Anthropic Overtakes OpenAI: The $30B ARR Milestone (2026)

TL;DR

On April 6, 2026, Anthropic disclosed that its run-rate revenue had surpassed $30 billion, up from approximately $9 billion at the end of 2025¹². That figure sits above OpenAI's roughly $24 billion annualized revenue (about $2 billion per month) as of the same window, marking the first time Anthropic has led OpenAI in disclosed annual run-rate revenue³⁴. The headline is contested: on April 13, OpenAI's chief revenue officer Denise Dresser sent a four-page internal memo arguing that Anthropic's accounting overstates the figure by roughly $8 billion, primarily through how Anthropic recognizes revenue share with AWS, Microsoft Azure, and Google Cloud partnerships⁵⁶.

The growth itself is not in dispute. Anthropic's run rate moved from $1 billion in January 2025 to $30 billion in April 2026 — a 30x jump in 15 months¹⁷ — and CEO Dario Amodei said the pace exceeded the company's own forecasts by a factor of eight⁷. The trajectory has been propelled by enterprise demand: roughly 80% of revenue comes from business customers, Claude Code alone hit a $2.5 billion run rate by February 2026, and 8 of the Fortune 10 now use Claude⁸⁹.

What You'll Learn

• What Anthropic actually disclosed on April 6, 2026, and how it compares to OpenAI's most recent ARR figures
• The full revenue trajectory: $87M (January 2024) → $1B → $9B → $30B
• Why OpenAI's CRO sent an internal memo accusing Anthropic of inflating revenue by $8B
• How Anthropic's enterprise-heavy mix differs from OpenAI's consumer-heavy mix
• What Claude Code's $2.5B run rate says about where the money is coming from
• How the funding stack ($380B Series G, up to $40B from Google, $900B talks) connects to the revenue story

What Anthropic Actually Said on April 6, 2026

The April 6 announcement was technically about compute, not revenue. Anthropic, Google, and Broadcom announced an expanded partnership for multi-gigawatt next-generation TPU capacity — a Broadcom SEC filing later disclosed the figure as 3.5 gigawatts coming online starting in 2027¹⁰¹¹. Buried in the announcement was a one-line disclosure that has become the more consequential headline:

Our run-rate revenue has surpassed $30 billion, up from $9 billion at the end of 2025, as demand for Claude continues to accelerate.¹

That single sentence pushed Anthropic past OpenAI in disclosed annualized revenue for the first time. OpenAI's most recently confirmed ARR figure — reported widely in March and April — sits at approximately $24 billion, equivalent to $2 billion per month in revenue, with Sacra estimating the company hit $25 billion in February³⁴¹². The two companies use different methodologies and reporting cadences, so the comparison is not perfectly apples-to-apples, but the relative position is what shifted.

The Revenue Trajectory

Anthropic's growth curve is the part of the story that nobody contests. The figures below come from a mix of Anthropic disclosures, Bloomberg reporting, and community trackers, and they line up across sources:

From $1 billion in January 2025 to $30 billion in April 2026 is a 30x move in 15 months. CEO Dario Amodei said the trajectory exceeded Anthropic's own forecasts by a factor of eight⁷. For context, Salesforce took roughly two decades to reach a comparable annual revenue figure⁷.

OpenAI's revenue is also growing rapidly, just from a higher absolute base earlier in the curve. Its disclosed trajectory: $2B (2023) → $6B (2024) → over $20B (end of 2025) → roughly $24B ($2B/month) in early 2026⁴¹⁶¹⁷. The reason the gap has compressed is not that OpenAI slowed down; it is that Anthropic accelerated faster.

Where the $30 Billion Is Coming From

Anthropic's revenue mix is the structural difference that explains both how it caught up and why OpenAI's CRO is contesting the headline.

The most recent breakdown the company has shared publicly (October 2025) shows roughly 80% of revenue comes from business customers — enterprises buying Claude via the API or through Claude for Work, Claude Code, and similar offerings — with the remaining ~20% from consumer subscriptions (Claude Pro, Max)⁸¹⁸. API-driven enterprise consumption accounts for around 70–75% of total revenue⁸.

OpenAI's mix is inverted. The company derives a large share of revenue from consumer subscriptions (ChatGPT Plus, Pro, and Team), with enterprise making up around 40% — up from roughly 30% a year earlier, but still well below Anthropic's enterprise concentration⁸¹⁷.

Within Anthropic's enterprise mix, one product is doing disproportionate work:

Claude Code launched publicly in May 2025 and became, by Anthropic's own account, the fastest-growing product in the company's history⁹¹⁵. By the company's reporting in early 2026, business subscriptions had quadrupled since the start of the year, and the number of customers spending $1 million or more annually crossed 1,000 in April 2026, roughly double the figure from February¹⁵²⁰.

Ramp's May 2026 AI Index (reporting April 2026 adoption data) showed a milestone in business adoption: Anthropic at 34.4% of business AI usage versus OpenAI at 32.3% — the first time Anthropic crossed OpenAI in that index²¹. Counterpoint Research's Q1 2026 LLM revenue analysis placed Anthropic at 31.4% global market share to OpenAI's 29.0%²².

The $8 Billion Accounting Dispute

On April 13, 2026, OpenAI's chief revenue officer Denise Dresser sent a four-page internal memo to OpenAI employees. The memo, which was widely leaked and reported on, argues that Anthropic's $30 billion figure is overstated by roughly $8 billion⁵⁶²³.

The mechanic Dresser objects to is how Anthropic recognizes revenue tied to its cloud-partner relationships across AWS, Microsoft Azure, and Google Cloud. When customers run Claude through partners — for example, through Amazon Bedrock, Microsoft Foundry on Azure, or Google Cloud's Vertex AI — Anthropic reports the gross amount of that consumption as Anthropic revenue, rather than recognizing only the net share that flows back to Anthropic after the partner's cut. Anthropic's position is that it is the principal in those transactions under standard accounting rules (ASC 606), which justifies gross-basis recognition⁵²³.

The arithmetic of the dispute is straightforward:

There is no independent verification of either treatment, and Anthropic has not publicly responded to the memo's specific allegations⁵⁶. Both companies are reportedly preparing for IPOs — Bloomberg's May 12 reporting suggests Anthropic is considering an offering as early as October 2026 — and how they report ARR in S-1 filings will be the first chance to resolve the dispute through audited disclosures²⁴.

For now, the most defensible framing is: Anthropic's gross run rate, on its own reporting basis, exceeded OpenAI's gross run rate in April 2026. The methodologies are not yet harmonized.

The Capital and Compute Stack

The revenue figures sit on top of a fundraising and infrastructure stack that has been moving almost as fast as the revenue itself. The relevant dates compressed into one timeline:

Two things stand out. First, Anthropic is running a multi-cloud compute strategy in parallel — Google Cloud (TPUs), AWS (Trainium and Nvidia GPUs), and Microsoft Azure (Nvidia GPUs via a $30B Azure compute commitment announced November 2025) — rather than committing to a single hyperscaler¹⁰²⁶. Second, the funding cadence is unusual: a $30B Series G in February at a $380B valuation, followed by talks of a $30B raise at $900B by May, would be a roughly 2.4x markup in three months. That kind of pace is normally associated with hype cycles; here it tracks an underlying tripling of run-rate revenue ($9B → $30B) in four months.

The Training-Cost Picture

One of the more analytically interesting comparisons is on training spend per dollar of revenue. Based on each company's disclosed long-term training cost projections — OpenAI at roughly $125 billion per year by 2030, Anthropic at roughly $30 billion per year by 2030 — Anthropic plans to spend approximately 4x less than OpenAI on training over the long run while currently generating comparable revenue²⁸.

This is a projection, not a current-period spend ratio, and both numbers will shift as model scales and inference economics evolve. But it points at a strategic divergence: OpenAI is positioning its consumer-product-led approach as one that requires very large frontier-training investments to defend, while Anthropic's enterprise-API model treats training spend more as a controlled production input than as a brand differentiator²⁸.

What This Actually Means

The crossover is significant for three reasons, none of which is "Anthropic has won."

First, it shifts the narrative inside enterprise buyers. For two years, "OpenAI is the market leader" was a defensible default in procurement decisions. As of Q2 2026, both Counterpoint Research's revenue data and Ramp's business-usage index put Anthropic ahead in the segments that matter for enterprise budgets²¹²². That changes the comparison from "OpenAI vs. challenger" to "two roughly comparable frontier vendors," with all the procurement-leverage implications that follow.

Second, it accelerates the consumer-vs-enterprise specialization in AI. OpenAI's roughly $24B ARR is heavily powered by ChatGPT subscriptions; OpenAI disclosed on February 27, 2026 that ChatGPT had reached 900 million weekly active users²⁹. Anthropic's roughly $30B is built almost entirely on enterprise API consumption and developer tooling. Both companies are technically frontier labs, but the businesses they are building underneath the labs are diverging rapidly.

Third, the accounting dispute matters for IPO-era discipline. If both companies file S-1s within 12–18 months, audited ARR figures will be reported on a standardized basis (likely revealing the gross-vs-net distinction Dresser raised). That filing will be the first test of whose framing the SEC's review process — and public-market investors — find acceptable. Until then, ARR comparisons should be read with a footnote attached.

The most important takeaway is also the simplest: in 15 months, Anthropic went from a credible but distinctly second-place AI lab to a company contesting the revenue lead. Whether that lead is real or accounting-driven matters less to enterprises right now than the fact that the conversation is taking place at all.

Bottom Line

Anthropic disclosed a $30 billion run-rate revenue on April 6, 2026 — passing OpenAI's roughly $24 billion ARR on a like-for-like reporting basis — driven by enterprise API consumption (80% of revenue), Claude Code's $2.5B run rate, and broad Fortune-1000 adoption¹⁸¹⁵. OpenAI's chief revenue officer has contested the headline number, arguing the figure is inflated by approximately $8 billion through gross-vs-net recognition of cloud-partner revenue⁵. The dispute won't be definitively resolved until both companies file audited disclosures (likely in IPO prospectuses), but the underlying growth — 30x in 15 months, exceeding Anthropic's own forecasts by a factor of eight — is not in serious dispute⁷²⁴.

For enterprise buyers, the practical takeaway is that the AI-vendor landscape has shifted from "one clear leader plus challengers" to "two roughly comparable frontier vendors with diverging product strategies." For investors, the more interesting metric to track over the next two quarters is not who is reporting the higher gross ARR, but how each company describes its revenue once auditors get involved.