Anthropic Targets Record IPO as Claude Revenue Surpasses $65 Billion

Anthropic is preparing to file its IPO registration documents publicly as early as the end of August, according to people familiar with the matter, with internal calculations targeting a share sale that would match or exceed SpaceX's $75 billion offering — the largest initial public offering in financial history, which ultimately raised $86.2 billion after the overallotment option was exercised. The timing would set up an October market debut at a valuation that investors have independently pegged at $2 trillion or more, a figure Anthropic has not confirmed or endorsed.

A second Bloomberg report published August 20 simultaneously confirmed that Citigroup is set to join Morgan Stanley, Goldman Sachs, and JPMorgan Chase as a top-ranked bank on the deal — a sign the underwriting process is well advanced. Anthropic declined to comment on both stories.

Revenue That Grew Sevenfold in Seven Months

Anthropic's annualized revenue run rate topped $65 billion by the end of July, according to figures the company shared with investors as part of a routine update — up from roughly $9 billion at the close of 2025 and $47 billion in May. The company's preliminary second-quarter revenue exceeded $11.5 billion, more than 14 times the $787 million it recorded in the same quarter a year earlier and more than double its first-quarter revenue of $4.73 billion.

Reaching $75 billion — the minimum needed to match SpaceX's initial record raise — would require Anthropic to sell a portion of the company at or above its current investor-assessed valuation. For context, SpaceX priced its shares at a fixed $135, raising $75 billion and valuing the company at $1.77 trillion, with the total climbing to $86.2 billion after the overallotment option was exercised. Clearing the bar depends on investor demand at whatever valuation Anthropic and its banks settle on — a number the company has explicitly declined to discuss even in private investor meetings, according to Bloomberg's sources.

What the run-rate number requires context to interpret: an annualized run rate takes a single month's performance and projects it across a full year. It is a standard startup metric, not a trailing twelve-month figure. For a company growing at Anthropic's pace, the July run rate likely overstates revenue actually earned in the first half of 2026 — a distinction that becomes material once GAAP accounting is applied in a public filing.

How Anthropic Actually Makes Money: Tokens, Agents, and Inference Cost

Understanding the revenue story requires understanding the business model. Anthropic earns primarily by selling access to Claude through an API priced per token — the unit of text the model processes and generates. Enterprise customers and developers pay a tiered rate depending on the model: Claude Haiku 4.5 starts at $1 per million input tokens and $5 per million output tokens; Claude Sonnet 5 at $3 and $15; Claude Opus 5 at $5 and $25. Roughly 80% of revenue comes from this enterprise and developer channel, with the remainder from consumer subscriptions and Claude Code.

Claude Code is the product that appears to have broken Anthropic's growth curve open. Launched publicly in May 2025 as a standalone coding agent, it enables developers to describe a software task in plain language and have Claude execute it end-to-end — writing code, testing it, and iterating — without constant human direction. The product grew from approximately $500 million in annualized billings in late 2025 to an estimated $8 billion by May 2026. Third-party analysts estimated the August figure near $15 billion, though Anthropic has not confirmed that specific number.

The economic improvement that made Q2 2026 Anthropic's first-ever profitable quarter — an adjusted operating income of $559 million, according to Bloomberg — rests on a single key ratio: how much compute costs relative to each dollar of revenue. In the first quarter of 2026, Anthropic spent an estimated 71 cents of every revenue dollar on inference compute — the cost of running models to serve user requests. By the second quarter, analysts estimate that figure had fallen to roughly 56 cents. Whether this reflects genuine infrastructure efficiency gains, temporary pricing discounts from Amazon and Google on contracted compute, or a favorable mix of high-margin products is something the public S-1 will need to clarify. The first operating profit in a single quarter is not a confirmed trajectory.

Anthropic's longer-term thesis is that inference costs will continue to fall as models become more efficient — potentially reaching gross margins of 77% by 2028, according to analyst projections. The company has secured approximately 3.5 gigawatts of compute capacity through agreements with Broadcom and Google, and has committed to $100 billion in compute spending with Amazon. Bankers are reportedly underwriting the IPO against 2028 revenue projections of $190 billion to $200 billion, according to a Reuters exclusive published August 15 — more than four times the July run rate.

The Questions the S-1 Must Answer

Anthropic's revenue trajectory is real and independently confirmed. But the S-1 will introduce complexity that private investor briefings have allowed the company to defer.

The most consequential complexity is revenue recognition. Anthropic reportedly records revenue from its cloud reseller partnerships — AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure — on a gross basis, counting the full end-customer spend as top-line revenue and recording platform payouts as expenses. A net-reporting alternative would record only Anthropic's own economics after paying the platforms. Analysts estimate this distinction could inflate headline revenue figures by 20 to 30 percent. Public investors will see audited GAAP figures for the first time when the S-1 lands.

A second question concerns a legal filing. According to financial analysts reviewing the company's court documents, CFO Krishna Rao stated in a sworn filing submitted to a federal court in March 2026 that the company's cumulative lifetime revenue had "exceeded $5 billion to date." The simultaneous existence of a $47 billion annualized run rate and a $5 billion lifetime figure at that point requires a definitional reconciliation — likely reflecting the difference between recognized GAAP revenue and annualized billing projections — that the S-1 will have to provide.

Where Anthropic Stands Against OpenAI

OpenAI filed its own confidential S-1 one week after Anthropic, on June 8, 2026. It has not announced an exchange, pricing timeline, or ticker, and the New York Times has reported it could delay its debut to 2027. The two companies offer investors meaningfully different profiles.

Anthropic's July run rate of $65 billion compares to OpenAI's estimated $40 billion, a figure Bloomberg reported last week. Anthropic grew from $9 billion to $65 billion in seven months — a sevenfold increase. OpenAI roughly doubled over the same period. At its last private-market valuation of $965 billion against a $47 billion run rate, Anthropic implied a revenue multiple of approximately 20 times forward sales. OpenAI's $852 billion valuation against a $25 billion run rate at the time translated to roughly 34 times — meaning Anthropic, by this narrow measure, was the less aggressively priced of the two at their respective last private marks.

The structural difference between the companies matters more than the valuation multiple. Anthropic derives roughly 80% of its revenue from enterprise and developer API consumption — a customer base with longer contracts, higher spending per account, and lower churn than consumer products. More than 1,000 customers spend over $1 million annually on Claude, up from approximately 500 in February 2026, according to company figures. OpenAI's revenue base includes a larger share of consumer subscriptions through ChatGPT, which drives higher brand awareness but historically carries lower contract values and higher sensitivity to price competition.

Anthropic also reported a positive adjusted operating income in Q2. OpenAI is projecting an operating loss of approximately $14 billion for 2026 and has not indicated a path to profitability before 2030, according to prior media reporting.

Risks the Prospectus Is Expected to Disclose

CNBC reported on August 21 that Anthropic's IPO filing is expected to include AI societal backlash as an explicit risk factor. CFO Krishna Rao is being asked about it in preliminary investor meetings, alongside questions about open-source model competition and what happens if data center construction slows.

Several risks extend beyond boilerplate language. The Pentagon designated Anthropic a supply chain risk earlier in 2026 after the company refused to remove safety guardrails at the administration's request — specifically, restrictions preventing its AI from being used for domestic mass surveillance and fully autonomous weapons. The designation, the first ever applied to an American company, may limit Anthropic's access to certain government contracts. CEO Dario Amodei has held White House discussions about the company's Mythos model, a frontier system with cybersecurity capabilities that has not been released publicly. Anthropic has contested the supply chain risk designation in court, and Reuters reported in August that the dispute was showing signs of easing across parts of the US government.

On the competitive front, open-source AI models continue to improve, with Meta's Llama family and other open-weight systems offering enterprise developers an alternative that eliminates per-token licensing costs entirely. Anthropic's case for premium pricing rests on safety, reliability, and advanced reasoning capabilities — arguments that become harder to sustain if open-source models narrow the performance gap.

Finally, the company's $600 billion or more in projected data-center commitments through 2030 represents infrastructure spending that will need to be absorbed regardless of whether revenue projections materialize. Amazon and Google, as both investors and primary infrastructure providers, create a related-party structure that investors and regulators will examine carefully in the S-1.

What the Public Filing Will Actually Settle

A public S-1 will do something no investor briefing has done: put Anthropic's finances under independent audit and SEC review. It will resolve the gross-versus-net revenue question, establish what GAAP revenue was actually recognized rather than run-rated, and clarify the terms of Anthropic's compute agreements with its largest customers who also happen to be its largest investors.

The $2 trillion valuation target is investor-generated, not company-confirmed, and the history of pre-IPO AI company valuations suggests that public investors price assets differently from venture funds absorbing the risk of multi-billion-dollar private rounds. SpaceX's shares opened below some expectations before recovering, and the book of large AI IPOs remains thin enough that comparables are difficult to establish.

What is not in dispute is the underlying growth rate. No enterprise software company in recorded history appears to have grown from $1 billion to $65 billion in annualized revenue in nineteen months. Whether that growth rate justifies a $2 trillion valuation depends on questions the prospectus will force into the open: how much of that revenue is gross versus net, how durable the inference cost improvements are, and whether the first operating profit in Q2 represents a structural turning point or a one-quarter artifact of favorable compute pricing. The public filing that lands before the end of August will begin answering them.