn on AI infrastructure in 2026 alone. The number compounds every year thereafter.
Therefore, the companies that control the power grids, the data centers, the chip supply chains, and the cooling infrastructure will control AI. Not because they are smarter. Because they own the physical layer that every AI model runs on. Everyone else will rent access to it at whatever price the infrastructure owners choose to charge.
The next generation of AI policy will not be about model safety or data privacy. It will be about who owns the pipes. That conversation needs to start now, before the pipes are owned by too few people to change.
