Anthropic $30B ARR: How Claude Overtook OpenAI in Revenue

TL;DR

Anthropic announced on April 7, 2026 that its annualized revenue run rate has crossed $30 billion, up from roughly $9 billion at the end of 2025 — a tripling in about four months¹. The figure places Anthropic above OpenAI's ~$25 billion ARR for the first time in either company's history². Alongside the revenue milestone, Anthropic disclosed that enterprise customers spending $1 million or more per year doubled from 500 to over 1,000 in less than two months since its Series G fundraise, and that it has expanded a compute partnership with Google and Broadcom to roughly 3.5 gigawatts of Google TPU capacity coming online in 2027¹³.

A week later, Bloomberg reported Anthropic is fielding unsolicited investor offers at valuations of about $800 billion or higher — more than double the $380 billion post-money valuation it closed in February⁴. The company has, so far, not accepted. Anthropic is in early talks with Goldman Sachs, JPMorgan, and Morgan Stanley about a potential IPO that could come as soon as October 2026⁵.

What You'll Learn

• How Anthropic's ARR grew from $1 billion in January 2025 to $30 billion in April 2026
• Why Claude Code is the single biggest driver of Anthropic's enterprise revenue
• The 3.5-gigawatt Google TPU deal and what it means for Claude's training pipeline
• How Anthropic's $30B ARR compares to OpenAI's $25B run rate
• The $800 billion valuation offers Anthropic has so far resisted
• Context on the IPO talks and why October 2026 is the date to watch

The Headline Number

On April 7, 2026, Anthropic announced that its annualized revenue run rate had surpassed $30 billion, driven by what CFO Krishna Rao described as "exponential growth" in the company's customer base¹. Bloomberg and The Register confirmed the figure on the same day⁶.

For context on the pace of growth:

That trajectory takes Anthropic from ARR parity with a mid-sized enterprise SaaS company to eclipsing the headline revenue of its largest rival in 15 months. The Register reported that the run rate more than tripled in roughly four months from the end of 2025⁶.

Passing OpenAI on Revenue — for the First Time

OpenAI's most recent disclosed ARR figure is approximately $25 billion. The Information reported on the $25 billion milestone in February 2026², and OpenAI's March 31, 2026 funding announcement did not revise that number upward publicly.

That $5 billion revenue gap — in Anthropic's favor — is the first time Anthropic has led this comparison since either company began reporting ARR. OpenAI closed a record-breaking $122 billion funding round on March 31, 2026 at an $852 billion post-money valuation, with Amazon committing up to $50 billion, Nvidia contributing $30 billion, and SoftBank contributing another $30 billion¹⁰. Even with that capital raise, OpenAI's revenue number sits below Anthropic's as of early April.

Two important caveats:

1. Run-rate revenue is a snapshot, not a full-year figure. ARR extrapolates the most recent month's revenue to twelve months. A spike in enterprise contract signings can move ARR faster than actual full-year revenue.
2. OpenAI has not updated its ARR publicly since February 2026. If OpenAI's ARR has moved upward since, the $5 billion gap could narrow or close.

Still, the direction matters. Anthropic grew its run rate from $9 billion to $30 billion in roughly four months while OpenAI's pace of disclosed growth has slowed from the sequential jumps it reported through 2025.

Claude Code: The Runaway Product

A major driver behind the acceleration is Claude Code, Anthropic's agentic coding product. Claude Code became generally available in May 2025. By November 2025 — six months after general availability — it had reached $1 billion in annualized run-rate revenue, a pace faster than any enterprise software product in Anthropic's cited comparisons¹¹.

As of the Series G fundraise disclosure in February 2026, Claude Code's run-rate revenue was $2.5 billion, with Anthropic noting that Claude Code's figure "more than doubled since the start of 2026"⁹. The product is now used by engineering teams at Netflix, Spotify, KPMG, L'Oréal, and Salesforce, among others¹¹.

Claude Code's growth explains a meaningful share of how a company whose flagship product is a chat interface managed to outgrow a rival whose flagship product has 900 million weekly users. Claude Code's economics are enterprise-first: multi-year contracts, per-seat pricing at developer-tooling rates, and high stickiness once integrated into a team's workflow. That profile converts revenue growth into ARR faster than consumer-driven usage does.

If you want a longer read on Claude Code's trajectory and Opus 4.7's benchmarks, see our Claude Opus 4.7 deep dive.

The Enterprise Customer Explosion

Alongside the $30 billion ARR figure, Anthropic disclosed that its count of customers spending $1 million or more per year on Claude had doubled from 500 to over 1,000 in less than two months since the Series G close in February¹.

To put the number in perspective:

A 40x+ increase over 24 months in seven-figure enterprise accounts is unusual even for the current AI market. It reflects two patterns Anthropic has emphasized: deep adoption within existing accounts (customers starting at smaller contract sizes and expanding), and new logo wins at banks, insurers, healthcare systems, and legal services firms that historically take 18-24 months to onboard any new AI vendor.

The Series G materials also noted that eight Fortune 10 companies are Claude customers⁹ — companies whose procurement cycles alone typically mean their decision to license was made months or quarters before the contract closed.

The 3.5-Gigawatt Google TPU Deal

Bundled with the ARR announcement was the disclosure of an expanded compute partnership with Google and Broadcom. The headline figure: approximately 3.5 gigawatts of Google TPU capacity will come online for Anthropic starting in 2027, supplied through Broadcom¹³.

The deal expands on an earlier agreement struck in October 2025 for more than one gigawatt of TPU capacity³. The new capacity roughly triples the prior commitment. Tom's Hardware and CNBC confirmed that the majority of new infrastructure will be located in the United States and that it extends the $50 billion American AI infrastructure commitment Anthropic made in November 2025³¹².

Why TPUs and not GPUs? Anthropic is unusual among frontier labs in that it trains across multiple hardware vendors. Its production stack spans Google TPUs, Amazon Trainium, and Nvidia GPUs, with Claude models trained using a combination of all three. The expanded Google deal reflects TPU pricing and availability relative to H200-class GPU scarcity, and it tightens an already-deep Google-Anthropic alignment — Google has invested more than $3 billion in Anthropic directly (reportedly holding a ~14% stake) and is a Claude distribution partner via Vertex AI.

Broadcom's statement noted that Anthropic's consumption of the new TPU capacity is contingent on "continued commercial success" — a standard clause in this size of compute contract, and one that signals Broadcom shares some supply-chain risk if growth decelerates³.

For context on the chip-supply dynamics at this scale, our AI costs breakdown walks through the math on training-run economics and how hyperscaler deals reshape frontier lab unit economics.

The $800 Billion Valuation Offers

Seven days after the ARR announcement, Bloomberg reported that Anthropic had received multiple unsolicited offers from venture investors to fund a new round at valuations of roughly $800 billion or higher⁴. TechCrunch's April 15 follow-up confirmed the details and noted that Anthropic was "shrugging off" the offers, though it had not ruled out raising additional capital¹³.

Some math on the valuation:

An $800 billion valuation would more than double the Series G mark in roughly two months. It would also close most of the valuation gap with OpenAI ($852 billion post its March 31 close). The ARR multiples differ sharply though: at $852 billion and $25 billion ARR, OpenAI trades at roughly 34x. At $800 billion and $30 billion ARR, Anthropic would trade at roughly 27x — a lower multiple despite its revenue lead.

The rationale for holding out, per reporting by TechCrunch and Bloomberg: Anthropic has ample cash from the $30 billion Series G it closed in February and prefers to move to an IPO rather than extend the private-round cadence¹³.

The IPO Path

Anthropic is in early talks with Goldman Sachs, JPMorgan, and Morgan Stanley about a potential IPO that could come as soon as October 2026, with a target raise of more than $60 billion⁵. At that size, it would exceed Saudi Aramco's 2019 listing — which raised $29.4 billion after its greenshoe option and still stands as the largest IPO in history¹⁴ — by roughly 2x, and would be the single largest IPO ever by a wide margin if the $60 billion target holds.

An IPO at $800 billion-plus would:

• Create one of the top-five most valuable public companies at listing, depending on market conditions
• Give existing private-round investors an orderly liquidity path (vs. the secondary-tender markets that have dominated AI private-company liquidity so far)
• Subject Anthropic to quarterly disclosure requirements, which could meaningfully change what the AI industry knows about unit economics at scale

Two things to watch:

1. Regulatory scrutiny. Anthropic is simultaneously litigating a Pentagon blacklisting dispute. A DC Circuit panel denied Anthropic's bid for a temporary injunction on April 8, 2026, meaning the supply-chain-risk designation remains in effect while the case plays out¹⁵. Government procurement exposure is a risk factor IPO investors will scrutinize.
2. OpenAI's response. OpenAI is reportedly targeting its own IPO in 2026 or 2027 at a ~$1 trillion valuation¹⁶. If Anthropic's IPO lands first, at a premium ARR multiple, it could compress OpenAI's pricing expectations.

How Anthropic Is Training 4x Cheaper

One reason Anthropic's unit economics support a large valuation is reported training efficiency. Per SaaStr's April 2026 summary, Anthropic spends roughly four times less on model training than OpenAI does for comparable frontier capability¹⁷. Several factors contribute:

• Multi-vendor hardware strategy. Training across TPUs, Trainium, and GPUs avoids supply-constrained pricing on any single chip class.
• Constitutional AI and reinforcement-learning-from-AI-feedback pipelines. Anthropic's alignment methods use fewer human annotators than OpenAI's RLHF-heavy approach.
• Concentrated engineering team. Anthropic's model-training team is measurably smaller than OpenAI's, producing a lower overhead component per training run.

None of this is independently audited — the ratios come from Anthropic's disclosures and industry reporting rather than a third-party analysis. But the cash-flow implications are real: at a given ARR, a lab that spends less on training generates more free cash flow, which translates into a defensible premium multiple.

For a deeper look at the chip-supply side of this story, our Cerebras IPO and wafer-scale chip post walks through how compute partnerships shape AI lab economics.

What This Changes in the Market

Enterprise AI buyers. Procurement committees that have been waiting to see which frontier lab "wins" now have a revenue-ranked answer. Expect Claude Enterprise deal sizes to climb, and expect OpenAI's enterprise team to respond with pricing and integration incentives.

Open-source labs. The gap between the top-two closed-source labs and the next tier — including open-weight efforts like Z.ai's GLM-5.1 and Meta's Muse Spark — is a function of capital access as much as model quality. Our GLM-5.1 analysis shows an open-weight model briefly leading on benchmarks while still trailing on customer scale.

Infrastructure vendors. Broadcom, Google Cloud, AWS, Microsoft Azure, and CoreWeave all benefit from the Anthropic-OpenAI-Meta capital arms race. The 3.5-gigawatt TPU commitment alone exceeds the total installed TPU capacity of most sovereign AI programs.

Regulators. A private company earning a $30 billion run rate and carrying an $800 billion implied valuation without public-company disclosure obligations is the exact profile that accelerates political pressure for AI transparency rules. Expect IPO timing to partially reflect this — Anthropic may prefer to file S-1s rather than have disclosure obligations imposed by legislation.

The Bottom Line

Anthropic crossed $30 billion in annualized revenue in early April 2026, overtaking OpenAI's last-disclosed $25 billion. The milestone was accompanied by a tripling of its compute footprint (via the 3.5-gigawatt Google TPU deal) and a doubling of its largest enterprise accounts in under two months.

The follow-up news — unsolicited $800 billion valuation offers and IPO talks for a potential October listing — is what typically happens when a private company's revenue trajectory outruns its current price. Whether Anthropic holds the revenue lead for the rest of 2026 is the open question. For now, though, the AI industry has a new revenue leader, and the gap between Anthropic's enterprise-first motion and OpenAI's consumer-first motion is delivering measurably different results on the financial side.

References