Anthropic's valuation has shrunk, putting pressure on domestic large model companies.

From $2 trillion to $1 trillion, the global AI valuation system is undergoing a reality check.

Anthropic's IPO was originally regarded as a grand capital feast for the global artificial intelligence industry.

The developer of the Claude large model plans to list on Nasdaq, becoming one of the first top-tier AI labs in the global capital market to undergo public market scrutiny. According to previous investor expectations, its listing valuation was once expected to reach $2 trillion.

But before this feast officially kicks off, the price has already changed.

First, the $2 trillion valuation expectation was lowered to $1.5 trillion. Then, along with the adjustment of tech stocks, rising financing costs for AI infrastructure, and investors re-examining the company's financial performance, some potential investors began to discuss valuations of around $1 trillion.

The drop from $2 trillion to $1 trillion means the market's valuation expectation for Anthropic may have been cut by half.

For an AI company whose revenue is still growing at a high speed, $1 trillion is still a staggering figure, but what is really worth paying attention to is why investors are no longer willing to pay the previously high price for it.

The impact of this incident may go far beyond Anthropic itself.

Over the past two years, Chinese large model companies have also enjoyed the valuation premium brought by the global AI industry. Enterprises such as Zhipu AI, MiniMax, Moonshot AI, and DeepSeek have become the focus of the capital market.

As the world's most representative AI companies begin to accept valuation revaluation, domestic large model companies will also face a new test.

Revaluating Anthropic

Anthropic is by no means a company lacking growth stories.

On the contrary, its revenue growth rate still ranks among the top echelons of the global technology industry.

According to the IPO documents submitted by the company to the U.S. Securities and Exchange Commission, in the first quarter of 2026, Anthropic recorded revenue of 47.31 billion U.S. dollars; revenue in the second quarter reached 115.55 billion U.S. dollars; and revenue in the third quarter further increased to 169.56 billion U.S. dollars.

From the second quarter to the third quarter, revenue increased by about 47% quarter-on-quarter.

This figure is not low, but compared with the previous growth rate, it has slowed down significantly. Revenue in the second quarter increased by about 144% compared with the first quarter, while the quarter-on-quarter growth rate in the third quarter dropped to around 47%.

In other words, Anthropic is still growing rapidly, but the acceleration of growth is declining.

This is precisely one of the reasons why investors are re-evaluating its valuation.

Previously, the market was willing to give Anthropic a valuation close to 2 trillion U.S. dollars, largely because investors believed that generative AI would become a continuously expanding market, and Anthropic would have the opportunity to become one of the most important infrastructure providers in this market.

The penetration of Claude in enterprise clients, software development and agent applications has supported this expectation.

However, the pricing method of the capital market is changing.

When a company's revenue scale has reached tens of billions of U.S. dollars, investors will no longer only care about how much more it can grow, but also start to ask: How much cost does the growth require? Can new revenue cover new computing power? Can profit margins stabilize?

Anthropic's financial data has exactly exposed these issues.

In the first three quarters of 2026, the company's cumulative revenue reached 332.42 billion U.S. dollars, far higher than the 25.45 billion U.S. dollars in the same period of the previous year.

In the same period, the company's operating loss calculated in accordance with U.S. Generally Accepted Accounting Principles was 1.26 billion U.S. dollars.

If calculated based on the company's self-defined adjusted operating profit, the situation is different: the adjusted operating profit in the first three quarters reached 40.81 billion U.S. dollars.

There is a clear difference between the two sets of figures.

Taking the third quarter as an example, Anthropic recorded an adjusted operating profit of 24.63 billion U.S. dollars, but according to the GAAP standard, the company still posted an operating loss of 5.67 billion U.S. dollars.

The difference mainly comes from items such as share-based payment expenses and non-cash charitable donations. In the third quarter, the company excluded approximately 10.7 billion U.S. dollars in share-based payment expenses and 19.6 billion U.S. dollars in non-cash charitable donations from the adjusted operating profit respectively.

Adjusted profits can reflect part of the improvement in operating efficiency, but investors still need to judge how many of these adjustment items are sustainable, and how the dilution cost brought by equity incentives should be included in the valuation.

The bigger problem lies in computing power.

Anthropic disclosed in its IPO documents that as of September 30, 2026, and the subsequently signed computing power procurement and leasing agreements, the cumulative committed expenditure is about 8954 billion U.S. dollars.

Of this amount, about 4475 billion U.S. dollars is expected to be incurred by the end of 2030, and the amount to be paid in the next 12 months is about 698 billion U.S. dollars.

This is not a bill that needs to be paid immediately, but a long-term commitment distributed over many years in the future, involving computing power procurement and infrastructure leasing.

As of the end of September 2026, Anthropic has approximately 1466 billion U.S. dollars in liquidity, including cash, short-term investments and available credit lines.

The figure looks large enough, but it also illustrates one thing: the competition for cutting-edge large models is not only a competition for R&D investment, but also a competition for capital organization capabilities.

Enterprises must lock in data center, chip and power resources in advance to ensure that model training and commercial services are not affected by computing power shortages.

The problem is that infrastructure investment occurs first, and revenue realization occurs later.

If model demand continues to grow, these investments can eventually be transformed into scale advantages; if revenue growth slows down, or a new generation of models quickly changes the market competition pattern, the long-term computing power contracts signed in advance may become a heavy financial burden.

The change in Anthropic's valuation actually reflects that investors are becoming cautious about this long-term investment.

Pressure is coming for Chinese large model companies

The reason why the change in Anthropic's valuation is worthy of attention by the Chinese market is that the valuation system of domestic large model companies is also built on the judgment of future market space.

For a period of time in the past, the capital market's attention to large model companies mainly focused on three aspects:

First, whether the model capability can reach the world's leading level; second, whether Token consumption can continue to grow; third, whether the enterprise can occupy an important position in the global AI industry chain.

These indicators are of course important, but they are not directly equal to revenue, let alone profit.

In a China AI financing research report released by the U.S. think tank Rhodium Group in September 2026, a set of data worthy of attention was given.

The seven leading Chinese AI model developers counted by this institution include DeepSeek, Moonshot AI, Zhipu AI, MiniMax, Alibaba, ByteDance and Kuaishou.

The combined annual recurring revenue (ARR) of the AI model business of these companies is about 10.7 billion U.S. dollars, which is only equivalent to about 10% of the relevant revenue scale of OpenAI and Anthropic combined.

The user scale and model call volume of Chinese large model companies are growing, but in terms of commercialized revenue, there is still a clear gap with leading U.S. enterprises.

What is more noteworthy is the relationship between valuation and revenue.

According to Rhodium Group's estimates at that time, the ratio of Anthropic's valuation to ARR was about 21 times, and that of OpenAI was about 34 times.

In contrast, the ratio of Zhipu AI is about 46 times, that of Moonshot AI is about 50 times, and that of DeepSeek is as high as 163 times.

Although these figures come from the research institution's estimates of enterprise valuation and ARR, not the public market price-to-sales ratio under a unified standard, and cannot be directly regarded as fully comparable transaction multiples, they at least illustrate one problem: the valuations of some Chinese large model companies already contain quite high expectations for future growth.

This means that once the valuation anchor of global AI companies changes, it will be difficult for domestic enterprises to stay completely out of the matter.

In the past, the market was willing to pay a premium for Chinese large model companies' technological breakthroughs, domestic substitution and global competition potential.

Now, investors may raise more specific questions: How much of these companies' revenue comes from real paying customers? Can the API business form a stable scale effect? After the model call volume increases, will the gross profit margin also improve synchronously?

Judging from the disclosed financial data, these questions are not easy to answer.

Take Zhipu AI as an example.

In the first half of 2026, the company achieved revenue of 954 million yuan, a year-on-year increase of about 400%; in the same period, gross profit was about 252 million yuan, R&D expenses reached 2.131 billion yuan, and the loss during the period was 2.072 billion yuan.

Revenue is growing very fast, but R&D investment is still far exceeding the revenue scale.

The situation of MiniMax is also similar.

In the first half of 2026, the company achieved revenue of 116.6 million U.S. dollars, a year-on-year increase of 283.1%. Among them, the revenue from the open platform and other AI enterprise services reached 73.9 million U.S. dollars, accounting for 63.4% of the total revenue.

From the perspective of business structure, MiniMax is extending from AI-native consumer products to enterprise services and the developer market.

But revenue growth has not immediately translated into profitability.

In the same period, the company's gross profit margin was 17.9%, and the adjusted net loss reached 293 million U.S. dollars.

Rhodium Group's study also found that the gross profit margins of Zhipu AI and MiniMax both declined.

In 2025, the gross profit margins of Zhipu AI and MiniMax were 41% and 25% respectively; by the first half of 2026, they had dropped to around 26% and 18%.

This shows that the contradiction faced by domestic large model companies is not the lack of demand, but there is still a gap between demand growth and profitability.

The cheaper the model, the harder commercialization becomes

Another problem faced by Chinese large model companies is that their business models are not completely the same as those of leading U.S. enterprises.

Anthropic's commercialization path is relatively clear: directly sell model capabilities to customers through Claude subscriptions, API calls and enterprise-level services.

For enterprises, as long as the model can help them improve software development efficiency, reduce operating costs or replace part of manual work, enterprises have reasons to pay for these capabilities.

But the competitive environment in China's large model market is more complex.

On the one hand, domestic cloud vendors, Internet companies and independent model enterprises participate in the competition at the same time; on the other hand, the popularization of open source models has greatly lowered the threshold for enterprises to deploy large models.

This has brought about a subtle change.

The stronger the model capability, the easier it is for users to obtain low-cost intelligent services, but model developers may not necessarily obtain revenue matching the value of the model.

For example, an enterprise can use an open-weight model to deploy on its own, or obtain API services through a third-party cloud service provider. Even if model developers obtain a large number of users, they may not necessarily directly obtain revenue from all calls.

Rhodium Group specifically mentioned this issue in its report.

Under the business model of open-weight models, if a third-party cloud service provider deploys the model and provides API services to the outside world, model developers may not necessarily get a share of the revenue.

Moonshot AI and Alibaba have begun to try to establish a revenue sharing mechanism with major model users, but this still requires time to verify.

At the same time, price competition in the domestic market is also squeezing the profit margins of model companies.

Rhodium Group's study points out that the willingness of Chinese enterprise customers to pay, fierce market competition and the wide application of open-weight models are all limiting the pricing power of large model companies.

This forms a contrast with Anthropic's situation.

The problem Anthropic faces is how to maintain high growth and high gross profit margin under huge computing power investment.

Chinese large model companies need to solve two problems at the same time: how to get more users for their models, and how to make users willing to pay a sufficiently high price for the models.

If enterprises only pursue Token call volume, they can quickly expand their scale by lowering prices.

But if the revenue growth rate cannot keep up with the rate of decline in inference costs, a larger scale does not necessarily mean higher profits.

From this perspective, model call volume is no longer a sufficient indicator to measure the commercial value of an AI company.

What really matters is how much revenue and gross profit an enterprise can retain for each additional Token.

The "blood transfusion dependency" of Chinese large models

There is another important difference between Anthropic and Chinese large model companies: financing structure.

Cutting-edge U.S. AI labs can obtain funds through large tech company investments, cloud computing partnerships and public market financing.

Anthropic's investors and partners include tech giants such as Google and Amazon. These enterprises provide not only capital, but also computing power and commercial distribution channels.

The financing structure of Chinese large model companies relies more on equity financing, bank loans and industrial capital support.

Rhodium Group's study shows that in the first eight months of 2026, the four enterprises Zhipu AI, MiniMax, DeepSeek and Moonshot AI received a total of 179 billion yuan in private equity investment, IPO financing and private placement funds.

Among them, Zhipu AI raised about 4 billion yuan through a Hong Kong stock IPO in January this year, and completed an H-share private placement of about 27 billion yuan in July.

Capital is still willing to provide funds for China's AI industry.

But it should be noted that the financing capacity of the capital market is not equal to the enterprise's own hematopoietic capacity.

For large model companies that are still in the R&D investment period, equity financing can help them maintain model iteration, purchase computing power and expand their teams.

But if the valuation continues to depend on the next round of financing rather than operating cash flow, once the market risk appetite declines, financing costs may rise significantly.

This is also the most important enlightenment of Anthropic's IPO for the domestic market.

It does not mean that the valuations of Chinese large model companies must fall following Anthropic, nor does it mean that the pricing system of the U.S. capital market can be directly applied to Chinese enterprises.

Chinese large model companies have a different industrial environment.

The development of domestic computing power, the digital demand of domestic enterprises, industrial policy support, and the low-cost expansion brought by open source models may all form an independent valuation logic.

But these factors ultimately still need to be implemented at the enterprise operation level.

The capital market can pay a premium for technological breakthroughs, but it is difficult to pay for unfulfilled commercial prospects forever.

For Zhipu AI, MiniMax, Moonshot AI and DeepSeek, what really needs to be proved in the future may not only be whether the models can continue to catch up with the world's leading level, but also whether they can establish a stable business model in the fierce price competition.

This means that the indicators the market needs to observe next will also change:

Whether revenue growth comes from continuously paying enterprise customers, whether gross profit margin can improve as the scale expands, whether inference costs continue to decline, whether R&D investment can be converted into product revenue, and whether operating cash flow can gradually cover computing power investment.

These indicators do not look as dazzling as model parameters and ranking lists, but they determine whether a large model company is just a technology lab, or a technology enterprise that can continuously create profits.

Anthropic's IPO may eventually become a pricing watershed for the global AI industry.

In the past, the capital market was willing to pay high prices for the future of AI.

Now, investors are beginning to demand