According to Woofun AI, Anthropic is sprinting into an initial public offering (IPO) with a valuation of $2 trillion, and its S-1 draft submitted to the US Securities and Exchange Commission was exclusively revealed by the Wall Street Journal on the 20th. The document revealed an extreme contrast in the company's financial performance: although revenue surged to $16.3 billion in the first half of the year, net loss reached $80.7 billion during the same period, and huge infrastructure expenses and potential compliance risks became the focus of market attention.
The explosive increase in revenue cannot hide the sharp expansion in the cost structure. Anthropic's revenue for the first half of the year reached $16.3 billion, an impressive increase of about 14.8 times compared to $1.1 billion in the same period last year.
However, computing and infrastructure-related costs underpinning this growth also soared from $2.3 billion in the same period last year to around $9 billion. Notably, the computational cost for the first half of the year alone surpassed $7.3 billion for the full year of last year, showing an exponential rise in computing power consumption.
Although the revenue growth rate nominally exceeded the cost increase, due to huge upfront investment, the net loss remained high, reaching US$80.7 billion, far higher than US$13 billion in the same period last year. According to data compiled by Woofun AI, of this net loss of $80.7 billion, $72 billion was not an operating loss, but accounting expenses due to changes in the fair value of convertible bonds and other financial products issued during the financing process.
This detail shows that the company's book losses are largely affected by fluctuations in the valuation of financial instruments, rather than simply the failure of business operations.
An even more serious challenge comes from the huge cloud computing infrastructure contracts that will need to be borne over the next 7 to 10 years. According to documents, Anthropic has signed a total of about 450 billion US dollars of cloud computing contracts, of which Google (GOOGL.US) accounts for 111 billion US dollars, followed by Amazon Web Services (AWS) (AMZN.US) with 110 billion US dollars, and Microsoft (MSFT.US) contributed 31 billion US dollars. Most of these contracts are irrevocable, meaning that Anthropic must meet payment obligations regardless of changes in market demand.
Additionally, contracts for leasing Nvidia (NVDA.US) chips through EnScale and Lambda could be as much as $86 billion. Coupled with Google (GOOGL.US) chip leasing obligations with Broadcom (AVGO.US), the amount exceeds $160 billion. Adding up all of these known contracts, the future infrastructure contract burden would exceed $600 billion.
If other contracts currently under negotiation are taken into account, the total amount could even rise to $750 billion. This long-term, highly rigid capital expenditure structure greatly limits the company's future financial flexibility. Any fluctuation in market demand may lead to serious resource idle or capital chain pressure.
Excessive customer concentration and the potential legal liability of artificial intelligence models pose another risk. According to the document, about 20% of Anthropic's revenue in the first half of the year came from two customers, and it is expected that it will still be highly dependent on a few large customers in the future.
This customer structure makes the company extremely vulnerable to changes in single customer decisions. Once major customers cut budgets or switch to competitors, revenue will risk a cliff-style decline. At the same time, legal liability for bad behavior that may be caused by artificial intelligence models is also viewed as a major hidden danger. In the document, Anthropic clearly mentioned the potential for threatening behavior, cyber attacks, and data breaches that an uncontrolled model could bring. The company warned that once a major artificial intelligence incident occurs, it may be difficult to cover the corresponding losses with the existing insurance mechanism alone.
This unpredictable technical risk can not only result in huge compensation, but also damage brand reputation, which in turn affects long-term commercial cooperation.
External expansion hindrances have further exacerbated the uncertainty in the IPO process. Local communities are increasingly opposed to data center construction, which may lead to project delays or cost overruns, directly affecting the supply of computing power. At the same time, cloud service provider limitations on computing power have also become a key factor limiting business expansion.
Despite rapid revenue growth, how to obtain stable access to computing resources and how to manage the burden of long-term contracts have become core issues that Anthropic must solve. In a context where computing power has become a scarce resource, Anthropic faces fierce competition with other tech giants for limited computing power, and any interruption in supply could have a fatal impact on the stability of its service.
Furthermore, the burden of long-term contracts makes the company lack sufficient room for adjustment in the face of market changes, further amplifying operating risks.