#Anthropic申请赴美IPO #每周来晒 The “most cash-burning” AI unicorn is about to go public! It spent $9 billion on compute in six months—how profitable is it really?

Anthropic formally filed an S-1 registration statement with the U.S. Securities and Exchange Commission (SEC) on October 26 Eastern Time, applying to launch an initial public offering of Class A common stock. It plans to list on the Nasdaq Global Select Market under the ticker “ANT.” The move has pushed Claude’s developer to the forefront of the U.S. stock market’s AI race, making it the first major foundational model company to pursue a public listing.

In stark contrast to Anthropic’s decisive move, its biggest rival, OpenAI, has clearly pushed back its IPO timeline. The two AI giants made opposite choices in the fall of 2026: one is opting to finance its enormous compute spending with the transparency and depth of capital offered by public markets, while the other is staying private and using a financing round of up to $30 billion to fill its pre-IPO funding gap.

I Targeting a Nasdaq Listing in Mid-November
Anthropic had previously confidentially submitted a draft listing application in June. The public filing on October 26 means the IPO process has entered a substantive phase. The company’s official statement confirmed that the S-1 covers a proposed offering of Class A common stock, with the ticker set as “ANT,” but the number of shares and price range have yet to be determined and will be disclosed in subsequent amendments. The IPO is being led by Goldman Sachs and Morgan Stanley, with around 20 other banks participating in underwriting, bringing the total to approximately 22 underwriters. Morgan Stanley is also acting as the price stabilization manager. According to several foreign media reports, Anthropic is targeting an IPO during the week of November 16, with the investor roadshow expected to begin during the week of November 9. In terms of its share structure, Anthropic has multiple classes of shares, including Class A, F, E, T and N. The seven co-founders, including CEO Dario Amodei and President Daniela Amodei, will retain 50.1% of the voting rights through a special Class F share, maintaining control of the company even after it goes public. Amazon and Google are its largest strategic shareholders. As of June 30, Amazon held $6.2 billion in convertible notes and could receive about one-fifth of the shares outstanding if they convert at the IPO price. Google currently holds about 7% and has the option to invest an additional $10 billion in non-voting shares at the IPO price.
Market expectations put Anthropic’s valuation between $1.5 trillion and $2 trillion. At recent investor events, some investors valued the company at around $1.5 trillion, below the previous expectation of $2 trillion, mainly because quarterly revenue growth has slowed somewhat. Even at $1.5 trillion, that would represent a premium of about 55% over Anthropic’s $965 billion valuation in its Series H funding round in May this year.
If the final price lands at the top of the range, the IPO’s proceeds and valuation could challenge or even surpass the record set by SpaceX this June, when it raised $86.25 billion at a valuation of about $1.77 trillion.

II Revenue Explodes Alongside Massive Book LossesAnthropic’s S-1 filing presents a set of sharply contrasting figures. Anthropic generated $16.286 billion in revenue in the first half of 2026, up about 1,300% from approximately $1.16 billion in the same period last year. Revenue in the first half alone far exceeded the $4.585 billion for all of 2025. Second-quarter revenue reached $11.555 billion. Third-quarter figures disclosed at an investor day showed revenue rising further to $16.8 billion, with the company recording adjusted profit and positive free cash flow for a second consecutive quarter.

At the same time, Anthropic posted a net loss of $80.748 billion in the first half, compared with a $13 billion loss in the same period last year. But the bulk of this figure did not stem from operating cash burn: about $72.4 billion came from the accounting treatment of convertible notes and SAFEs (Simple Agreements for Future Equity). As the company’s valuation soared, these early-stage financing instruments became liabilities with a higher book value, accounting for about 90% of the net loss. Operating losses were actually just $693 million, while adjusted operating profit was $1.618 billion and adjusted net profit was $2.671 billion.
The real pressure comes from compute spending. Anthropic’s capital expenditures in the first half came close to $9 billion, already exceeding the $7.3 billion for all of 2025. Even more significant are its long-term commitments: over the next seven to ten years, Anthropic must pay about $450 billion in cloud computing contract fees, most of which are non-cancellable. These include $111 billion to Google Cloud, $110 billion to Amazon Web Services, $31 billion to Microsoft, and up to $86 billion for renting Nvidia chips from Nscale and Lambda. Including more than $160 billion in commitments to rent Google chips through its deal with Broadcom, total future infrastructure costs will exceed $600 billion. The filing explicitly warns that capital expenditures are expected to “increase significantly,” and that net cash outflows from operating and investing activities may remain high even if financial performance improves.
Customer concentration is also listed separately among the risk factors: about one-fifth of the $16.3 billion in revenue in the first half of 2026 came from just two customers, and the company expects a “substantial portion” of its revenue to continue coming from a “limited number of customers” for the foreseeable future.

III A Comparison with OpenAI: Diverging Paths to Going Public
OpenAI’s IPO status provides a clear point of comparison as Anthropic files its registration statement. In terms of timing, OpenAI has clearly dropped out of the 2026 IPO window. CEO Sam Altman said in September that the company would not pursue an IPO this year, citing unfinished AI safety and alignment work. Remaining private, he said, would allow the company to make safety decisions that “may not be in the short-term commercial interest” when necessary. OpenAI’s CFO had previously mentioned 2027 as a target, but Altman himself has not committed to that timeline. OpenAI confidentially submitted a draft S-1 in June but has since chosen to put the process on hold. It was previously reported that OpenAI was close to completing an equity financing round of up to $30 billion at a pre-money valuation of about $1.25 trillion, co-led by UAE investors and a U.S. institution. The round is positioned as a pre-IPO “bridge arrangement” to fill a short-term funding gap. Foreign media have reported that OpenAI is accelerating efforts to secure investor commitments ahead of Anthropic’s expected mid-November listing, and its valuation could exceed $1.4 trillion.
In terms of valuation, the two companies’ private and expected public valuations are broadly in the same range, but they are moving in different directions. OpenAI’s latest private valuation is about $1.25 trillion, compared with $852 billion in its previous funding round in March 2026, including the funds raised.
Anthropic’s expected IPO valuation is $1.5 trillion to $2 trillion, while its private valuation in its Series H round in May was “only” $965 billion.
In other words, Anthropic is seeking a significant valuation jump through the public markets, while OpenAI has chosen to continue raising funds privately at a valuation of $1.25 trillion. Anthropic has explicitly argued that the transparency requirements of being a public company can actually strengthen AI safety, with public disclosures and external oversight serving as governance constraints. OpenAI’s position is that a private structure preserves greater flexibility for safety decisions, allowing it to address risks posed by frontier models without the pressure of quarterly results.

From a market-timing perspective, Anthropic’s IPO will directly test public-market investors’ tolerance for the valuations of foundational model companies, particularly against a backdrop of compute costs running to hundreds of billions of dollars and customer concentration that has yet to ease. OpenAI’s decision to hold back means this test will not happen to it for now—at least not in 2026.