AI News

Anthropic launches IPO roadshow, with valuation cap of about $1.02 trillion

YicaiUpdated Nov 11 16:53

AI summary

Anthropic launched its IPO roadshow on November 10, planning to issue 825 million shares at a price range of $70 to $78 per share, raising up to about $64.4 billion, with a valuation cap of about $1.02 trillion based on outstanding shares, lower than the previous market expectation of $2 trillion. The company also plans to raise about $26 billion through private placements to cornerstone investors such as Google, Nvidia, and AMD. The prospectus shows that its revenue in the first three quarters of 2026 was $33.24 billion, a year-on-year increase of 1206%, while it has signed approximately $900 billion in computing power and infrastructure purchase and lease commitments, of which about $451.6 billion is expected to occur before the end of 2030.

Why read this

SelectedAI score 85

The original article places Anthropic's IPO pricing, its $900 billion computing power commitment, and the market's shift in pricing logic from revenue growth to return on capital together, so readers can understand changes in the AI financing environment based on this.

Topics

Tags

Body · AI translation

Anthropic is bringing the AI industry's most pivotal capital race of the past few years into the public markets.

On November 10 local time, Anthropic launched its IPO roadshow, planning to issue 825 million shares at a price range of $70 to $78 per share, raising up to approximately $64.4 billion.

At the same time, the company also plans to raise approximately $26 billion through private placements to cornerstone investors including Google, Nvidia, and AMD.

Based on the roadshow documents and outstanding shares, Anthropic's corresponding IPO valuation ceiling is approximately $1.02 trillion; on a fully diluted basis, the valuation ceiling is approximately $1.4 trillion. This is a marked decline from the $2 trillion the market had once expected this year.

Beyond Anthropic, OpenAI has also cut its new funding target from roughly $30 billion to about $20 billion, with its valuation dropping from about $1.4 trillion to $1.25 trillion.

A Slowing Growth Curve

Looking solely at operating data, Anthropic has not experienced a traditional fundamental collapse.

The company's latest prospectus shows revenue of $33.24 billion in the first three quarters of 2026, up 1,206% year over year; third-quarter revenue was approximately $17 billion, still up about 47% quarter over quarter. As of the end of October, its Run-Rate Revenue was approximately $81 billion. More than 11,900 customers contributed at least $100,000 in revenue over the past 12 months.

The problem is that growth is decelerating from an extremely high level, while capital investment has not slowed in tandem.

Although revenue grew 12-fold year over year in the first three quarters, the company itself noted in its prospectus that as the revenue base expands and the AI market matures, past growth rates are not sustainable. For investors, valuation is therefore no longer just about how big the future can be, but is beginning to shift toward "how much revenue and profit can each dollar of compute investment actually generate."

Xie Siyuan, managing director at Yijing Capital, told Yicai that for companies like Anthropic, one cannot look only at the absolute scale of ARR; it is also necessary to observe revenue generated per unit of compute, the economics per unit of Token, and whether profit margins can continue to improve as scale expands.

This change also explains why Anthropic and OpenAI are simultaneously facing tighter financing conditions: when the market shifts from chasing revenue growth to calculating return on capital, a marginal decline in revenue growth will be amplified by the capital markets.

But this does not mean AI demand has suddenly disappeared. Xie Siyuan judges that inference demand, Agent applications, and enterprise-level applications are still expanding, and compute remains an important foundational resource for the AI industry. He prefers to understand the recent adjustment as repricing under supply and capital constraints, rather than a complete reversal of the demand logic.

However, the capital markets are beginning to demand a higher cost of proof.

Anthropic's prospectus shows that as of September 30 and thereafter, the company's commitments for compute and infrastructure procurement, leasing, and other items total approximately $900 billion, of which about $451.6 billion is expected to occur before the end of 2030, and about $70.7 billion needs to be paid in the next 12 months. The company explicitly states that a considerable portion of these agreements include minimum purchase volumes, take-or-pay, and other clauses, and that even if actual utilization is insufficient, it may still be obligated to pay.

Xie Siyuan is not pessimistic about this. He believes that AI may become a foundational resource similar to oil and water in the future, and the real risk is not massive investment, but whether competition can ultimately form a sufficiently strong advantage. Model companies continuously purchasing compute and expanding training scale are essentially building capital barriers.

Anthropic's $900 billion is also not cash or revenue cost that must necessarily be consumed in the future. The latest documents show that about $213.8 billion of it involves exercisable termination rights or preconditions such as financing, facilities, and regulatory approvals, accounting for about 24% of the total. Some future compute arrangements without fixed dollar amounts are not included in this figure.

Over the past two years, the logic most easily accepted by the market in the AI industry has been certain demand, scarce compute, increased capital expenditure, and growth in upstream orders. Under this logic, GPUs, storage, advanced packaging, servers, power, and data centers can almost all share in the growth expectations of model companies.

But recently the market has begun to ask the next question: who ultimately pays for this capital expenditure?

Since October, AI-related stocks have come under notable pressure. Goldman Sachs data shows that trend-following hedge funds last week experienced one of their worst weekly performances in two years, and whether massive AI capital expenditure can ultimately generate returns has become an important factor driving volatility in stocks and bonds. Earlier, Goldman Sachs Prime data showed that some hedge funds had net sold global AI stocks for multiple consecutive trading days, with AI positions declining markedly.

At the same time, the credit risk of new cloud providers such as CoreWeave has become an important window for the market to observe AI capital expenditure. Compared with traditional cloud providers that have stable cloud businesses, cash flow, and balance sheets, independent compute leasing providers rely more on financing and the capital markets. Once compute utilization, leasing prices, or customer orders change, the elasticity of their profits and valuations will be further amplified by leverage.

But this does not mean AI compute demand has already entered a downcycle. Goldman Sachs' recent judgment holds that political and regulatory resistance to data center construction may create supply constraints and further raise the scarcity value of existing compute infrastructure. What truly warrants vigilance are new compute companies that have not yet formed stable customer contracts but are highly dependent on financing expansion.

For the domestic market, this change also has mapping significance.

Previously, the market priced domestic GPUs more around share, compute scale, and domestic substitution, but if overseas AI leaders begin to face stricter tests of PS, cash flow, and ROIC (return on invested capital), it will also be difficult for domestic compute companies to price independently of fundamentals over the long term. Xie Siyuan judges that the valuations of some domestic projects are already close to the edge of a bubble.

At present, Anthropic's pricing is not fully determined. According to market sources, Anthropic's underwriters are preparing to raise the IPO valuation to $1.5 trillion or even higher, which means that the $70–78 issuance range Anthropic announced yesterday is not yet the final price.

At the same time, market participants revealed that Anthropic currently gives higher priority to raising enough capital rather than maximizing the IPO valuation, because the company will need large-scale investment in computing infrastructure in the future. Of the approximately $900 billion in procurement and leasing commitments already signed, about $451.6 billion is expected to occur before the end of 2030.

Source

Yicai