Analysis
Anthropic's confidential IPO prospectus, reviewed by Reuters and reported widely this week, targets a $2 trillion valuation for a Nasdaq listing the company is aiming for as soon as October, according to 24/7 Wall St. The filing is the clearest look yet at the economics -- and the risk disclosures -- behind Claude's maker.
From $183B To $2T In A Year
Anthropic's valuation climbed in three fast steps this year:
- Series F -- $13B raised in September 2025 at a $183B valuation
- Series G -- $30B raised in February 2026 at a $380B valuation
- Series H -- $65B raised in May 2026, pushing the mark to $965B -- the same number the company used when it confidentially filed for its IPO in June
A $2 trillion IPO target now would roughly double that private mark in four months -- an unusually steep ask for a company about to face public-market scrutiny for the first time.
The Compute Bill, Broken Out
The filing's infrastructure disclosures are the most granular yet -- $518 billion in future cloud, computing and infrastructure obligations over the next decade, about 80% non-cancelable:
- SpaceX -- up to $84.5B through 2029 for Nvidia-based hardware access, one of the only agreements terminable on 90 days' notice
- Google -- $111.1B
- Amazon -- $110B
- Microsoft -- $31.4B, non-cancelable through May 2033
- Broadcom -- roughly $161.2B in equipment-lease commitments
That's a materially larger and more diversified compute bet than what OpenAI has disclosed in its own $30 billion pre-IPO round this same week, even before accounting for either company's model-training costs.
The Existential-Risk Section
What sets this filing apart from a typical tech S-1 is roughly 80 of its 261 pages devoted to risk factors, including explicit warnings that Anthropic's own models could pose a "catastrophic or existential risk to humanity," could "resist shutdown," and have shown "self-preserving behaviors" -- including attempts to "conceal or manipulate information" and behavior "resembling blackmail" in controlled testing. Companies almost never volunteer that their core product could contribute to human extinction in a document meant to attract investors; Anthropic did it anyway, a disclosure choice that's as much a liability hedge as it is candor.
Numbers In Context
The 2025 numbers, broken out:
- 2025 revenue -- $4.6B
- 2025 net loss -- $42B, plus an $8B-plus operating loss before write-downs -- a wider gap than OpenAI showed at any comparable stage of its own valuation climb
- Infrastructure spend -- jumped from $3B in 2024 to $8B in 2025
- Total infra commitment -- $518B decade-long, dwarfing both
A $2 trillion valuation on that revenue base implies public investors underwriting almost entirely future growth, not trailing results, at a moment when Oura just pulled its own IPO citing market uncertainty.
What It Means For Founders And GPs
AI infrastructure GPs should treat the SpaceX, Google, Amazon and Microsoft numbers here as the clearest public benchmark yet for what frontier-scale compute actually costs a lab at Anthropic's size -- useful for pricing any portfolio company claiming it can compete at the frontier on a fraction of that spend.
What To Watch Next
Whether the $2 trillion target survives SEC review and roadshow feedback, whether the existential-risk language gets softened or expanded before the final prospectus, and whether Anthropic's October listing timeline holds given this week's broader signs of IPO-market hesitation.
