Artificial intelligence · News
Anthropic's $80.7 Billion Loss Exposes the Real Cost of Building AI Models
Published:

In brief
Anthropic reported a net loss of $80.7 billion for the first half of 2026, a draft IPO prospectus reviewed by Reuters shows. Most of that figure — about $72 billion — is a non-cash accounting charge tied to convertible notes, but the company's actual computing costs also surged, pointing to the huge capital outlays behind frontier AI development.
Anthropic, the maker of the Claude chatbot, disclosed a net loss of $80.7 billion for the first six months of 2026 in a draft prospectus for a planned initial public offering, according to Reuters, which reviewed the filing. The loss widened sharply from $13 billion in the same period of 2025.
The company's revenue also grew rapidly over the same period, rising from $1.1 billion in the first half of 2025 to $16.3 billion in the first half of 2026, Reuters reported citing the prospectus.
Where the $80.7 billion came from
The bulk of the headline loss is not tied to day-to-day operations. Reuters reported that about $72 billion of the first-half loss reflects an accounting charge from a rise in the estimated value of convertible notes — bonds that could eventually convert into Anthropic shares. As the company's valuation has climbed, the fair value of those notes has risen too, and accounting rules require that increase to be booked as a loss even though no cash changes hands.
Olhar Digital, citing a Wall Street Journal analysis of the same filing, reported that Anthropic's full-year 2025 net loss of $42 billion was likewise driven mainly by this kind of non-cash accounting adjustment tied to the convertible notes, rather than by cash operating costs.
Stripped of that accounting effect, Anthropic's underlying business still lost money. For full-year 2025, Reuters reported an operating loss of about $8.1 billion on roughly $4.6 billion of revenue, excluding the fundraising-related liability remeasurement.
The real driver: computing costs
Even excluding the accounting charge, the cost of training and running AI models is rising fast. According to Olhar Digital's reporting on the prospectus, Anthropic spent nearly $9 billion on computing capacity and infrastructure in the first half of 2026, up from $2.3 billion in the same period of 2025 — and already higher than the $7.3 billion it spent in all of 2025.
Those costs are set to keep climbing. Yahoo Finance's review of the prospectus detailed Anthropic's long-term cloud and chip commitments: $111 billion to Google Cloud, $110 billion to Amazon Web Services, $31 billion to Microsoft, up to $86 billion for Nvidia GPU leasing through firms including Nscale and Lambda, and over $160 billion for Google-designed chips supplied via Broadcom. Total noncancelable commitments were put at roughly $450 billion, with as much as $150 billion more under negotiation — figures RTÉ's report on the Reuters filing rounded to a decade-long obligation of about $518 billion.
Revenue is growing faster too
Anthropic's business is scaling alongside its costs. Reuters reported that second-quarter 2026 revenue more than doubled from the first quarter to $11.6 billion, and that the company told prospective IPO investors its third-quarter revenue jumped 45% from the previous quarter to $16.8 billion.
| Metric | H1 2025 | H1 2026 |
|---|---|---|
| Revenue | $1.1 billion | $16.3 billion |
| Net loss | $13 billion | $80.7 billion |
| Computing/infrastructure spend | $2.3 billion | ~$9 billion |
Why this matters for the IPO
Anthropic is reportedly aiming for a stock market listing around November 2026. RTÉ's report on the Reuters filing said some investors were weighing a lower valuation of around $1.5 trillion following the latest disclosures, while Yahoo Finance's review of the prospectus cited discussions of a valuation of up to $2 trillion. RTÉ noted that the company's path to a listing is being watched as a signal for the broader AI sector.
Yahoo Finance reported that the prospectus devotes close to a third of its content to risk disclosures, including warnings about AI models exhibiting what the filing describes as self-preserving behaviors or generating manipulative responses, and potential legal liability from harm caused by its models.
What is known and what is not
- Known: Anthropic's H1 2026 net loss was $80.7 billion, per the draft IPO prospectus reviewed by Reuters.
- Known: About $72 billion of that loss is a non-cash accounting charge tied to convertible-note valuations, not operating cash costs.
- Known: Computing and infrastructure spending nearly quadrupled year over year to about $9 billion in H1 2026.
- Not yet known: The final IPO valuation, listing date, and whether the company's operating losses (excluding accounting charges) will narrow as revenue scales.
Questions
- Did Anthropic actually lose $80.7 billion in cash?
- No. Reuters reported that about $72 billion of the $80.7 billion first-half 2026 loss was a non-cash accounting charge reflecting a rise in the estimated value of convertible notes, not an operating cash loss.
- How much is Anthropic spending on computing?
- According to the draft IPO prospectus cited by Olhar Digital and Reuters, Anthropic spent nearly $9 billion on computing and infrastructure in the first half of 2026, more than its full 2025 total of $7.3 billion.
- How fast is Anthropic's revenue growing?
- Reuters reported revenue rose from $1.1 billion in H1 2025 to $16.3 billion in H1 2026, with third-quarter 2026 revenue reaching $16.8 billion, up 45% from the previous quarter.
- When is Anthropic planning to go public?
- RTÉ's report on the Reuters filing cited an expected listing around November 2026, with some investors weighing a valuation of around $1.5 trillion, while Yahoo Finance cited discussions of a valuation of up to $2 trillion.
- How much has Anthropic committed to cloud and chip providers?
- Per Yahoo Finance's review of the prospectus, commitments include $111 billion to Google Cloud, $110 billion to AWS, $31 billion to Microsoft, and up to $86 billion and over $160 billion respectively for Nvidia and Google-designed chips, totaling roughly $450 billion in noncancelable commitments.
How the world is telling it
- PortugueseOlhar Digital's coverage, drawing on a Wall Street Journal analysis of the prospectus, framed the story around the scale of the headline loss, emphasizing that most of the deficit was an accounting effect while still highlighting the multibillion-dollar real compute spending as evidence of the capital intensity of AI development.
Sources
- Olhar Digital — Anthropic revela prejuízo de R$ 435,8 bilhões e custos bilionários com IA
- RTÉ (Reuters) — Anthropic logged net loss of over $80 billion in H1, 2026
- US News (Reuters) — Anthropic Recorded Net Loss of Over $80 Billion in First Half of This Year, WSJ Reports
- Yahoo Finance — Anthropic's New Prospectus Shows Revenue Is Growing Fast, So Are Its Compute Costs, Losses And AI Risks
This piece was prepared with artificial intelligence and passed an automated source check before publication. If you see an error, ask for a correction. Request a correction or reply · How we make this
Related

Artificial intelligenceArticle
How Neural Networks Learn: The Math Behind ChatGPT and Gemini
Neural networks learn by making a prediction, measuring how wrong it was with a loss function, and then adjusting internal values called weights backward through the network — a process called backpropagation paired with an optimisation method called gradient descent. The same mechanics, scaled to billions of parameters, underlie how systems such as ChatGPT and Gemini are first trained and then adapted to specific tasks through fine-tuning.

Artificial intelligenceArticle
How AI Generates Art and Music, and Why "True Creativity" Remains Disputed
AI image tools like Midjourney and Stable Diffusion build pictures by learning to reverse a noise process, while music tools like Suno and Udio compress audio into tokens and predict them step by step, guided by a text prompt. Whether either process counts as genuine creativity is unresolved: researchers are split between calling it pattern-based "remixing" and crediting it with measurable creative performance.
Artificial intelligenceArticle
Why Moonshot Deliberately Weakened Its New Kimi AI Model's Hacking Skills
Chinese AI lab Moonshot says it removed cyber-related training data from its new Kimi K3.1 Pro model so the open-weight system would not push the "frontier" of offensive hacking capability. The move, announced Saturday, October 24, 2026, comes as governments and security researchers warn that increasingly capable AI could make sophisticated cyberattacks cheaper and easier to carry out.
Artificial intelligenceArticle
How AI Text Detectors Work, Explained Through a Prize-Winning Novel Dispute
AI text detectors do not prove who wrote something; they estimate a probability by measuring statistical patterns such as word predictability and sentence-length variation. A detector called Pangram 4.1 recently attributed 99% of a €20,000 prize-winning Spanish-language novel to artificial intelligence, prompting its publisher to suspend promotional events while it investigates.