Artificial Intelligence

Anthropic Posts $11.5B Quarter, Its First Profit

Bloomberg reports Anthropic's Q2 2026 revenue reached $11.5 billion with positive operating income, beating the $10.9 billion forecast the company gave investors in May. Two subsidized cloud deals with Google and Amazon did much of the work.

Artificial IntelligenceBy Servola Tech Desk2026-08-154 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Anthropic Posts $11.5B Quarter, Its First Profit

Key takeaways

Bloomberg reported on August 14, 2026, citing people familiar with internal documents, that Anthropic's second-quarter revenue reached more than $11.5 billion, a fourteen-fold jump from $787 million a year earlier and well above the $4.73 billion booked in the first quarter of 2026. The company posted positive adjusted operating income for the first time, though none of these figures has been confirmed by Anthropic itself or audited.
The number beats what Anthropic told its own investors to expect. The Wall Street Journal reported on May 20, 2026, that internal projections shared during fundraising put second-quarter revenue at $10.9 billion and operating profit at $559 million, a forecast CNBC, PYMNTS and The Decoder corroborated within a day. The actual print, as reported by Bloomberg, came in roughly $600 million ahead of that projection.
The same May reporting traced the margin improvement to a falling compute-to-revenue ratio, from 71 cents per dollar of revenue in the first quarter to 56 cents in the second, driven by higher-margin coding workloads, tokenizer gains in Opus 4.7 that lifted realized prices 20 to 30 percent without a list-price change, and two large compute commitments: $40 billion with Google for TPU capacity and $33 billion with Amazon for AWS capacity.
Anthropic filed confidentially for an initial public offering targeting an autumn 2026 listing, working with Morgan Stanley, Goldman Sachs and JPMorgan Chase, the same week OpenAI made its own confidential filing. Both Google and Amazon are Anthropic equity investors as well as its compute suppliers, which means the two deals doing the most work on this quarter's margin are financing arrangements, not a standalone commercial cloud bill, and an autumn prospectus is where that distinction has to survive scrutiny.

A Number Anthropic Has Not Confirmed

Every account of Anthropic's second quarter starts with the same caveat: the $11.5 billion figure comes from Bloomberg's sources inside the company's own documents, not from a statement Anthropic issued or an audited filing. That distinction matters more than usual here, because Anthropic is mid-way through a confidential IPO process, and internal numbers shared with prospective investors are, by definition, meant to be persuasive before they are meant to be exact.

What is not in dispute is the shape of the trend. Revenue went from $787 million in the second quarter of 2025 to $4.73 billion in the first quarter of 2026 to more than $11.5 billion in the second quarter, a run of growth few software businesses of any kind have matched, and the company recorded positive adjusted operating income for the first time in the same period.

Where the Margin Actually Came From

The Wall Street Journal's earlier reporting, corroborated by CNBC, PYMNTS and The Decoder within a day of publication, gives the clearest account of how the margin moved. Anthropic's compute cost fell from 71 cents to 56 cents for every dollar of revenue between the first and second quarter, a 15-cent swing in a single quarter that would be remarkable for any infrastructure-heavy business.

Three things drove it: a growing share of revenue from coding products billed per seat or per task, which carry thicker margins than raw inference; a tokenizer change in Opus 4.7 that raised the number of tokens processed per request by roughly 47 percent, letting Anthropic lift realized prices 20 to 30 percent without touching its published per-token rate; and two multi-year compute commitments, $40 billion with Google for TPU capacity and $33 billion with Amazon for AWS capacity, both struck as part of funding rounds in which Google and Amazon also took equity.

The Question an Autumn Prospectus Has to Answer

Google and Amazon are not neutral vendors here. Both are Anthropic shareholders as well as its two largest compute suppliers, and the compute commitments behind this quarter's margin gain were negotiated as part of the same funding rounds that gave them their stakes. That does not make the $11.5 billion fake, but it does mean a chunk of the margin improvement reflects the pricing an investor-supplier is willing to offer, not the price a standalone enterprise customer would be quoted for the same capacity on the open market.

For anyone evaluating the autumn listing, that is the number worth isolating before the headline profit figure sets a valuation multiple: what would this quarter's operating income look like if the TPU and AWS capacity were billed at the rate Anthropic's own paying customers face, rather than the rate its cloud-investor-suppliers extended it. Bloomberg's report notes that OpenAI's own annualized run rate already tops $40 billion, though the two companies may not calculate the underlying metrics the same way, a reminder that quarter-to-quarter comparisons across labs are harder to make than the headline numbers suggest.

What Has to Happen Before Autumn

A confidential filing is not a prospectus, and Morgan Stanley, Goldman Sachs and JPMorgan Chase still have to turn investor-facing projections into audited financial statements that a public listing requires. That process is exactly where informal, sales-oriented numbers meet accounting rules, and it is common for the two to diverge, sometimes by a little and sometimes by a lot.

Anthropic filed in the same week as OpenAI, and the coincidence sets up a natural comparison the market will draw whether either company wants it or not: two frontier labs, both claiming a path to profitability, both leaning on cloud-provider relationships that mix vendor and investor in ways a public-market income statement has never had to disclose before. Whichever prospectus lands first will set the template the other is measured against.

Frequently asked questions

How much revenue did Anthropic report for the second quarter of 2026?

Bloomberg reported more than $11.5 billion, citing people familiar with internal company documents. Anthropic has not issued its own statement confirming the figure, and it has not been audited.

Is this the first time Anthropic has been profitable?

According to Bloomberg's reporting, Anthropic recorded positive adjusted operating income in the second quarter of 2026 for the first time. The company had been loss-making on that measure in prior quarters, including a near-break-even first quarter of 2026.

Why did the Wall Street Journal report a lower figure in May?

The Journal's May 20, 2026 report described an internal projection shared with prospective investors during fundraising, of $10.9 billion in revenue and $559 million in operating profit. The actual second-quarter result, as later reported by Bloomberg, came in above that forecast.

What role do Google and Amazon play in Anthropic's margin improvement?

Both companies are Anthropic equity investors and its two largest compute suppliers. Compute commitments worth $40 billion with Google and $33 billion with Amazon, agreed alongside funding rounds in which both took stakes, are among the factors reported to have cut Anthropic's compute-to-revenue ratio from 71 cents to 56 cents per dollar.

An unaudited number leaked from internal documents is not proof that large language models have solved their unit economics, and the two biggest levers behind this quarter's margin gain were negotiated with a company's own investors, not with the open market. What the number does prove is that Anthropic's growth curve has not bent, and that the autumn prospectus race with OpenAI is now a contest over whose profitability survives being priced at market rates rather than investor rates.

AnthropicClaudeIPOAI InfrastructureCloud ComputeGoogleAmazon

Anthropic Unblocks Biology AI, Keeps Bioweapons Locked

Anthropic rewrote and retrained Fable 5's biology safety classifier, cutting false-positive refusals on benign biology and medical queries by about 85%, while virology and toxicology stay hard-blocked.

5 min read

Cybersecurity

Hidden AI Reasoning Leaked Via Cheaper Sibling Models

Researchers decoded 315,320 encrypted reasoning blocks from public AI transcripts and recovered 367 personal-data artifacts and 182 credentials - proof a flagship model's private reasoning is only as safe as its cheapest sibling.

5 min read

Decisions

The Third Pivot Was the Warning Magic Leap Missed

Magic Leap filed a WARN notice for 193 layoffs effective 1 October 2026, its fourth strategic pivot after $4 billion raised. The lesson is the pattern, not the pivot.

4 min read

← All articles