Anthropic's Own Numbers
Published: October 1, 2026 | Category: AI | By Mahesh
Anthropic reported a net loss of nearly $42 billion for 2025 in a draft IPO prospectus reviewed by Reuters, even as the company targets a valuation above $2 trillion for its public listing. According to CNBC's reporting on the document, roughly $34 billion of that total loss was an accounting charge reflecting a rise in the estimated value of financing that could eventually convert into Anthropic shares, not money the company actually spent running its business.[1] Strip that charge out and the operating loss, what the business actually lost running itself, was just over $8 billion, up from about $3 billion the year before.
Why a Company Can "Lose" Money It Never Spent
The mechanism behind the $34 billion charge is specific and worth explaining in plain terms, since it is the entire reason the headline loss figure looks so much larger than the business's actual cash burn. According to Yahoo Finance's detailed reporting, which reviewed the prospectus directly, the more valuable Anthropic gets, the more the early backers' convertible financing claims are worth on paper, and accounting rules require the company to record that rising value as a loss on its own books, even though no cash actually left the company.[2] Both Google and Amazon originally invested in Anthropic through convertible notes, debt instruments that convert into equity later, and as Anthropic's valuation has climbed toward its targeted $2 trillion IPO price, the accounting value of those conversion rights has climbed right along with it, flowing straight onto Anthropic's income statement as a paper loss.
The Part of This Story Forbes Named Directly
What makes this detail genuinely striking is not just that Anthropic's loss is mostly an accounting artifact, it is where the other side of that same artifact actually landed. According to Forbes' reporting, the identical rising valuation that produced Anthropic's $34 billion paper loss simultaneously boosted the recorded value of Google's and Amazon's own holdings in Anthropic, and both companies have already recognized real financial benefits from it.[3] Amazon's July earnings report showed net income jumping to $62.6 billion from $18.2 billion a year earlier, a gain Forbes' reporting attributes largely to markups on Amazon's Anthropic stake, while Google disclosed in its own most recent earnings statement that it held $124.3 billion in private startup holdings, a figure substantially inflated by its own Anthropic position. The same accounting mechanism that makes Anthropic look like it lost $42 billion is simultaneously making Google and Amazon look more profitable, a detail that complicates any simple reading of who actually benefited financially from Anthropic's valuation climb.
The $518 Billion Commitment That Makes This More Than an Accounting Story
Beyond the headline loss figure, the prospectus discloses a forward spending commitment of a scale that puts Anthropic's financial position in sharper relief. According to reporting from IJR, which reviewed the filing's specific contract terms, Google is owed at least $111.1 billion between April 2026 and July 2033 under Anthropic's cloud agreement, and Amazon is separately owed $110 billion between May 2026 and April 2036.[5] Roughly 80% of the full $518 billion commitment cannot be canceled, meaning Anthropic has locked itself into paying these amounts regardless of how its own revenue trajectory actually plays out over the coming decade. Against that scale of locked-in obligation, Anthropic held $20.28 billion in cash, cash equivalents and short-term investments at the end of 2025, according to multiple outlets reviewing the same filing.
This connects directly to a pattern Depth Grid covered in detail on September 21, when an internal OpenAI presentation projected $278 billion in cash burn through 2030. Reading the two disclosures together, both of the industry's two most valuable AI labs are now showing public evidence of spending commitments that run into the hundreds of billions of dollars, each locked in years ahead of the revenue growth that is supposed to eventually justify them.
Why Nearly a Quarter of Anthropic's Revenue Comes From Two Companies
The prospectus discloses a specific customer concentration detail that sits awkwardly alongside Google and Amazon's dual role as both investors and infrastructure suppliers. According to Augment's detailed review of the filing, the prospectus warns that nearly a quarter of Anthropic's revenue came from just two customers in 2025.[6] Fortune's own reporting on the broader $518 billion spending plan separately confirmed that Anthropic committed more than $100 billion over 10 years to Amazon Web Services specifically in April, with Amazon's own balance sheet carrying its Anthropic holdings at around $190 billion as of June 30, split between preferred stock and convertible notes valued mostly off the prices of Anthropic's own private funding rounds. Google and Amazon are therefore simultaneously two of Anthropic's largest shareholders, two of its largest infrastructure suppliers, and reportedly a meaningful share of its actual customer revenue, a three-way relationship concentrated in exactly two companies.
How the Revenue Growth Still Reads As the Genuinely Strong Part of This Story
It would be a mistake to read this filing as purely a story of financial strain, since the underlying growth numbers remain genuinely remarkable even accounting for the scale of the losses sitting alongside them. Reuters' own reporting, cited by Yahoo Finance, confirmed revenue reached nearly $4.6 billion in 2025, a twelvefold increase from the prior year, while compute and infrastructure spending tripled to $7.33 billion, more than half of the company's $12.65 billion in total operating costs. That spending pattern, infrastructure costs rising even faster than the business itself, is consistent with Depth Grid's own coverage of Anthropic's business trajectory earlier this month, and reflects a company still firmly in a capital-intensive buildout phase rather than one approaching the kind of operating leverage that would narrow its losses on its own.
Why the IPO Timing Still Points to a Specific Window
Augment's reporting notes the IPO is reportedly expected to proceed at a valuation above $2 trillion with a listing timed after the US midterm elections, a detail worth connecting to Depth Grid's earlier coverage of Anthropic's Accenture safety evaluator deal, which cited the same November IPO timeline. A valuation above $2 trillion would represent more than double Anthropic's own $965 billion private-market estimate from May, and multiple outlets note that gap reflects a reported target rather than a measured market change, since a private funding round and a public listing price are set through entirely different processes with different investor bases involved.
What to Watch Next
The clearest near-term signal will be whether Anthropic's final, publicly filed S-1 confirms the same $42 billion loss figure and $518 billion spending commitment this leaked draft shows, since the numbers in a confidential draft prospectus can still shift before the document becomes official. Also worth tracking closely is how public market investors ultimately price the distinction between Anthropic's $42 billion headline loss and its considerably smaller $8 billion operating loss once the company actually lists, since that gap, and whether investors treat it as a minor accounting footnote or a genuine source of valuation risk, will be one of the clearest early tests of how rigorously the market scrutinizes AI labs' complex, investor-linked financing structures.
Read More on Depth Grid
- OpenAI's Own Numbers Show It Runs Dry By 2028
- Anthropic Hired a $70B Consultant to Check Its Own Safety
- Anthropic's IPO paperwork is about to admit, in writing, that people don't want its data centers
- Nvidia Built a Kill Switch AI Agents Can't Touch
Article by Mahesh | Depth Grid


Discussion (0)
Comments are disabled for this post.