Anthropic, the company behind Claude, has signed a six-year deal worth $10 billion to buy computing capacity from Volta Infra Holdings, an AI infrastructure startup barely a few months old and backed by Nvidia. The information was reported by Bloomberg and confirmed by people close to the negotiation, in a move that underscores how the race for chips and data centers has stopped being merely a battle of models and become a war over infrastructure.
What stands out is not only the size of the check — $10 billion is among the largest compute orders ever placed by an AI lab — but who is on the other side. Volta was founded in early 2026 by former Brookfield Asset Management executives and counts Nvidia among its investors. In other words: a company that has barely completed its first year of existence is selling, in one shot, more than a decade of computing capacity to one of the leading names in artificial intelligence. It is as if a newly formed construction firm signed a contract to raise the world's tallest building before finishing its first house.
To put this in context, Anthropic has struck a series of cloud deals in recent months. The logic is simple: training and operating models such as Claude demands thousands of cutting-edge GPUs, and no company can build data centers fast enough to keep pace with the labs' appetite. The solution has been to rent capacity from third parties — from hyperscaler giants to so-called neoclouds, startups specialized in AI computing. Volta represents exactly this new category, and Nvidia, by investing in it, creates a parallel market for its own GPUs.
The deal, however, does not come without risks. Relying on such young infrastructure means betting that Volta will deliver the chips, energy, and operations it promised. If anything goes wrong — installation delays, cooling problems, or volatility in the power supply — Anthropic becomes exposed to a critical bottleneck at the very heart of its operation. Moreover, $10 billion spread over six years is a heavy financial commitment for a company that still needs to prove its revenue model can sustain that scale.
The broader implication is that AI infrastructure has become the most contested asset of the moment. Labs that once competed for talent and benchmarks now compete for megawatts and chip access. Nvidia, by financing neoclouds like Volta, creates an ecosystem in which its GPUs keep circulating even as cloud giants try to diversify suppliers. And the question that lingers is whether the AI computing market is large enough to sustain so many players promising the same thing, or whether we are facing an infrastructure bubble about to burst.
There is also a strategic dimension worth attention. By spreading its computing contracts across several providers — including cloud giants and now neoclouds like Volta — Anthropic is trying to avoid becoming hostage to a single partner, something that has already cost other AI companies dearly in the past. This diversification is a bet on redundancy: if one supplier fails, the operation does not stop. The same reasoning that led Apple to develop its own chips so as not to depend on a single manufacturer now appears at the AI infrastructure layer. The difference is that here the capital involved is far larger and the payback horizon uncertain. With $10 billion committed, the question that lingers is whether these long-term contracts will prove to be a growth anchor or a burden that eats into the company's profits over the coming years, especially if the AI investment cycle cools earlier than expected.
Sources: TechCrunch, Yahoo Finance (Bloomberg), FourWeekMBA, Venture Atlas
✓ Independent sources cross-checked and verified before publishing
