Anthropic IPO: Revenue, Valuation and the Cost of AI

Anthropic could soon become one of the most valuable companies ever to enter the stock market. But the most interesting number in its IPO may have nothing to do with the share price.

For years, investors have been pouring money into artificial intelligence through Microsoft, Alphabet, Nvidia and other established technology companies. Those businesses offer exposure to the AI boom, but they also make it difficult to isolate the economics of developing advanced AI models. Microsoft’s cloud business includes far more than AI. Alphabet has search advertising. Nvidia sells the hardware powering the entire industry.

Anthropic is different. The company behind Claude and Claude Code is far more directly exposed to the business of building and selling advanced AI systems. Its financial statements could offer an unusually clear view of an industry that has attracted enormous investment without yet establishing what sustainable profitability looks like at scale.

That opportunity is getting closer. Reuters reported on October 22 that Anthropic was pressing ahead with plans for a possible mid-November 2026 IPO, with a public prospectus potentially arriving in late October. The timetable remains subject to change, and no final offering price or valuation has been established.

Meanwhile, financial information reviewed by Reuters already points to a business growing at an extraordinary pace while committing staggering sums to computing infrastructure. The Anthropic IPO could become a referendum on whether the economics of frontier AI can eventually justify the money being invested in it.

Anthropic’s IPO is approaching, but the deal is not done

Anthropic confirmed in June 2026 that it had confidentially filed for a US initial public offering. A confidential submission allows a company to begin the regulatory review process before publishing its registration documents for investors.

That is an important distinction. Filing confidentially does not mean shares have been priced, an IPO has been completed or the company has received a final market valuation.

The timetable has already shifted. Earlier reports pointed toward a possible September or October debut. By October 22, Reuters was reporting that Anthropic was targeting the middle of November, potentially beginning investor marketing during the week of November 9.

The company is attempting to list in a less welcoming market than the one that supported some of the AI industry’s earlier fundraising rounds. Higher interest rates and growing scrutiny of AI infrastructure spending have contributed to pressure on several technology companies. Reuters reported that other planned listings had been postponed or withdrawn.

Valuation expectations have also moved. Anthropic raised $65 billion in May 2026 at a confirmed private post-money valuation of $965 billion. Reports later described ambitions approaching $2 trillion, but Reuters reported on October 15 that some investors were discussing a figure closer to $1.5 trillion.

Those figures should not be treated as interchangeable. The May valuation was attached to a completed private financing round. The later numbers reflect reported expectations and investor discussions, not an agreed public offering price.

For prospective investors, the difference could be enormous. But the more revealing question is what financial performance would be required to support any of those valuations.

The company behind Claude is already generating serious revenue

Anthropic began as a research-focused AI company, but its commercial business now extends well beyond a chatbot subscription.

Claude is available to individual users through paid plans, while businesses can access the company’s models through enterprise products, developer APIs and cloud distribution partnerships. Claude Code has also become an important product for software development, allowing users to delegate increasingly complex programming tasks to AI.

This business model matters because different customers generate different kinds of revenue and costs.

A consumer subscription offers relatively predictable monthly billing, but usage can vary dramatically between customers. An enterprise contract may involve larger commitments, administrative controls and integrations into corporate systems. API customers often pay according to how much processing they consume, which can connect revenue more directly to computational demand.

These are not economically identical products, even when they rely on the same underlying models.

Anthropic’s growth has been remarkable. In May, the company reported that its annualized revenue run rate had exceeded $47 billion. Reuters subsequently reported that the figure had surpassed $65 billion by the end of July.

However, an annualized run rate is not the same thing as revenue already earned over twelve months. It extrapolates a recent level of business activity into a hypothetical full year. In a company growing this quickly, that measure can convey momentum, but it should not be confused with audited annual sales.

More concrete figures emerged in October. According to Reuters, Anthropic told prospective investors that third-quarter 2026 revenue reached $16.8 billion, up from $11.6 billion in the second quarter and $4.7 billion in the first.

That represents approximately 45% quarter-over-quarter growth in the third quarter. It is an exceptional expansion for a company operating at this scale.

But revenue growth is only half the story. The other half is what Anthropic must spend to produce it.

The cost of building frontier AI is almost impossible to ignore

Traditional software businesses can become extremely profitable because distributing another copy of a finished product costs relatively little. Once the software has been developed, the marginal cost of serving another customer can be modest.

Frontier AI operates differently.

Training advanced models requires large clusters of specialized processors, extensive engineering work and substantial electricity. After training, the models must continue running whenever customers send requests. This process, known as inference, consumes computing resources every time the service is used.

That means successful AI products can generate considerable ongoing costs precisely because customers use them heavily.

According to a September 28 Reuters investigation based on Anthropic’s draft IPO prospectus, the company spent $7.33 billion on computing and infrastructure in 2025, roughly three times its 2024 expenditure. Total operating expenses reached approximately $12.65 billion.

Revenue grew twelvefold to nearly $4.6 billion in 2025, yet the company recorded an operating loss exceeding $8 billion, excluding certain accounting adjustments associated with financing arrangements.

The reported net loss was approximately $42 billion, although Reuters explained that roughly $34 billion reflected an accounting charge related to the increasing estimated value of financing instruments that could convert into Anthropic shares. That charge should not be confused with cash spent operating the business.

This distinction is essential. An IPO prospectus may contain several different measures of profitability, each answering a different question.

Net income reflects accounting rules, including potentially substantial noncash adjustments. Operating income offers a view of the underlying business before some financing-related effects. Adjusted profitability can exclude selected expenses, but its usefulness depends on exactly what management removes.

Cash flow reveals something else again: how much money the company actually generates or consumes through its operations and investment commitments.

For Anthropic, the central financial question is whether revenue can grow faster than the cost of training, operating and continuously improving its models.

Why Anthropic’s $518 billion infrastructure commitments matter

One of the most striking disclosures reported by Reuters concerns Anthropic’s future computing requirements.

The draft prospectus described approximately $518 billion in cloud, computing and infrastructure obligations extending into future years. This is a reported contractual commitment figure, not a statement that Anthropic has already spent $518 billion or that the entire amount will immediately appear as capital expenditure on its balance sheet.

The distinction matters because Anthropic can obtain computing capacity through cloud agreements rather than owning every data center and processor itself.

Amazon is a major investor and infrastructure partner, while Anthropic has also developed relationships across the broader computing ecosystem. These partnerships give the company access to the enormous resources required to compete with OpenAI, Google and other frontier-model developers.

They also create dependencies.

Anthropic must secure enough computing capacity to train competitive models and serve customers reliably. But commitments made during a period of explosive demand could become burdensome if model prices fall, competitors gain market share or customers reduce their usage.

Investors should therefore distinguish between computing capacity already paid for, future contractual obligations, operating expenses and capital expenditures. Each has different implications for liquidity and financial flexibility.

A company can report rapid revenue growth while simultaneously increasing its exposure to long-term infrastructure costs. The crucial question is whether those obligations translate into profitable customer demand.

The prospectus could reveal whether AI margins are improving

Anthropic reportedly told investors in October that it had achieved adjusted profitability during the third quarter. That would represent an important development, but it requires careful interpretation.

Adjusted profitability does not necessarily mean positive net income under generally accepted accounting principles, nor does it guarantee positive free cash flow.

Investors will need to examine how the company calculates its adjusted results, which expenses are excluded and whether the improvement can be sustained.

The relationship between revenue and inference costs will be especially revealing.

If newer models become more efficient, Anthropic may be able to serve the same workload using fewer computing resources. That could improve margins even as usage expands. Better software optimization, more efficient hardware and improved model architectures could all contribute.

But customers may respond to lower prices by demanding more powerful models, longer reasoning processes and larger workloads. Anthropic could then find itself delivering substantially more computation without capturing a proportional increase in revenue.

There is also competitive pressure. OpenAI, Google and other developers regularly release new models, while open-weight alternatives offer businesses different ways to deploy AI. Features that command premium prices today may become standard capabilities tomorrow.

The prospectus should help investors understand whether Anthropic is developing durable pricing power or participating in a market where technical improvements are rapidly passed on to customers through lower prices.

Those two outcomes could produce very different financial businesses, even if AI adoption continues accelerating.

Customer concentration is another risk hiding behind the growth

Enterprise adoption is a major part of Anthropic’s commercial appeal. Large organizations can integrate Claude into software development, customer service, research and internal workflows, creating substantial recurring demand.

However, enterprise revenue is not automatically stable.

Reuters reported that nearly one-quarter of Anthropic’s 2025 revenue came from just two customers. The draft prospectus also warned that some major customers were not committed under long-term agreements and could reduce or discontinue spending.

That concentration creates a risk that is easy to overlook when focusing on headline growth rates.

If a large customer switches models, renegotiates pricing or decides to build more AI capabilities internally, the effect on revenue could be significant. A customer might also distribute workloads across multiple AI providers rather than relying exclusively on Claude.

Investors should pay attention to the proportion of revenue coming from the largest customers, the duration of commercial commitments and the relationship between direct enterprise sales and distribution through cloud platforms.

Another important distinction is between customers paying for AI experimentation and customers embedding it into essential business processes. The latter may offer more durable demand, but the prospectus will need to provide evidence rather than simply describe growing adoption.

Why Anthropic is not another Microsoft or Alphabet

Microsoft, Alphabet and Meta have all committed enormous resources to artificial intelligence. Yet buying shares in those companies is not equivalent to investing directly in the economics of frontier AI models.

Microsoft has a diversified software and cloud business. Alphabet combines search advertising, YouTube, cloud computing and other operations. Meta’s advertising business generates cash that can support its AI infrastructure ambitions.

These established revenue streams provide financial resilience. They also make it difficult to isolate how much value comes specifically from advanced AI development.

Anthropic offers a more concentrated proposition. Its success depends heavily on the commercial value of Claude, its developer tools and the broader platform built around its models.

That concentration could make the company easier to analyze as an AI business, but it also exposes investors more directly to changes in model economics.

If inference becomes cheaper and enterprise demand remains strong, Anthropic could benefit substantially. If AI models become increasingly interchangeable and pricing collapses, the company may face pressure that a diversified technology giant can absorb more easily.

The same concentration that makes Anthropic an attractive benchmark for the industry also makes its risks unusually visible.

How much is Anthropic actually worth?

Reported valuation discussions have ranged from approximately $1.5 trillion to $2 trillion, while the company’s completed May financing established a $965 billion private valuation.

Those figures reflect different moments and different circumstances. None should be mistaken for a final IPO valuation.

Reuters reported in August that Anthropic was projecting approximately $190 billion to $200 billion in revenue for 2028, according to people familiar with its financial forecasts. Bankers and investors were reportedly examining valuation multiples based on those future expectations.

That approach is understandable for a business growing so quickly that historical revenue may become outdated almost immediately. It is also risky because the calculation depends on assumptions about growth several years into the future.

Revenue projections are not guaranteed outcomes. Even if Anthropic reaches the forecast sales level, investors would still need to determine how much of that revenue translates into operating profit and cash generation.

A high revenue multiple can sometimes be justified by strong margins, recurring customer relationships and a durable competitive advantage. It becomes harder to defend if growth requires continuously rising infrastructure spending or if pricing power deteriorates.

The IPO prospectus should make it possible to examine the assumptions behind the valuation rather than relying on the size of the number itself.

The unusual governance question behind Anthropic’s IPO

Anthropic’s corporate structure adds another dimension to the offering.

The company operates as a public benefit corporation, a structure intended to allow consideration of its stated public-benefit mission alongside shareholder interests.

Reuters reported in September that its draft IPO documents described a proposed Founder LLC governance arrangement involving Anthropic’s co-founders. The structure was designed to preserve influence over the company’s mission as it approached the public markets.

For investors, that raises questions about voting rights, accountability and the relationship between financial objectives and the company’s AI safety commitments.

A shareholder might agree with Anthropic’s long-term approach to safety while still wanting to understand how strategic decisions are made, who can challenge management and whether public investors will have meaningful influence.

These questions are particularly relevant for a company operating in an industry where product development, regulation and safety concerns can directly affect commercial performance.

What the Anthropic IPO could teach us about the AI revolution

The most important result of Anthropic becoming a public company would not necessarily be the amount raised or the valuation achieved on its first trading day.

It would be the emergence of a more detailed financial record for one of the companies developing the world’s most advanced commercial AI models.

Investors could begin tracking how revenue growth compares with computing expenditure, whether customer relationships become more durable, how infrastructure commitments evolve and whether profitability improves as the business expands.

Quarterly financial reporting would also make it easier to examine the gap between AI industry forecasts and actual operating performance.

That transparency could have consequences far beyond Anthropic. It would provide a new reference point for evaluating OpenAI and other privately held developers, as well as the enormous infrastructure investments being made by cloud providers and semiconductor companies.

But one company’s results would not settle the economics of the entire industry. Anthropic has its own customer base, products, financing arrangements and infrastructure partnerships. Other AI developers may operate under materially different conditions.

There is also a larger question that no IPO prospectus can answer immediately: whether today’s most advanced AI models will retain enough differentiation to command attractive prices as the technology improves and competitors catch up.

Anthropic’s commercial growth suggests that businesses are willing to pay substantial amounts for advanced AI. Its reported infrastructure obligations show how expensive meeting that demand can become.

The Anthropic IPO could finally put both sides of that equation in front of public-market investors. The valuation will attract the headlines. The margins, cash flow and computing commitments will tell the more important story.

Anthropic IPO FAQ

Is Anthropic publicly traded?

No. As of October 24, 2026, Anthropic has not completed its IPO. It confirmed a confidential filing in June, and Reuters has reported that a public prospectus and potential November listing are being planned, subject to change.

What is Anthropic’s IPO valuation?

No final IPO valuation has been confirmed. Anthropic raised capital at a $965 billion post-money private valuation in May 2026. More recent reported IPO valuation discussions have included approximately $1.5 trillion and earlier expectations approaching $2 trillion.

How much revenue does Anthropic generate?

According to Reuters, Anthropic reported $16.8 billion in revenue for the third quarter of 2026, compared with $11.6 billion in the second quarter. These figures were reported from a presentation to prospective investors, rather than a publicly filed quarterly earnings report.

Is Anthropic profitable?

Reuters reported that Anthropic told investors it was profitable on an adjusted basis in the third quarter of 2026. That is not the same as confirmed GAAP net profitability or positive free cash flow. Its full financial statements and the definitions of adjusted metrics will be important.

How does Anthropic make money?

Anthropic earns revenue through Claude subscriptions, enterprise services, developer API usage and products such as Claude Code, including distribution through cloud platforms.

Why does the Anthropic IPO matter for AI investors?

It could provide a relatively direct view of the revenue, computing costs, customer concentration, financing requirements and profitability of a frontier AI developer. These economics are harder to isolate inside diversified technology companies.