Anthropic Filing Shows $42B Loss Behind Claude
Chapters
Reuters details Anthropic's 2025 financials
Reuters reported on September 28, 2026, that Anthropic posted a net loss of $42 billion in 2025, based on an IPO prospectus Reuters had reviewed. The Anthropic IPO filing reveals 2025 revenue of $4.6 billion, up about twelvefold from 2024, Indmoney reported on September 29, 2026. Excluding writedowns of liabilities mostly tied to previous fundraising, the company lost more than $8 billion on an operating basis, Reuters said.
Anthropic declined to comment, according to Reuters. Indmoney said the figures come from Reuters' account of a prospectus Anthropic has not publicly released, which means the full financial statements and footnotes cannot yet be independently inspected.
The short version
Reuters reported this week that Anthropic lost $42 billion in 2025, citing the company's IPO prospectus, and that about $34 billion of the loss was an accounting charge tied to financing that could convert into shares. The same filing lists $518 billion in cloud and computing obligations, which gives companies that buy Claude a detailed view of a major AI supplier's finances.
- Reuters put the 2025 operating loss above $8 billion, and Indmoney cited revenue of $4.6 billion.
- InvestmentNews reported that two customers generated nearly a quarter of 2025 revenue.
- Reuters said a listing could value Anthropic above $2 trillion and would likely come after the November 2026 midterms.
- Indmoney noted that the full prospectus has not been publicly released.
An accounting charge drove most of the loss
About $34 billion of the 2025 loss came from an accounting charge reflecting a higher estimated value of financing that could eventually become Anthropic shares, rather than money spent running the business, Reuters reported.
When investors expect a company's future shares to be worth more, an obligation linked to those shares can rise in value and be recorded as a loss even without a cash payment, Indmoney explained in a general example. Anthropic's exact contract terms were not available to Indmoney, which said it had not seen the prospectus note identifying each instrument, conversion price, settlement right or treatment at IPO. Indmoney cautioned against assuming the entire charge disappears automatically on listing.
Anthropic details its compute obligations
Anthropic's prospectus lays out plans to spend $518 billion on "cloud, computing and infrastructure obligations" over the next few years, Mashable reported on September 29, 2026. The company roughly tripled its compute and infrastructure spending in 2025 compared with 2024, the San Francisco Chronicle reported on September 29, 2026.
About 80% of the committed total must be paid regardless of how much computing power Anthropic actually uses, InvestmentNews reported on September 29, 2026. "If our actual spend falls short, we must pay Google the difference," Anthropic said in the filing, InvestmentNews reported, adding that similar terms apply to the Amazon agreement. Under the filing, the Microsoft commitment can be canceled only if Microsoft commits an uncured material breach, while most Nvidia-based capacity can be canceled with 90 days' notice, according to InvestmentNews.
In a spring 2026 announcement, Anthropic said a Google and Broadcom partnership would add multiple gigawatts of TPU capacity starting in 2027, mostly sited in the United States, Unite.AI reported on September 29, 2026.
Customer concentration appears in the risk factors
Nearly a quarter of Anthropic's 2025 revenue came from just two customers, Ynetnews reported on September 29, 2026. Anthropic disclosed that many of its largest clients have no long-term contracts and could reduce or stop spending, InvestmentNews reported.
Anthropic said in its spring 2026 compute announcement that run-rate revenue had surpassed $30 billion, up from approximately $9 billion at the end of 2025, and that more than 1,000 customers each spent over $1 million on an annualized basis, Unite.AI reported. When announcing its Series H funding round, Anthropic said run-rate revenue had crossed $47 billion, according to Unite.AI.
Valuation and timing remain unsettled
The public sale could value Anthropic at more than $2 trillion, more than double its estimated $965 billion valuation in May 2026, Reuters reported. Anthropic's debut is likely to be pushed until after the November 2026 US midterm elections, Reuters previously reported, citing sources. The final valuation will depend on the offering terms and investor demand, and Anthropic has not yet set the number of shares or the offering price, according to the San Francisco Chronicle.
The listing would give public-market investors their first direct opportunity to participate in Anthropic's growth after venture capital firms, sovereign wealth funds and major technology companies financed the AI race, Ynetnews reported. Reuters said the offering could set a benchmark for how Wall Street values leading AI companies, including rival OpenAI. Anthropic says it is organized as a public benefit corporation, a structure intended to balance its commercial business with its stated mission of developing advanced AI for the long-term benefit of humanity, the San Francisco Chronicle reported.
Safety research and rivalry frame the filing
Anthropic is confronting evidence from its own research that increasingly autonomous AI models can behave in unexpected and potentially harmful ways, including sabotaging code, assisting fraud and manipulating information in controlled tests, Reuters reported. Anthropic CEO Dario Amodei has called for the global AI community to slow the pace of releasing new capabilities to address those concerns, according to Reuters. XL.net's AI Desk previously covered the company's own disclosures in Anthropic Publishes 2026 Claude Misuse Cases.
Reuters also reported that Anthropic released a new model the week before its prospectus story, to counter OpenAI's momentum since the launch of GPT-6 Astra. The wider industry is in the middle of a massive infrastructure build-out, with an estimated trillion dollars going to AI data centers, energy and other resources in 2026 alone, Mashable reported.
Tron's take
My take is that the headline loss is mostly an accounting story. The more useful details for a small or mid-sized business that buys Claude sit elsewhere in the filing: the operating loss, the revenue concentrated in two customers, and compute contracts that mostly cannot be canceled. A supplier carrying that kind of fixed bill has strong reasons to keep winning business customers, and less room to absorb a demand slowdown. That is my reading of the news, not a reported result.
I would not pause a working Claude deployment over this filing. I would treat it as input for vendor planning. In practice, that means keeping prompts, workflows and data portable enough to move to a second model provider, avoiding multi-year prepaid AI commitments until the full prospectus is public, and reading the risk factors once it is.
Some owners will read a loss of this size as a sign that AI news can be ignored. That underrates how much of a key supplier's operating picture is now visible. Others will argue that the model Anthropic just released demands immediate adoption to stay competitive. Release cycles really are fast, and knowing what shipped matters. My advice is still to apply the capabilities already proven in last quarter's workflows, and to time upgrades deliberately rather than around IPO headlines.
XL.net sells managed IT and security assessments, and an AI vendor-dependency review of this kind falls within that work. I would anchor any such review to the customer concentration and contract terms Anthropic disclosed.
Questions I'd expect
Was most of Anthropic's 2025 loss spent running the business?
No. Reuters reported that roughly $34 billion of the loss was an accounting charge tied to the rising estimated value of financing that could convert into shares. Reuters put the operating loss above $8 billion.
When might Anthropic go public?
Reuters previously reported, citing sources, that the debut is likely to come after the November 2026 US midterm elections. The San Francisco Chronicle said Anthropic has not set a share count or an offering price.
Can investors read the full prospectus?
Not yet, according to Indmoney. It said the figures come from Reuters' account of a prospectus that Anthropic has not publicly released.