Anthropic has locked in $35bn of Nvidia-backed compute in Texas. That is not annual revenue and it is not a typo, it is the size of a single infrastructure commitment, and it tells you exactly where the AI arms race has gone. The fight stopped being about clever models a while ago. It is now a contest over who can secure the most silicon and the most electricity, and the sums have gone properly silly.
The numbers behind the headline
The Texas commitment does not stand alone. It sits on top of a broader compute-and-capital push: Anthropic has committed around $30bn to Microsoft Azure compute and up to one gigawatt of Nvidia AI hardware, while Nvidia is investing up to $10bn into Anthropic and Microsoft up to $5bn, a combined cheque of as much as $15bn. Financial media pegged that deal at an implied valuation near $350bn.
The scale is easier to grasp against Anthropic’s own growth. The company disclosed a $47bn run-rate revenue in mid-May 2026, rising to roughly $65bn by the end of July, after Dario Amodei described an 80x annualised jump earlier in the year. Its last confirmed private valuation was about $965bn from a Series H that closed in late May, and investors are reportedly circling a $2tn valuation for an IPO that could land as soon as October 2026.
| Metric | Figure |
|---|---|
| Texas compute commitment | $35bn (Nvidia-backed) |
| Azure compute commitment | ~$30bn + up to 1GW Nvidia hardware |
| Nvidia / Microsoft investment | up to $10bn / up to $5bn |
| Revenue run-rate (May → Jul 2026) | $47bn → ~$65bn |
| Last private valuation | ~$965bn (Series H, May 2026) |
| Reported IPO target | ~$2tn (as soon as Oct 2026) |
Why Texas, and why now
Cheap land, a permissive build regime and, above all, power. The binding limit on the next generation of AI is no longer the GPU order, it is the grid, and Texas has room and watts to spare. When you read “AI compute”, picture a substation, not a laptop. Locking in a gigawatt of hardware years ahead is a bet that demand keeps compounding fast enough to fill it.
The circular-money problem
Here is the part that should give a careful reader pause. When the chip supplier is also an investor in the customer, the deal starts to look circular: Nvidia sells the GPUs, helps finance the buildout, and books the revenue, while Anthropic commits to spending that only pays off if the boom keeps compounding. It is a bet on a bet. If demand keeps climbing, everyone looks like a genius. If it wobbles, tens of billions of contracted concrete and silicon become a very heavy thing to hold.
The investment read (not investment advice)
You cannot buy Anthropic directly yet, which shapes everything below.
Bull: a company going from a $30bn to a $65bn run-rate inside a year, with Nvidia and Microsoft writing cheques and an IPO reportedly targeting $2tn, is capturing demand at historic speed. If you want exposure now, the cleanest proxies are Nvidia (which sells the hardware and holds a stake) and Microsoft (Azure host and investor), both of which win if Anthropic’s spend materialises.
Bear: the $2tn IPO chatter values Anthropic at roughly 30x a $65bn run-rate that itself is barely a year old, in a market where the same handful of players fund each other. Circular financing flatters revenue on the way up and amplifies pain on the way down. A single quarter of softening enterprise AI demand could reprice the whole chain, Nvidia included.
Neutral: watch three things before the IPO: whether the revenue run-rate keeps climbing at anything like this pace, whether Anthropic actually fills the gigawatt of capacity it is booking, and whether independent (non-Nvidia, non-Microsoft) money shows up at the $2tn mark. If outside investors balk at that price, that is your signal.
What this means
The story under the $35bn is that frontier AI has become one of the most capital-intensive industries on the planet, funded by a tight loop of the same names. For a retail investor, Anthropic is a private bet you access secondhand through Nvidia and Microsoft, and the key risk is not that AI fails but that the money gets ahead of the demand. That concentration is efficient right up until the moment it is fragile, and nobody rings a bell at the top. (None of this is investment advice; do your own research.)
Related on Top Tool Stack: Nvidia’s Fine, the Market Sold Off Anyway · The Pentagon’s AI Rollout
Did you know: Anthropic’s reported ~$2tn IPO target would value it near the entire market capitalisation of some of the world’s largest banks, on a revenue run-rate barely a year old. The scale of this business has left the realm of ordinary intuition.