Payment will be made within 48 hours. The valuation is 6 trillion.

The supply of projects in popular sectors falls short of demand, and the valuations of unicorns are skyrocketing. This phenomenon is not limited to the current domestic primary market. When a startup asks investors to complete their payment intentions within 48 hours, it is a sign of the times, indicating that in the spring of 2026, the financing narrative in the AI sector is entering a new stage.

Multiple foreign media reports indicate that Anthropic is advancing a new round of financing worth up to $50 billion, with its valuation aiming for $900 billion (approximately 6 trillion RMB). The soaring valuation figure within a month is quite interesting. First, Bloomberg and Business Insider reported that Anthropic had received multiple pre - financing offers with a valuation of $800 billion. However, TechCrunch, citing six informed sources, gave a valuation range of $850 billion to $900 billion and claimed that Anthropic required investors to submit placement plans within 48 hours.

Some sources also said that due to the surging demand from investors, the final valuation is likely to exceed this figure. This also means that Anthropic's potential valuation has exceeded that of OpenAI - its top competitor, which just completed a record - breaking $122 billion financing in late March, reaching a valuation of $852 billion.

Of course, Anthropic's deal has not been finalized yet. In response, the company's spokesperson declined to comment. But that doesn't really matter. You know, in the primary market, this kind of lightning - fast subscription almost only occurs for the top - tier assets. Since the release of ChatGPT at the end of 2022, Anthropic has been chasing OpenAI, and this chasing momentum has been particularly strong in the past year. Now, the continuously growing revenue and valuation figures at least show that it has finally caught up.

Annualized ARR exceeds $40 billion, aiming for listing

Going back to May a year ago, Anthropic officially launched Claude Opus 4 and Claude Sonnet 4 at its first developer conference, kicking off nearly a year of technological upgrades. Among them, Opus 4, as the flagship model, supports continuous independent work for 7 hours and scored 72.5% in the SWE - bench programming benchmark test, surpassing GPT - 4.5 of the same period and establishing its advantage in handling complex enterprise - level tasks. Subsequently, Claude Opus 4.5 was iteratively launched, breaking through the 200,000 - token ultra - long context window and being able to fully process entire books and millions of lines of code. The implementation of technology has also accelerated simultaneously. The Claude model has become the only large model that covers the three major cloud platforms of AWS, Google Cloud, and Azure.

During this stage, Anthropic clearly played two aces in its technology roadmap: security and efficiency, and refused to relax AI ethical restrictions for short - term benefits. During this period, Anthropic even publicly rejected a $200 million contract from the US Department of Defense, maintaining the red line of "prohibiting use in lethal weapons and large - scale surveillance", which consolidated the trust of high - compliance customers.

Different from OpenAI, the core driving force behind Anthropic's growth is the explosion of enterprise - level demand and the improvement of the product matrix. On the one hand, 8 of the Fortune 500 companies have become in - depth customers of Claude, and the customized orders from industries such as finance, healthcare, and law have increased significantly. On the other hand, intelligent agent products such as Claude Code (programming assistant) and Claude Cowork (office assistant) have been launched, forming a complete matrix of "basic model + vertical application", and the value per customer has increased significantly.

Benefiting from Claude Code, just last month, Anthropic announced that its annualized revenue (ARR) had reached $30 billion, which has exceeded that of OpenAI. By May 2026, this figure had become $44 billion. In addition, SemiAnalysis pointed out that Anthropic's gross profit margin for inference business increased from 38% in May 2025 to over 70% in May 2026, with cost - efficiency far exceeding the industry average.

For Anthropic, which is making rapid progress in business, the most urgent thing at the moment is computing power. At the beginning of this month, Anthropic launched its latest model, Claude Opus 4.7, and claimed it to be the best product currently available for the mass market. Earlier, Anthropic launched Claude Mythos Preview, a model with advanced network security features that is only available to some specific companies. Mythos has triggered a series of high - profile meetings among government officials, technology companies, and bank executives, which is also one of the reasons for Anthropic to seek new financing - Anthropic needs to purchase a huge amount of computing resources to run the Mythos model.

In recent weeks, computing power has even become the most concerned issue for Anthropic. Anthropic has reached an agreement with Amazon. Amazon has agreed to invest $5 billion in Anthropic immediately, and may add up to $20 billion in the future if specific business milestones are met, with a total upper limit of $25 billion. In return, Anthropic will get up to 5 gigawatts of computing power from Amazon Trainium chips for training and deploying the Claude model, and nearly 1 gigawatt is expected to be online by the end of 2026.

Anthropic has also announced with Google and Broadcom that it will obtain approximately 3.5 gigawatts of next - generation TPU computing power, which is expected to be online in 2027. Google then made an additional investment, promising to provide another 5 gigawatts of TPU computing power in the next five years. Google said it plans to invest up to $40 billion in Anthropic, with an immediate investment of $10 billion and an additional $30 billion if performance targets are met in the future.

So, like OpenAI, Anthropic is also eager to go public, both to give an explanation to the capital market and to have more funds to purchase computing power.

Techcrunch further reported that despite strong market demand, some early investors, especially those who invested in 2024 or earlier, have chosen to abandon this round of financing. These investors are waiting to cash out in the IPO expected to take place later this year. Earlier, the Financial Times also revealed that Anthropic has selected a law firm to start preparations for an IPO and may go public as early as this year.

The FOMO game of capital

The ongoing round of financing is also regarded as the last round of financing before listing.

In September last year, Anthropic conducted a Series F financing, raising $13 billion, with a post - investment valuation of $183 billion. It was led by ICONIQ, Fidelity, and Lightspeed, and followed by top - tier capital such as Blackstone, Goldman Sachs, and the Qatar Investment Authority. After this round of financing, Anthropic officially joined the club of AI companies with a valuation of over $100 billion, forming a duopoly prototype with OpenAI.

In February this year, in the Series G round, Anthropic raised $30 billion, with a post - investment valuation of $380 billion. It was led by the Singaporean sovereign wealth fund GIC and the hedge fund Coatue, and Google and Amazon increased their strategic investments. This set the second - largest private financing record globally, second only to OpenAI's $40 billion financing.

In the latest round aiming for a $900 billion valuation, the most iconic change in this process is not the amount but the extreme compression of the transaction rhythm. The traditional financing process in the primary market is: due diligence, negotiation, determination of terms, submission to the investment committee, and payment, which takes weeks or even months. However, this round of financing by Anthropic presents a completely different logic: the subscription window opens, investors scramble for quotas, and the shares are locked in within 48 hours.

Behind this, Anthropic's growth logic has been upgraded from tool - type products to enterprise - level infrastructure: it is no longer just a "promising artificial intelligence startup" but has become the fastest - growing B2B company in the history of the software industry. 8 out of the top 10 Fortune companies have become its paying customers, and the number of enterprise customers with an annual consumption of over $1 million has exploded from a single - digit number in the early stage to hundreds or even thousands. This means that Claude is no longer a pilot project for enterprises but has gradually evolved into a necessary tool deeply embedded in the core business process.

Once the AI model becomes the "water, electricity, and coal" of the digital age, the financing logic will change fundamentally. Capital is no longer investing in future growth but seizing the core entry point. Missing out on Anthropic may mean missing out on the core track of enterprise AI services in the next decade. This kind of anxiety makes top - tier capital willing to give up the bargaining power and compress the decision - making cycle.

Of course, both Anthropic and OpenAI face unavoidable structural pressures: the cost of model training is increasing exponentially, and a single ultra - large - scale training can cost billions of dollars; the inference cost remains high, and the computing power consumption brought by the large - scale invocation of enterprise customers requires continuous huge capital support. More importantly, the global enterprise AI demand is growing explosively, and the computing power reserve directly determines the delivery ability and market share.

For leading AI companies, financing means more abundant survival resources. Delaying financing may cause a company to fall behind in the computing power arms race, and then be overtaken by competitors in terms of customer acquisition and market gap. This life - and - death pressure forces Anthropic to complete financing as quickly as possible, and capital is also willing to cooperate in speeding up. After all, the later the investment, the less the quota.

Anthropic was not a favorite of capital in the early stage and was rejected by 21 top - tier VCs. Now, its valuation has soared dozens of times, and it has become a "hot commodity" pursued by global capital. This dramatic reversal has completely changed the behavior pattern of investors. It is imaginable that in order to get a share of Anthropic, many top - tier investment institutions have abandoned the traditional due - diligence process, simplified the investment committee approval, and even made payments first and completed the procedures later, just to lock in the shares in time.

For them, the cost of missing out on Anthropic is far higher than the risk of high valuation. After all, so far, Anthropic is the only target that can compete with OpenAI, and its scarcity is irreplaceable.

This article is from the WeChat official account "China Venture Capital" . Author: Liu Yanqiu. Republished by 36Kr with permission.