Anthropic raised 11.5 billion U.S. dollars in Q2, recording an astounding 14-fold growth rate that leaves Wall Street's valuation system at a loss for words.

Wall Street no longer knows how to value AI companies.

Anthropic's growth rate has once again exceeded its own forecasts.

According to investor documents seen by Bloomberg, Anthropic's preliminary revenue in the second quarter of this year exceeded 115 billion US dollars, surging more than 14 times from 787 million US dollars in the same period last year; it also more than doubled compared with 47.3 billion US dollars in the first quarter of this year.

Anthropic also recorded positive adjusted operating profit in the second quarter. The relevant figures are still preliminary data, and the final results may be adjusted.

Just the day before, Reuters exclusively disclosed an even more staggering figure: Anthropic expects that by 2028, the company's annual revenue will reach 190 billion to 200 billion US dollars. The news came from two people familiar with the company's financial situation, and this figure had not been previously reported.

As Anthropic speeds up toward its IPO, Wall Street has begun to use this "Anthropic two years from now" to determine how much Anthropic is worth today. This also turns a problem that originally sounded almost crazy into a realistic one.

Two years later, can a company that mainly makes money by selling AI model capabilities really achieve 200 billion US dollars in annual revenue?

The newly disclosed quarterly revenue of 115 billion US dollars suddenly makes the seemingly crazy "200 billion US dollar story" much less distant. But what Wall Street is betting on is whether AI can become a business with sufficiently lucrative profits.

Growth is so fast that traditional valuation methods can no longer keep up

Anthropic's growth curve is already quite rare.

In the second quarter of last year, the company's revenue was only 787 million US dollars; it rose to 47.3 billion US dollars in the first quarter of this year, and exceeded 115 billion US dollars in the second quarter.

If we simply annualize the latest quarterly revenue, Anthropic is now equivalent to a company with an annual revenue of about 460 billion US dollars.

This is basically consistent with the data Anthropic announced in May this year. At that time, when the company completed a new round of financing, it stated that its annualized run-rate revenue had exceeded 470 billion US dollars. This round of financing reached 650 billion US dollars, with a post-money valuation of 9650 billion US dollars.

This scale was almost built in just over a year. At the beginning of 2025, Anthropic's annualized run-rate revenue was only about 10 billion US dollars; by August 2025, it had exceeded 50 billion US dollars. The company said its revenue run rate has achieved more than 10 times annual growth for three consecutive years.

The acceleration this year is even more obvious. The financial forecasts previously obtained by Reuters show that Anthropic originally expected its second-quarter revenue this year to reach at least 109 billion US dollars, and record a quarterly operating profit of about 5.59 billion US dollars. The preliminary revenue disclosed now has exceeded 115 billion US dollars, which is higher than the previous forecast.

This is also why the extremely exaggerated "200 billion US dollar revenue" target for 2028 has begun to enter Wall Street's valuation models.

Based on the current annualized revenue level of about 460 billion to 470 billion US dollars, Anthropic needs to expand by about 4 times more to reach 200 billion US dollars.

For a mature software company, this is still an extremely aggressive forecast. But when placed against Anthropic's growth curve over the past few years, the situation becomes more complicated.

Wall Street starts pricing Anthropic for two years from now

Reuters, citing four people familiar with the matter, said that in the pre-listing valuation discussions for Anthropic, banks and investors are using the EV-to-Revenue (enterprise value-to-revenue) multiple, and directly adopting the 2028 revenue forecast.

For high-growth software companies that have not yet achieved mature profits, it is not uncommon to use revenue instead of profits for valuation. But looking directly two years ahead is not common practice.

Why? Anthropic is growing too fast. Investors are therefore facing a rather special problem.

If Anthropic is valued based on today's profits, its huge investments in GPUs, model training, inference, talent and data centers will make it look extremely expensive.

Assuming that these infrastructures can eventually support an AI platform that is several times larger and has continuously improving profit margins, today's investments may just be the costs during the rapid expansion period.

Wall Street is actually betting on the second scenario. Reuters said that high-growth companies such as Palantir, Cloudflare and SpaceX are currently being used by investors as valuation references for Anthropic. There are even discussions in the market about whether Anthropic can reach a valuation of 2 trillion US dollars.

Palantir: It has become the benchmark for investors to value companies with high growth and deep involvement in AI business.

Cloudflare: It provides a direct comparison with high-growth software and infrastructure companies.

SpaceX: It provides a case showing that a company's valuation can partly depend on expectations for its future scale, rather than current financial fundamentals.

Similar precedents have appeared among some of the fastest-growing companies listed recently.

People familiar with the matter said that Cerebras's backers cited 2028 revenue expectations before the company's IPO this year. Before SpaceX went public at a record valuation in June this year, its financial forecasts even extended to 2029.

This also reflects the difficulty of valuing an AI company.

Who on earth will pay for the 200 billion US dollars?

After the Reuters report was released, the related discussion quickly climbed to the top of the Reddit tech community.

Supporters believe that judging from Anthropic's growth rate over the past three years, 200 billion US dollars is not as incredible as it seems at first glance. Especially when enterprise AI is still in its early stage, a large number of companies have not yet actually deployed generative AI into their core workflows, and the potential market is still huge.

But skeptics have raised the question: Is the same growth logic that drove revenue from tens of billions of dollars to hundreds of billions of dollars still applicable?

Some netizens pointed out that it is one thing to get the market to pay tens of billions of dollars a year for Claude; it is completely another thing to get enterprises to finally pay nearly 200 billion US dollars. Enterprises need to prove that AI can continuously generate sufficiently large return on investment to support this scale of long-term expenditure.

Others have raised the question: If Anthropic, OpenAI and other AI giants all expect to reach hundreds of billions of dollars in revenue in the future, which part of the economic system will the huge expected revenue of the entire AI industry eventually come from?

No matter where it comes from, 200 billion US dollars in revenue must eventually correspond to 200 billion US dollars in real customer spending.

Anthropic's biggest risk: AI is getting cheaper

Some users are turning their attention to open-weight models, local models, and increasingly affordable competing products.

What enterprises need may not always be the most powerful model in the world, but a model that is good enough to complete tasks and costs less. Once a large amount of AI capabilities become commoditized, enterprises will have the incentive to continuously look for lower-cost alternatives, and the pricing power of model vendors will also come under pressure.

This is the most easily overlooked layer of risk in Anthropic's "200 billion US dollar forecast".

Revenue essentially depends on two variables: usage volume, and price. The market now generally believes that the volume of AI calls will increase sharply. But if the price per unit of model capability drops just as fast in the future, the explosion of usage volume does not necessarily mean that revenue will explode at the same speed. There may even be a rather counterintuitive result: the faster AI technology advances, the faster today's most expensive cutting-edge capabilities will turn into cheap commodities tomorrow.

The moat that Anthropic really needs to build is not just leading the model rankings, but embedding Claude into enterprise workflows, making it increasingly difficult for customers to replace it.

The current growth of Claude in programming and professional work scenarios is exactly one of Anthropic's most important bargaining chips. Reuters previously pointed out that the rapid adoption by enterprise customers and Claude Code is an important driving force for the company's revenue surge in recent years.

Revenue is not the hardest part, profit is

In the past, the market often asked when AI companies would become profitable? Now, the question has further become: Can this profit be sustained, and can it grow rapidly as the scale expands? This is crucial to Anthropic's valuation.

One of the biggest differences between AI and traditional SaaS is that every model call requires real computing resources. When a traditional software company adds its 1 millionth user, the marginal server cost may be very low; but every time a large language model generates a token, it corresponds to GPU, power and data center resources.

Therefore, to support a valuation close to a trillion US dollars or even higher, Anthropic must also prove that revenue can eventually grow faster than computing costs.

This is exactly the core assumption of investors in the Reuters report: as the efficiency of models, chips and inference improves, and fixed costs such as personnel are diluted by a larger revenue scale, Anthropic's profit margins will continue to expand in the future.

If this assumption holds, then the huge computing power invested today may be similar to Amazon's logistics infrastructure in the early years, which was built in advance for future scale. If it does not hold, an AI company whose revenue rises while its computing power bill also rises synchronously will hardly enjoy the high valuation multiples of traditional software companies.

At least one thing is clear now: The market has underestimated Anthropic's growth rate more than once. But the truly difficult part of the "200 billion US dollar story" has only just begun.

It requires enterprise AI spending to continue growing at a high speed; it requires Claude to maintain sufficiently strong product competitiveness; it requires open models and low-price competitors not to push down prices quickly; and it also requires computing power costs to drop fast enough so that revenue can eventually be converted into large-scale profits.

Quarterly revenue of 15 billion US dollars has already proved that Anthropic's ability to sell AI is far stronger than most people expected. But to support the trillion-dollar valuation that Wall Street is discussing, its next step must prove a more ancient business question: How much of this money can finally be retained?

This article is from the WeChat official account Synced (ID: almosthuman2014), Author: Sia, published with authorization from 36Kr.