“Cracks in Expectations for AI Growth”: Anthropic’s Valuation Slides Ahead of IPO as Long-Term Investment Burdens and Regulatory Headwinds Reignite Bubble Concerns

Anthropic’s valuation plunges as earnings and investment risks mount ahead of listing
Tighter US data center regulations and a technology stock sell-off also weigh on investor sentiment
Practical applications and revenue generation by companies adopting AI will be crucial as growth potential faces a test

The valuation of Anthropic, an artificial intelligence (AI) company pursuing a Nasdaq listing, is declining. With revenue growth slowing and concerns mounting over the impact of massive long-term infrastructure investments on profitability, investors are repeatedly lowering their valuation expectations. External factors—including tighter US data center regulations, weakness in technology stocks, and demand for and practical applications of AI models in industry—are also cited as reasons for growing market caution.

Anthropic’s Declining Valuation

According to a Reuters report on the 10th (local time), Anthropic proposed an IPO price range of USD 70–78 per share in an amended securities registration statement filed with the US Securities and Exchange Commission (SEC). This implies a fully diluted valuation of USD 1.26 trillion–1.41 trillion, with the upper end roughly 30% below the USD 2 trillion discussed in the market early last month. The initial public offering (IPO) is expected to raise up to USD 64.4 billion.

Anthropic adjusted its proposed offering price because investor expectations have been falling rapidly since last month. On the 15th of last month, Reuters reported that some institutional investors who attended Anthropic’s investor presentation considered a valuation of around USD 1.5 trillion appropriate. Bloomberg subsequently reported on the 6th that potential Anthropic investors were discussing a lower valuation of USD 1.2 trillion–1.3 trillion. Some investors reportedly lowered their valuations to USD 1.1 trillion.

Growing Doubts About Business Growth

Weakening earnings growth underlies the pronounced decline in valuation. At an investor presentation last month, Anthropic announced that its third-quarter revenue this year rose 45% from the previous quarter to USD 16.8 billion. Given that second-quarter revenue more than doubled from the preceding quarter, this represents a substantial slowdown. Anthropic also has nearly USD 895.4 billion in purchase and lease commitments related to AI computing infrastructure. Of this, USD 69.8 billion is scheduled for payment before October 2027, while spending expected through the end of 2030 totals USD 447.5 billion. Most of the remainder is scheduled for 2031–2045. Under this structure, the financial burden of long-term contracts could increase considerably if future revenue growth falls short of expectations or partners face deteriorating financing conditions.

There are also substantial doubts about long-term growth potential. Goldman Sachs, the lead underwriter for Anthropic’s IPO, told potential investors that the company’s annual revenue would surge from USD 4.59 billion last year to USD 52 billion this year, USD 136 billion in 2027, and USD 685 billion in 2030. This scenario assumes revenue will increase roughly 150-fold in five years. Despite this revenue growth, however, Goldman Sachs projected that Anthropic’s operating loss would widen to USD 40.4 billion next year and that free cash flow (FCF) would register a net outflow of USD 65.7 billion in 2027. Operating income is not expected to turn positive until 2029, and free cash flow until 2030.

Structural Risks Come Into Focus

The business structure itself also has vulnerabilities. According to Anthropic’s securities registration statement, its five largest customers accounted for roughly 25% of total revenue from January through September this year. Many major customers either have no long-term contracts or have contracts nearing expiration, leaving them able to reduce spending or discontinue services at short notice. Revenue generated through online marketplaces operated by cloud providers such as Amazon Web Services (AWS) also accounted for nearly 44% of the total during the same period. This means that reduced AI investment by key customers or changes to agreements with cloud providers could directly affect results. Anthropic also estimated that roughly 5–7% of revenue during the period came from customers violating its usage policies or commercial contract terms, and said it was difficult to predict the scale and timing of any resulting revenue decline.

Corporate governance is also cited as a major post-IPO risk. Anthropic designed a governance structure that gives Class F shares held by an entity affiliated with its founders 50.1% of total voting rights. This would make it difficult for public investors to exercise meaningful influence over board composition or major management decisions after the listing. In addition, Anthropic stated in its registration statement that, “As a public benefit corporation, we may make decisions that conflict with shareholders’ short-term interests in pursuit of AI safety and the long-term public benefit, and such management decisions could lead not only to weaker short-term results but also to a decline in the share price after listing.” This means that even highly profitable projects could face delays in development or commercialization if they conflict with public-interest objectives.

Table 1. Factors Pressuring Anthropic’s Valuation

External Headwinds Accumulate

The external market environment is also deteriorating by the day. Changes in the US political landscape, in particular, have emerged as a new source of uncertainty for the AI industry. With Democrats securing majorities in both the Senate and the House in the midterm elections on the 3rd, tighter regulation of AI infrastructure, including data centers, has become more likely. Democrat Gina Hinojosa, who won the Texas gubernatorial election, pledged to halt new data center construction, while a Democratic candidate advocating a moratorium on related development projects also won in Georgia. Goldman Sachs estimated that the Democratic election victory could temporarily affect roughly 45% of planned US data center expansion capacity.

This political uncertainty weakened investor sentiment across technology stocks. On the 5th, the US Nasdaq Composite fell 5.56% from the previous trading session, its largest decline in roughly 19 months. The Nasdaq 100 also slid 6.4%, wiping out approximately USD 2.5 trillion in market capitalization in a single day, while the Philadelphia Semiconductor Index dropped 12%. Risks involving the leadership of Nvidia, the dominant player in AI chips, also emerged as a new headwind. The US Federal Bureau of Investigation (FBI) and Nvidia’s board launched investigations after allegations surfaced that a person with whom Nvidia CEO Jensen Huang was having an affair had ties to Chinese intelligence agencies. As the controversy spread, investor concerns over national security risks quickly intensified, and calls grew in the US Congress to accelerate legislation restricting China’s access to advanced AI chips.

Industry Outlook Depends on Demand

Another risk is that resolving the AI industry’s bubble concerns depends more on companies adopting the technology in their actual operations than on those supplying it. Major AI developers such as OpenAI, Anthropic, and Google are rapidly improving the capabilities of large language models (LLMs). But unless corporate users translate these advances into productivity gains, cost savings, and new revenue, market demand sufficient to justify massive infrastructure investment will be difficult to sustain over the long term. Companies therefore need to go beyond simply purchasing access to AI services, redesign workflows across their organizations, and deliver tangible economic results.

Global companies that led AI adoption are already showing signs of caution toward indiscriminate use. For example, on the 27th of last month, Amazon revised its AI usage guidelines for software developers and moved to curb so-called “tokenmaxxing,” the practice of increasing AI usage for its own sake. Amazon Vice President Turk Turkowitz warned, “There are problems with using AI model token consumption or the volume of code written as measures of productivity,” adding, “Technology use unrelated to actual work outcomes can increase costs and needlessly complicate tasks.” The AI industry has thus fully entered a transitional phase in which it must prove its tangible economic value, moving beyond technology development and the race for early adoption.