Due to slowing revenue growth, investors are discussing lowering Anthropic's IPO valuation to $1.5 trillion.

On October 14, Anthropic hosted a closed-door Investor Day at its San Francisco headquarters. The attendees were primarily the company's most prominent investors, who met face-to-face with CEO Dario Amodei and CFO Krishna Rao.

The figures released by the company were impressive enough: revenue in the third quarter reached 168 billion US dollars, up 45% from the previous quarter; it turned profitable on an adjusted basis, and quarterly cash flow also turned positive.

But the market did not respond with applause.

According to Reuters, four people familiar with the matter revealed that following Anthropic's latest financial disclosures, some existing investors and potential investors set to participate in the upcoming IPO are considering lowering their valuation expectations. The valuation discussed at the Investor Day is around 1.5 trillion US dollars, while previously the company and its investors had set a valuation of around 2 trillion US dollars for the November share issuance.

From 2 trillion to 1.5 trillion, that is a 25% discount. And this is less than a month before the originally scheduled IPO roadshow to be launched in the week of November 9.

Growth is slowing, which is the core of the problem

Anthropic's revenue curve is very steep, but its slope is changing.

47 billion US dollars in the first quarter, 116 billion US dollars in the second quarter, 168 billion US dollars in the third quarter. From Q1 to Q2, the quarter-on-quarter growth reached 147%, with revenue more than doubling. From Q2 to Q3, the quarter-on-quarter growth dropped to 45%.

Growth continues, but the acceleration is declining. Multiple investors expressed disappointment over the slowdown in sequential growth in the latest quarter, as the growth rate in the previous quarter was at the "doubling" level.

Converted based on the single-quarter revenue of 168 billion US dollars in Q3, Anthropic's current annualized revenue run rate is about 670 billion US dollars. The 2 trillion US dollar valuation corresponds to a price-to-sales ratio of about 30x. If calculated based on the full-year revenue of 46 billion US dollars last year, the multiple exceeds 40x.

This valuation multiple is not cheap for an ordinary software company. And the 30x price-to-sales ratio implies a premise: growth can be sustained, and the cost curve can eventually be flattened by revenue.

The problem is that on the other end of the cost curve is an even more staggering bill. The prospectus shows that Anthropic has committed to investing at least 5180 billion US dollars in cloud computing services, computing power and infrastructure in the next ten years, averaging about 518 billion US dollars per year, close to 80% of its current annualized revenue run rate.

For the full year of 2025, Anthropic recorded revenue of 46 billion US dollars, a net loss of 420 billion US dollars, and an operating loss of more than 80 billion US dollars. Computing and infrastructure expenditures reached 73.3 billion US dollars, accounting for more than half of its total operating expenditures of 126.5 billion US dollars.

Financing risks in AI infrastructure are amplifying

Anthropic's valuation cut is not an isolated event. It is taking place amid a global tech stock sell-off, whose core focus is AI infrastructure financing.

On the same day, CoreWeave, the GPU cloud computing company backed by NVIDIA, saw its share price plummet by 15.46%. The yield on its unsecured bonds is approaching 15%, and the credit market predicts its default probability within five years is as high as 63%. CoreWeave's debt scale increased from about 150 billion US dollars in the previous year to 520 billion US dollars, and its interest expenditure in the second quarter rose to 640 million US dollars from 267 million US dollars in the same period last year.

Panic spread rapidly. Nebius fell 10.08%, and Oracle fell 8.83%. The KOSPI Composite Index plummeted by 7.6%, with the share prices of Samsung Electronics and SK Hynix both falling by more than 9%, and the exchange launched the SIDECAR mechanism to suspend programmatic selling.

Investors' concerns are focused on one point: more and more funds for AI infrastructure construction are relying on borrowing. CoreWeave is a typical case, whose hardware procurement is highly dependent on debt. The surge in bond yields means the market demands higher risk premiums to be willing to provide funds. Elon Musk replied "Yikes" to CoreWeave's bond yield trend chart on social media, further amplifying market sentiment.

This concern is spreading from the infrastructure layer to the model layer. If the financing cost of computing power leasing companies rises, the cost for model companies to obtain computing power will also rise accordingly. And the 5180 billion US dollar infrastructure investment committed by Anthropic is premised on its ability to obtain continuous financial support at a reasonable cost.

Valuation benchmarks for Chinese companies are also facing similar pressures

Anthropic's valuation adjustment also has reference significance for Chinese large model companies that are sprinting for IPOs.

In January this year, Zhipu AI and MiniMax were listed on the Hong Kong Stock Exchange one after another. After listing, their performance in the capital market was strong, and their combined market capitalization once reached the level of hundreds of billions of US dollars. The market paid for the story of "China's version of Anthropic", and Zhipu AI's price-to-sales ratio once exceeded 500x, while MiniMax's exceeded 600x.

But there is a magnitude gap in revenue volume. According to estimates by Rhodium Group, the total annual recurring revenue of seven leading AI developers in China is about 10.7 billion US dollars, only equivalent to about 10% of the sum of OpenAI and Anthropic. Among them, OpenAI's ARR is about 400 billion US dollars, and Anthropic's is about 650 billion US dollars.

The contrast in valuation multiples is even more obvious. The valuation/ARR multiples of OpenAI and Anthropic are about 34x and 21x respectively, while those of Moonshot AI and DeepSeek reach 50x and 163x respectively.

This means that the valuations of Chinese large model companies are more based on the "scarcity premium" and "future growth expectations" rather than the current revenue scale. When a company with an annualized revenue of 670 billion US dollars like Anthropic needs to lower its valuation expectations, Chinese companies whose revenue volume is an order of magnitude smaller will face even greater pressure on their valuation logic.

However, the financing paths of Chinese AI companies are significantly different from their US counterparts. Anthropic's computing power investment mainly relies on corporate cash flow and equity financing, while the AI infrastructure investment of leading Chinese tech companies is increasingly relying on debt financing. Tencent has issued about 38 billion RMB in notes for financing this year, with capital expenditure reaching 96.3 billion RMB in the first half of the year, up about 110% year on year. Alibaba plans to invest 380 billion RMB in cloud and AI infrastructure construction in the next three years.

According to calculations by Huatai Securities, the total capital expenditure of 8 leading domestic internet giants in 2026 is expected to increase to about 1.07 trillion RMB, a substantial increase from 455 billion RMB in 2025. The sources of funds for these investments are shifting from own cash flow to a combination of debt and equity financing.

A question that needs to be answered

Back to Anthropic, at the Investor Day, the company told potential investors that it achieved adjusted profitability in the third quarter. But the word "adjusted" is the key — it did not say it was profitable under the GAAP standard, nor did it disclose which specific items were adjusted out. Combined with the background that the operating loss last year exceeded 80 billion US dollars, what the market needs to judge is: how solid is the conversion rate of an AI lab turning revenue into profit.

Anthropic's IPO is expected to precede OpenAI's. OpenAI confidentially submitted its IPO application in June, but its listing time has been delayed again and again, and it is currently expected to be in early 2027. Anthropic's issuance result will become an important case for the global capital market to test the business model of cutting-edge AI enterprises. If the valuation drops from 2 trillion to 1.5 trillion, it means that the public market's pricing logic for AI companies is shifting from "technical narrative" to "revenue quality, profitability sustainability and capital use efficiency".

For Chinese large model companies that are queuing up for IPOs, this signal also needs to be taken seriously.

This article is from the WeChat official account "Tech Business", written by Tech Business, and authorized for release by 36Kr.