Anthropic’s 2025 Compute Bill Exceeded Revenue by $2.7 Billion
Editorial Team
••📖5 min read
Anthropic’s reported 2025 compute spending outstripped sales despite rapid growth. Customer concentration and cloud partners with competing interests complicate its IPO case.
Anthropic’s reported $7.33 billion compute bill was about 160% of its nearly $4.6 billion revenue in 2025.
Revenue grew twelvefold while compute spending roughly tripled, but the company still reported an operating loss of about $8 billion.
Its cloud distributors are also investors, computing suppliers and AI competitors.
Anthropic’s spending on computing and cloud capacity to train and run AI models exceeded its entire 2025 revenue by about $2.7 billion. The historical gap exposes the scale of the business challenge behind the AI developer’s initial public offering ambitions: rapid sales growth must eventually cover infrastructure and the rest of its expenses.
The figures come from a September 30 account of a prospectus circulated among investors, which attributed the revenue and operating-loss figures to Reuters. It described a company seeking a $2 trillion valuation while planning hundreds of billions of dollars in future infrastructure spending. Those were reported ambitions, rather than a completed offering or money already spent.
Anthropic was founded in 2021 by researchers who left rival AI developer OpenAI over concerns about safe AI development. It develops the Claude AI system, including Claude Code, a tool that lets software developers delegate engineering tasks from their terminal. Anthropic introduced Claude Code as a limited research preview on February 24, 2025.
Magica chart of reported 2025 revenue and compute spending, which includes training and running AI models. Source: Los Angeles Times.
Revenue grew faster, but the gap remained large
The September account reported these 2025 figures from the prospectus:
UC Berkeley and NVIDIA are releasing tools for AI agents to assemble and improve robot programs in simulation. A warehouse operator reports a useful production update, while unsuccessful fixes and benchmark results show why testing, speed and existing infrastructure still matter.
Editorial Team
Computing and cloud capacity for training and running models
$7.33 billion
Total operating expenses
$12.65 billion
Operating loss
About $8 billion
Dividing the reported compute spending by revenue gives roughly $1.60 per dollar of sales. Subtracting revenue from compute spending gives the approximately $2.7 billion gap. Both calculations use rounded figures for the same full year.
The strongest counterweight is growth: sales rose much faster than compute spending. Spending therefore fell substantially relative to revenue even though its dollar amount increased. That improvement did not produce an operating profit in 2025.
The ratio also has limits. The reported compute category covers both training and running models; it does not isolate the expense of serving customer requests. It cannot establish the margin on an individual Claude subscription or a unit of model usage.
The September report separately put the company’s total on-paper loss at $42 billion. That is a different measure from operating loss and does not establish that Anthropic paid out $42 billion in cash during the year.
Cloud partners have several seats at the table
A September 29 report based on the confidential filing said Amazon and Alphabet’s Google routed 47% of Anthropic’s 2025 sales through their cloud marketplaces. They were simultaneously major investors, computing suppliers and AI competitors.
The report put marketplace sales at about $2.16 billion and calculated roughly $351 million in distribution fees, or about 16 cents per dollar of those channel sales. It said Anthropic records partner fees under sales, marketing and partnerships expenses. That percentage applies to marketplace revenue, rather than all company revenue.
The same account said about $3.8 billion of revenue came from usage-based payments for Claude, compared with $789 million from subscriptions. Anthropic said it expected usage-based sales to remain the substantial majority of revenue.
Magica chart of reported 2025 usage-based and subscription revenue. Payment models are distinct from cloud-marketplace distribution channels. Source: Reuters via Investing.com.
Concentration and safety complicate the expansion plan
The September 30 prospectus account said two customers generated nearly a quarter of 2025 revenue without long-term contracts. That concentration matters when evaluating the company’s reported plan to spend $518 billion on chips and data centers in the coming years. The plan rests on expected growth; the report does not supply a year-by-year spending schedule.
The sought $2 trillion valuation was roughly double the reported $965 billion valuation in the previous private funding round. A valuation is the price investors assign to the company, not the amount of fresh cash it would receive from an offering.
The same report said Anthropic’s prospectus devoted 80 of 261 pages to risk factors, including warnings that increasingly autonomous models could manipulate, blackmail or otherwise harm people. These are company disclosures about possible harms, not evidence that every described outcome has occurred. For a developer founded around safe AI development, those warnings also put product reliability at the center of its expansion case.
Anthropic’s company-reported business model links model improvement and growth to continued investment. Source: Anthropic S-1. Source: Anthropic, PBC - S-1.
What the next investment decision requires
The S-1 signature pages date the registration statement to October 26, 2026. The financial figures above describe 2025 as reported in September; they should not be read as a measure of current profitability.
Investors assessing the offering need subsequent financial results, the timing and terms of infrastructure obligations, and evidence that large customers keep buying. They also need to distinguish the economics of serving customers from the cost of developing future models. Whether sales growth can support the next spending commitments remains the consequential question.
Cloud Pathways lets customers coordinate large TPU workloads from one Python client, extending the system Google used for Gemini training. General availability on GKE follows earlier customer access; substantial reported usage and internal recovery results still leave customer cost savings unmeasured.
NVIDIA’s GR00T 2 predicts video and robot actions together to transfer skills across bodies. Company tests show gains from human video, but selected tasks, corrective data and recovery agents qualify its strongest success rates.
Anthropic says its new AI model produced circuit boards, enclosures and firmware for three prototypes that needed no electrical fixes. The demonstration avoided factory revisions, according to the company, while leaving certification, manufacturing scale and total costs unproven.
Meta is reportedly building a consulting channel for corporate AI sales after announcing an enterprise platform led by former MongoDB CEO Chirantan Desai. Its hiring remit also covers advertising and messaging, while prospective partners already work with rival AI suppliers.
OpenAI’s $200 Chromatic: Codex Edition adds AI-assisted development to an existing cartridge-game platform. The public bundle includes a programmable cartridge, but creation requires a computer, established tools and playtesting; the store now lists a November 19 restock.
PwC US plans a December launch for a critical-thinking and judgment certificate, with completion targeted by June 2027. It will assess staff through performance reviews as the firm seeks new ways to develop skills once practiced through routine work.
Linnea Quigley and Beverly Randolph say Aftermath departed from the musical they agreed to make, exposing a gap between a marketed cast reunion and performers’ control over their screen identities.
Anthropic’s Claude Fable 5.2 and Mythos 5.2 bring lower token prices and stronger technical performance. The release also limits capabilities that could accelerate frontier AI development, as new evidence puts the pace of research close to a sensitive policy boundary.
Meta’s smallest Muse model matches Claude Haiku 5.5’s rounded intelligence score at about one-third of the benchmark task cost. The case for delegating agent work depends on caching, task difficulty and higher charges for long requests.
SpaceX proposes a gas-fired power plant and super-compute facility on about 444 acres in Starbase, Texas. A December hearing will consider local incentives, while the public notice leaves the tax concession and operating demands unspecified.
The California state senator elected to Nancy Pelosi’s open seat wants a federal AI regulator, independent safety checks and penalties for developers. His California record shows both the reach and limits of that agenda; political support from an Anthropic-funded group adds scrutiny.
Apple Creator Studio subscribers can buy extra Logic Pro assistant usage for $49.99 in the U.S. The optional packs last up to a year, but Apple’s public support document does not quantify their capacity.
Boston Dynamics has fitted its electric Atlas humanoid with scooter wheels to test learned controls against real motor limits. The research feeds into the walking robot as Hyundai plans factory deployment, but the demonstration offers no quantified reliability gains.
Nominal raised $240 million at a $3 billion valuation, three times its March valuation, to expand its hardware engineering software into AI agents and lifecycle applications. Its defense foothold and company-reported workflow gains provide a foundation, but do not establish the newer agents’ performance or economics.
The former Megaspeed executive sought backing for a data centre project worth several hundred million dollars, citing GPU access and Alibaba business. The reported pitch adds commercial detail to scrutiny of her ties to Nvidia’s chief, while espionage and chip-smuggling allegations remain unproven.
President Donald Trump offered an unnamed executive’s alleged reversal as evidence that AI risks can be managed. His account identifies neither the conduct at issue nor the legal basis for a jail threat, leaving its value as a safety precedent unproven.
The Texas startup is building factory-made data centers for AI customers, with testing reportedly planned for early 2027. Its financing backs a manufacturing strategy whose cost and power-access advantages remain to be demonstrated.
Baidu says its automated research system has improved inference for the forthcoming ERNIE-5.5-Flash, alongside public benchmark gains. The results show bounded engineering progress, while the cost of finding those improvements remains undisclosed.
Nvidia’s shares fell as other chip stocks rebounded after its board opened an independent inquiry into reports of a foreign intelligence effort targeting CEO and co-founder Jensen Huang. The board retains confidence in him and says it has no reason to believe secrets were compromised; China rejects the spying allegation.
