The AI Behind Claude Needs More Computer Power Than Big Tech Can Provide
Anthropic cannot rent enough computing power from Amazon, Google and Microsoft to keep up with demand for Claude, according to details from its S-1 filing reported by The Wall Street Journal.
An S-1 is the paperwork a company files before selling shares to the public. The filing says all three big cloud providers now see Anthropic as a competitor, not just a customer. That leaves its basic supplies in the hands of rivals. TechCrunch summarized the filing on Oct. 20, 2026.
Anthropic now says it could spend more than $750 billion on cloud contracts, including deals still being discussed. That replaces an earlier plan for at least $518 billion over ten years with six partners. Yahoo Finance reported that $518 billion plan in early October.
The capacity problem
Much of the $518 billion plan depends on deals that cannot be canceled, according to Reuters. Anthropic must pay even if needs change. In return, it gets promised access to AI chips, electricity and data center space.
To get more room, Anthropic is leasing its own data centers and working with outside data center companies. The filing warns those outside companies are new and may be less dependable than the big providers.
Big providers have years of practice with power, fast internal networks, fixing failures and shifting work to backup regions. Newer companies, often built fast to serve AI demand, may not have that depth. Anthropic is warning future shareholders about that gap.
It has been lining up other supplies for months. In April, it expanded work with Google and Broadcom and reserved several gigawatts of new TPU chips starting in 2027. TPU is Google's AI chip. Anthropic called it its largest computing deal to date. Anthropic
In May, it made a computing deal with SpaceX to add a large amount of capacity, and separately raised usage limits for Claude. Anthropic That same month, it raised $65 billion at a $965 billion value and said the money would pay for more computing for Claude. Anthropic
The financial picture
The filing lists $16.3 billion in first-half revenue and an $80.7 billion net loss. Two customers made up about one-fifth of sales in that period. That focus will get close attention before the IPO.
Growth is fast. Anthropic told investors third-quarter revenue alone was $16.8 billion, up 45% from the prior quarter. One quarter was larger than the whole first half.
It is aiming for a November IPO. Earlier reports pointed to sales talks starting in mid-October at the earliest and a stock debut as soon as mid-November at a value of up to $2 trillion. Yahoo Finance
Two other risks stand out. The filing says Anthropic could owe damages for harmful AI behavior beyond what insurance covers. In June, it took its two newest AI models offline after an export control order from the Trump administration. TechCrunch
Anthropic calls itself an AI safety and research company building "reliable, interpretable, and steerable AI systems."
The broader context here is one technology workers have seen before. Platform owners rent space to future rivals, then change the terms once the renter gets big. It happened with operating systems and apps, then with retail and market sellers, then with cloud hosts and software companies.
In my view, the $750 billion number is less a shopping list and more a message about bargaining power. Anthropic is telling investors it needs huge, steady computing, its current landlords cannot or will not provide enough, and it will sign firm deals with less proven suppliers while building chip programs of its own.
Worth flagging for planners, those risks connect. If growth slows, Anthropic still pays. If growth stays strong, its suppliers are also its competitors. If an outside site fails, there is no big-provider backup. The optimistic case here is the usual one for technology. Firm, long-term orders make it possible to finance and build new supply.
