Anthropic Targets $1.3 Trillion Valuation in 825 Million-Share IPO

By Rebecca Stone

Anthropic plans to sell 825 million shares at $70 to $78 each, an offering that would value the frontier AI vendor at roughly $1.3 trillion and raise more than $61 billion, according to an amended S-1 filing made public Tuesday.

The proposed valuation sits below the $2 trillion figure some investors floated over the summer. It still ranks among the largest technology offerings in history. The debut will test whether a standalone AI model developer can convert rapid demand growth into a durable business, and whether Wall Street will fund it without the diversified revenue that Google and Microsoft draw from advertising, enterprise software, cloud and consumer products.

What does the IPO change for the AI market?

Going public ahead of rival OpenAI gives investors the first direct benchmark for pricing pure-play AI revenue. "Anthropic's IPO will establish the first public benchmark for how investors value AI revenue," said Monte Malukas, founder and CEO of Stockatomics, a financial data and research platform.

Malukas also flagged circular financing between AI developers and their suppliers. "Some of the demand for Anthropic's chips and cloud capacity is financed by capital from the very suppliers providing that infrastructure," he said.

Nvidia illustrates the overlap. Anthropic trains and runs its models on Nvidia GPUs, and the chipmaker has agreed to buy roughly $10 billion of Anthropic's Class N shares in a private placement after the IPO, subject to closing conditions.

Matt Hasan, an economist, AI strategist and CEO of AIresults Global Advisory, said the offering will clarify how frontier AI companies balance revenue growth, infrastructure spending and profitability. "A successful IPO doesn't mean the industry's business model is sustainable over the long term," Hasan said. "Investors will ultimately judge these companies on their financial performance, not just technological promise."

What do the financials show?

Third-quarter revenue reached $16.96 billion, up from $11.56 billion in the second quarter. The reported net loss narrowed from $66.3 billion to $23.29 billion, though accounting adjustments tied to convertible debt heavily affected both figures.

The company disclosed about $900.9 billion in total spending commitments, nearly all tied to computing capacity and infrastructure. Anthropic expects to spend roughly $451.6 billion of that by the end of 2030, with most remaining commitments extending through 2045.

Malukas called this an asset-liability mismatch. "Some infrastructure commitments extend up to 20 years, while individual AI models may have commercial lives of only one to two years," he said. "If models become obsolete faster than the contracts expire, Anthropic could end up paying for computing capacity it no longer needs."

Jacopo Viganò, head of research at equity research publisher Golden Bear Capital, cited a second cost pressure. "Anthropic bought $5 billion in compute through short-term agreements in Q3 at significantly higher costs than comparable capacity under longer-term contracts, while its adjusted operating margin fell from 19% to 15%," he said.

Longer-term deals improve unit costs but create liability for capacity Anthropic may need less of. "The valuation depends partly on how accurately it can buy future capacity before it knows future demand," Viganò said. Because much revenue comes from consumption-based services, Anthropic risks committing to more compute than customer demand ultimately supports.

Hasan said better computing efficiency will not necessarily fix this. "If Anthropic uses lower computing costs to support more workloads and increase consumption, its overall spending could continue to rise, making it harder to improve profitability," he said.

What should enterprise customers do?

Organizations running Claude for core applications, custom agents and automated workflows must weigh Anthropic's long-term stability as a vendor. IPO capital could expand compute capacity and enterprise support, and public reporting adds visibility into financial health. But margin pressure could move pricing, product access and commercial terms.

Viganò cautioned against assuming current API prices survive renewal. "Anthropic's subscription terms are generally 12 months or less, while some of its infrastructure commitments extend into the 2030s," he said. Before budgeting, customers should test whether projects remain viable at higher renewal costs and assess switching difficulty.

Malukas argued public-market incentives could shift pricing power toward the vendor. "As a public company, the lab will face pressure to demonstrate revenue growth and improve margins, which could influence how much of its declining per-task costs it passes on to customers," he said. He advised enterprises to negotiate price protections and model-deprecation notice periods into contracts and to maintain a second qualified model provider.

Hasan echoed the concentration risk. "Supplier concentration and the costs of switching providers are serious economic considerations for organizations that increasingly depend on AI models," he said.

The core question remains whether growth converts into profit. Valuation depends on existing customers spending more as pilots move to production, rather than on slower, costlier new-customer acquisition. Anthropic must offset falling per-task prices against $900.9 billion in commitments, and access to power and chips could constrain expansion.

"CIOs need to understand how long their pricing and capacity guarantees will last before making long-term commitments to an AI provider," Viganò said.