Anthropic has spent months building toward one of the largest public offerings in history, and its confidential paperwork just revealed exactly how much that ambition is costing behind the scenes.
The company has kept its IPO timeline flexible, with Reuters reporting the debut is likely to be pushed to after the November U.S. midterm elections.
The number at the center of the filing is big enough to rank among the largest infrastructure commitments any company has ever disclosed.
What makes it notable is not just the size of the spending, but how little flexibility Anthropic actually has to walk any of it back if its own growth ever slows down.
Inside Anthropic’s $518 billion commitment
Anthropic told investors in its confidential IPO prospectus that it expects to spend at least $518 billion over the next decade building AI infrastructure with six partners. The plan ranks among the largest AI buildout commitments on record, according to CNBC.
The company disclosed that about 80% of that total is non-cancelable or requires payment regardless of actual usage. It means Anthropic is locked into paying most of the bill, even if demand for its models comes in below expectations.
The individual pieces are enormous on their own. Anthropic owes Google at least $111.1 billion between April 2026 and July 2033. Amazon is owed $110 billion between May 2026 and April 2036. Microsoft is owed $31.4 billion between November 2026 and May 2033.
More AI:
- Nvidia just made a move Wall Street wasn’t ready for
- Microsoft just took sides in AI policy fight
- OpenAI just disclosed something genuinely alarming
The Microsoft commitment can be canceled only in the event of Microsoft’s uncured material breach. The Google and Amazon agreements both include a must-pay clause. If actual spend falls short, Anthropic must pay the difference.
Anthropic is also carrying about $161.2 billion in Broadcom-related equipment lease obligations that are largely non-cancelable.
Anthropic framed all of this as a necessary bet. The company told investors that future demand for advanced AI systems is likely to exceed available supply and will be limited principally by the availability of compute, Reuters reported.
The SpaceXAI and AMD deals add more layers
Not every agreement in the Anthropic filing carries the same rigid terms. It disclosed agreements with SpaceXAI that could result in up to $84.5 billion of spending on Nvidia-based computing capacity through 2029.
Those commitments are largely cancelable with 90 days’ notice, offering a much softer structure than most of its other deals.
The filing also detailed a deepening relationship with AMD. The chipmaker committed to supply AI computing capacity expected to exceed $20 billion, building on an earlier $5 billion computing agreement the two companies struck in July, according to TheStreet.
Anthropic has also leaned on SpaceX for compute. Anthropic agreed to pay $1.25 billion per month for computing capacity under a deal that SpaceX disclosed as running through May 2029.
That arrangement alone amounts to roughly $15 billion a year from a single supplier, Reuters reported.
Why Anthropic is building its own infrastructure
The company told investors it is increasingly shifting away from a cloud-only model toward dedicated data centers and directly leased chips. This is a strategic pivot toward gaining more direct control over its infrastructure, rather than renting it from outside partners.
That shift comes with a risk Anthropic itself flagged directly. The company depends on Amazon, Google, and Microsoft for computing infrastructure and distribution. All three are simultaneously developing their own rival AI models. It’s a structural conflict of interest built into some of Anthropic’s most important partnerships, Reuters reported.
Those same partners have enormous financial exposure to Anthropic’s IPO. Amazon has committed more than $100 billion over 10 years to AWS technologies. It also secured up to 5 gigawatts of Trainium chip capacity and committed to invest up to $20 billion more beyond its original stake.
What it means for the Anthropic IPO
Anthropic filed its draft registration statement with the SEC in June. Investors currently expect the company to go public at a valuation near $2 trillion, which would exceed SpaceX’s $1.77 trillion valuation at its record-setting June IPO, Reuters reported.
The financials behind that valuation are striking. Revenue rose 12-fold to nearly $4.6 billion in 2025. But Anthropic lost $42 billion on a net basis and more than $8 billion at the operating level. The company spent $7.33 billion on compute and infrastructure last year alone, up threefold from 2024.
Anthropic’s filing offers a rare, detailed look at just how locked in these infrastructure bets really are. The $518 billion figure is not a spending target. Anthropic can simply scale back if growth disappoints. It is a bill that comes due regardless.
That reality will likely shape how public market investors price the company once its shares finally start trading. The IPO will be the first time Wall Street can put a public number on what this level of locked-in spending is actually worth.
