Anthropic Revenue Run Rate Nears $100 Billion as Its November IPO Test Looms

Anthropic’s revenue run rate will reportedly exceed $100 billion this year as the Claude maker prepares for a possible November stock-market debut. The figure, reported by The New York Times and relayed by Bloomberg, would place Anthropic among the fastest-growing technology companies ever recorded.

That headline creates an extraordinary first impression. Anthropic reported a $47 billion run rate in May and reportedly exceeded $65 billion by late July. Reaching $100 billion would represent another sharp increase before public investors can examine the underlying contracts, expenses, and revenue recognition choices.

That last point matters because a run rate is not the same as annual revenue. It annualizes a recent sales period to estimate what a company would generate if that pace continued. The metric captures momentum, but it does not guarantee that customers will maintain their spending for a full year.

The reported growth also intensifies Anthropic’s contest with OpenAI. Claude’s strength in coding and enterprise work has helped Anthropic overtake its better-known rival on several reported financial measures. An IPO would move that comparison from private briefings into audited public disclosures.

Anthropic Revenue Run Rate Approaches a New Threshold

The reported $100 billion figure signals exceptional momentum, but it remains a projection based on Anthropic’s recent sales pace.

The New York Times reportedly cited people familiar with Anthropic’s finances when describing the company’s latest trajectory. Anthropic declined to comment on the report, according to Bloomberg. No public filing currently verifies the precise $100 billion figure.

The supporting trend is less speculative. Anthropic said in May that its run-rate revenue had crossed $47 billion. The company disclosed that number while announcing a $65 billion financing round at a $965 billion post-money valuation.

Anthropic attributed the increase to enterprise adoption across industries. It said organizations were putting Claude into core operations, while individuals increasingly used the system for everyday work. The company also highlighted Claude Code and Cowork as important products serving professional users.

By late July, Anthropic’s reported run rate had surpassed $65 billion. Documents described by Axios showed more than $11.5 billion in preliminary second-quarter revenue. That was more than double the reported $4.73 billion from the first quarter.

Those quarterly figures provide a firmer foundation than an isolated annualized estimate. They suggest that Anthropic’s growth reflects collected or recognized sales across a meaningful period, rather than one unusually strong week. However, the company remains private, and outsiders cannot yet inspect a complete income statement.

The progression is striking. Anthropic reported a $14 billion run rate in February, $47 billion in May, and more than $65 billion by July. The latest report places its expected annualized pace above $100 billion before year-end.

Such a sequence does not mean Anthropic will record $100 billion of revenue during 2026. A company reaching that run rate in December would have operated at lower levels during earlier months. Full-year recognized revenue would therefore remain below the year-end annualized figure.

Seasonality can also distort the calculation. Large enterprise contracts may arrive in clusters, while product launches can produce temporary usage increases. Reuters Breakingviews observed in August that Anthropic’s monthly expansion had already shown uneven periods despite its wider upward trend.

The latest claim still matters because it changes the burden of proof. Anthropic is no longer presenting Claude as a promising product searching for a market. It is presenting an enterprise AI business operating at the scale of an established technology leader.

That framing will follow the company into its IPO. Public investors will want to know how much of the run rate comes from recurring consumption, committed contracts, cloud resellers, and customers experimenting with unusually large AI budgets.

Enterprise Claude Adoption Is Driving the Surge

Anthropic’s growth appears tied to paid work inside companies, giving its revenue story more substance than consumer attention alone.

Coding has become one of Claude’s most important commercial use cases. Developers use Claude Code to inspect repositories, generate changes, troubleshoot failures, and coordinate longer engineering tasks. These activities consume substantial computing capacity because they involve repeated model calls and large amounts of context.

That usage model creates a direct connection between productive work and revenue. A developer who keeps an agent active throughout the day generates more consumption than someone occasionally asking a chatbot a simple question. Anthropic benefits when Claude becomes part of a recurring development process.

The same pattern is spreading beyond software engineering. Companies are testing Claude in customer support, research, legal review, financial analysis, life sciences, and internal operations. These deployments can generate sustained demand when they move from demonstrations into routine workflows.

Anthropic has worked to meet that demand through several cloud channels. The company says Claude is available through Amazon Web Services, Google Cloud, and Microsoft Azure. That distribution lets enterprise customers buy model access within infrastructure relationships they already maintain.

The strategy reduces friction for procurement, security review, and billing. It also means Anthropic shares parts of its commercial relationship with cloud platforms. Those arrangements become important when comparing its revenue with competitors using different accounting methods.

Anthropic’s May funding announcement also described agreements for large amounts of future computing capacity. AWS remained its primary cloud and training partner, while Google, Broadcom, and SpaceX were identified as additional infrastructure providers.

Those commitments reveal the mechanism beneath the sales growth. Anthropic needs more accelerators, memory, networking, power, and data-center capacity whenever customers send more work to Claude. Revenue expansion therefore brings an unusually direct increase in infrastructure demand.

This relationship separates an AI laboratory from a conventional software company. Traditional software can often serve additional users at modest incremental expense. Frontier models require expensive inference, which is the computing performed whenever a deployed model generates an answer or takes an action.

Higher usage is valuable only when revenue grows faster than the cost of serving it. Anthropic must therefore improve model efficiency, negotiate favorable infrastructure agreements, and guide customers toward tasks that produce measurable economic value.

Enterprise buyers face a related calculation. A model can appear expensive per request while reducing the cost of a completed project. Conversely, inexpensive access provides little value if employees must repeatedly correct unreliable results.

Harrison Rolfes of PitchBook told Axios that a premium model can deliver a lower cost per completed task when users need fewer retries or human corrections. That argument helps explain why professional adoption can support high revenue even amid competition from cheaper models.

Still, customers will eventually demand evidence beyond impressive demonstrations. They will compare Claude’s output with labor costs, failure rates, security risks, and competing systems. Teams building a searchable knowledge base, for example, need reliable retrieval and traceable answers rather than raw model capability alone.

The reported Anthropic revenue run rate therefore measures more than popularity. It suggests that companies are paying for AI to perform recurring work. The IPO will test whether that spending has become durable infrastructure or remains an unusually intense adoption cycle.

The OpenAI Comparison Is Not Apples to Apples

Anthropic appears to lead OpenAI on reported run-rate revenue, but different accounting treatments complicate the apparent gap.

OpenAI provides the clearest competitive reference because it created the modern consumer chatbot market and remains a defining AI brand. Yet Anthropic has built a stronger reported position in enterprise and coding revenue, placing OpenAI under pressure before either company becomes public.

In August, OpenAI’s reported annualized revenue stood near $40 billion. Anthropic’s reported figure had already passed $65 billion. The new projection would widen that headline difference considerably if OpenAI’s pace did not rise at a similar rate.

The numbers cannot be compared without understanding how the two companies record cloud-partner sales. Anthropic reportedly recognizes the full value of certain Claude transactions completed through cloud providers. It then records the provider’s share as an expense.

OpenAI reportedly recognizes only its own portion of some comparable sales. One company can therefore show higher top-line revenue even when the customer spends the same amount and the model provider ultimately retains a similar share.

Accounting professor Francine McKenna explained the distinction through the principal-versus-agent question. The central issue is which company controls the customer relationship and bears responsibility for delivering the service.

Neither treatment is automatically improper. Contract terms and operational responsibilities can support different conclusions. However, those choices affect the revenue figure investors see before examining expenses.

The reported difference does not erase Anthropic’s lead. An accounting analysis estimated that switching Anthropic from gross to net presentation would have reduced its earlier reported run rate by approximately 6% to 10%. Based on the figures available then, Anthropic would still have remained well ahead.

The comparison becomes more meaningful when investors examine gross profit. Gross profit subtracts the direct cost of delivering a service from recognized revenue. It can show whether high sales translate into attractive economics after paying infrastructure partners.

Cash flow will add another layer. Anthropic may recognize revenue from customers while committing much larger sums to future computing capacity. Investors must separate current operating performance from the capital needed to train and serve future models.

OpenAI also retains significant competitive advantages. ChatGPT has broad consumer recognition, an enormous user base, and growing enterprise adoption. OpenAI can introduce new models or distribution agreements that alter the revenue comparison within a quarter.

Anthropic’s current advantage rests more heavily on professional tasks. Claude’s reputation in coding has positioned the company where model usage can become frequent, measurable, and deeply embedded. That focus appears to have converted technical preference into enterprise spending.

Going public first would strengthen Anthropic’s position in another way. A successful offering could provide capital and a publicly traded currency for infrastructure agreements, acquisitions, and employee compensation. It would also establish a market valuation before investors evaluate OpenAI through the same process.

However, first-mover status comes with exposure. Anthropic would become the first major independent frontier-model company required to report detailed quarterly results. Any slowdown, margin disappointment, customer concentration, or unexpected infrastructure expense would be visible.

The rivalry is therefore shifting. Anthropic and OpenAI once competed mainly through model benchmarks, chatbot features, and developer enthusiasm. Their next contest will involve revenue quality, gross margins, cash generation, and the credibility of their financial disclosures.

A $100 Billion Headline Cannot Answer the Margin Question

The central uncertainty is not whether Anthropic is selling more AI, but how much economic value remains after the cost of providing it.

Run-rate revenue is useful for describing direction. It is much less useful for assessing profitability because it says nothing about direct costs, future commitments, customer concentration, or the timing of cash collections.

Anthropic’s models operate on costly infrastructure. Each customer interaction consumes computing resources, and complex agentic tasks can involve long conversations, tool calls, retries, and extensive context. Growing demand can increase both revenue and expense at the same time.

The company has pursued an unusually large supply plan to support its expansion. Its May announcement referenced agreements with Amazon for up to five gigawatts of capacity and with Google and Broadcom for another five gigawatts involving next-generation TPUs.

A gigawatt measures power, not currently installed computing performance. These agreements therefore indicate planned scale rather than immediately available capacity. They also underline how much physical infrastructure lies behind a seemingly digital service.

The IPO filing should reveal how Anthropic accounts for those commitments. Investors need to see which agreements are firm obligations, how long they run, and whether minimum payments apply when customer demand weakens.

Revenue concentration presents another question. A small number of cloud platforms or large enterprises might account for a substantial share of sales. That arrangement can accelerate growth but give important customers more negotiating leverage.

Consumption-based revenue can also fluctuate. A company might expand an AI pilot rapidly, then reduce usage after reviewing expenses or output quality. Another customer might switch models when a competitor delivers adequate performance at a lower cost.

These risks do not invalidate the reported growth. They explain why a $100 billion run rate cannot function as a complete valuation argument. Public investors will assess how reliably Anthropic converts usage into retained revenue and cash.

Anthropic has already taken the formal first step toward disclosure. The company submitted a confidential S-1 to the US Securities and Exchange Commission on June 1. It said the offering remained subject to SEC review, market conditions, and other factors.

The filing did not set a share count or offering price. Confidential review allows the company and regulators to address disclosures before a public prospectus appears. The process does not guarantee that the IPO will happen in November.

Once the registration statement becomes public, investors should receive audited financial statements and detailed risk factors. They should also learn more about related-party arrangements, stock-based compensation, infrastructure contracts, and the company’s corporate structure.

Revenue recognition will receive particular attention. Anthropic must explain why it treats itself as the principal in cloud-mediated sales and how that policy complies with accounting standards. SEC correspondence released later can provide additional context.

The company’s public-benefit status adds another layer. Anthropic PBC describes safety as central to its mission, while its rapid commercial expansion requires enormous capital and continued product releases. Public shareholders may press the company to reconcile those commitments when they create short-term financial tension.

The skepticism should remain proportionate. Anthropic’s progression from $47 billion to more than $65 billion was supported by company statements and financial reporting. The unverified part is the precise timing and durability of the reported advance toward $100 billion.

That distinction matters. The story is not that Anthropic invented a financial number from nothing. It is that a private company’s extraordinary growth claim has reached the point where only audited disclosures can answer the most important questions.

Three Signals Will Define Anthropic’s November IPO

The public filing, customer retention, and inference economics will determine whether Anthropic’s revenue surge supports a durable public company.

The first signal is Anthropic’s public registration statement. It should show recognized quarterly revenue rather than only an annualized snapshot. Investors can then compare reported sales with gross profit, operating expenses, losses, cash flow, and infrastructure obligations.

That filing should also clarify the difference between gross and net revenue. If Anthropic provides enough detail about cloud sales, analysts can reconstruct a more comparable measure for OpenAI and other model providers.

Strong audited numbers would reinforce the reported Anthropic revenue run rate. A large gap between the promotional run rate and recognized revenue would weaken it, especially if seasonal contracts explain much of the increase.

The second signal is customer retention. Anthropic needs enterprises to expand real deployments after initial trials, not simply sign short-lived contracts during an AI investment rush. Net revenue retention, remaining performance obligations, and customer concentration would help measure that durability.

Enterprise case studies will be useful only when they describe completed work and measurable adoption. Announcements about experiments or partnerships do not prove that customers will maintain consumption after reviewing their return on investment.

Buyers should watch their own usage as carefully as Anthropic’s headline numbers. The relevant questions concern completed tasks, error rates, employee time saved, oversight requirements, and total inference expense. A model that performs well in a benchmark can still fail an operational cost test.

The third signal is inference economics. Anthropic must show that rising model usage creates improving margins rather than an equally fast increase in computing bills. Efficiency gains can come from better hardware, model optimization, caching, routing, and more favorable infrastructure agreements.

Reuters Breakingviews argued that slower sequential growth could have an upside if it helps Anthropic improve profitability before listing. That observation captures the tradeoff facing the company. Maximum expansion is not automatically the best outcome when every additional task carries a significant service cost.

Competitor responses will influence those economics. OpenAI can reduce prices, introduce more capable models, or use its distribution partners to win enterprise contracts. Google can combine Gemini with its cloud platform and workplace software. Open-weight models can place additional pressure on routine workloads.

Anthropic’s position remains formidable. It has translated Claude’s professional appeal into reported financial scale, secured vast amounts of capital, and formally started the IPO process. Its progress has already forced competitors and investors to reconsider assumptions about who leads commercial AI.

The November timing remains reported rather than guaranteed. Market conditions, SEC review, financial preparation, or a strategic decision could delay the offering. Anthropic itself has only confirmed that it submitted a confidential filing and retains the option to proceed.

That uncertainty makes the next disclosure more important than another private revenue update. The market needs numbers that reconcile Anthropic’s sales pace with its cloud accounting, compute costs, and cash requirements.

For developers and enterprise buyers, the outcome will affect more than one company’s valuation. A successful IPO backed by durable usage would strengthen the case that AI agents have become a major software market. Weak margins or unstable retention would show that adoption has moved faster than sustainable economics.

Watch the public S-1 first, then compare audited revenue with the reported run rate. Next, examine whether customers keep expanding Claude deployments. Finally, follow gross margins and infrastructure commitments. Those signals will reveal whether Anthropic is approaching November with a historic business or merely a historic headline.