OpenAI vs Anthropic Revenue Dispute: $74B Gross ARR vs $41B, and Which Number to Trust

Anthropic's tracked annualized revenue hit $74.1 billion in July 2026, 1.79x OpenAI's $41.3 billion — but OpenAI says that gap is inflated by roughly $8 billion in cloud-accounting differences. That's the short answer. The longer answer is more interesting.

Two companies filed confidential S-1s seven days apart in June 2026, and institutional investors evaluating both are now stuck comparing revenue figures built on two different accounting methodologies. Anthropic books its Claude sales through AWS and Google Cloud on a gross basis; OpenAI books its ChatGPT and API sales through Microsoft on a net basis. Both are GAAP-legal. Neither is directly comparable without adjustment, and OpenAI's own CRO put a specific number on the distortion in a leaked internal memo.

The OpenAI-Anthropic Revenue Accounting Dispute, Explained

The OpenAI-Anthropic revenue accounting dispute centers on one question: should a cloud partner's commission count as your revenue? Anthropic books gross sales through AWS and Google Cloud, including the cut those partners keep; OpenAI books net sales through Microsoft, excluding Microsoft's cut. Both are legal under GAAP, but they produce headline ARR figures that overstate or understate the underlying business by double-digit percentages, and the gap has become central to how investors price both companies' pending IPOs.

The dispute became public on April 13, 2026, when OpenAI's chief revenue officer Denise Dresser sent a four-page internal memo to staff — later reported by Forbes and BigGo Finance — arguing that Anthropic's then-$30 billion annualized run rate was overstated by approximately $8 billion because of how Anthropic accounts for its AWS and Google Cloud revenue-sharing deals. Using OpenAI's preferred net methodology at the time, Anthropic's comparable figure would have been closer to $22 billion — which, notably, would have placed it below OpenAI's own $25 billion run rate rather than above it. The framing war has only intensified as both companies approach a public listing, since a $32.8 billion headline gap and an adjusted ~$0 gap tell two completely different stories to the same investor.

Neither company has fully settled the argument in public. Anthropic has not issued a line-by-line rebuttal of the $8 billion figure, and it has continued to report ARR on a gross basis in the months since, including in briefings around its May 28, 2026 Series H. OpenAI, for its part, has kept reporting net figures even as its own consumer-advertising pilot — which reportedly hit $100 million in ARR within six weeks — adds a new revenue line that could eventually invite its own gross-vs-net questions if OpenAI ever monetizes ad inventory through a third-party ad-tech partner rather than directly.

ARR by the Numbers: OpenAI vs Anthropic, July 2026

The table below lays out both companies' tracked revenue trajectory across 2026, plus the structural differences behind the numbers. All figures are third-party estimates; neither company has published audited public financials as of this writing.

Figures are 2026 estimates blended from TickerTrends' ARR tracker, Forbes, BigGo Finance, Bloomberg funding-round coverage, and company announcements, checked July 2026. Neither company has published audited financials; tracker figures are unofficial and subject to revision.

Gross vs Net Revenue Accounting: Where the $8 Billion Gap Comes From

Gross revenue accounting books the full amount a customer pays, even the slice a distribution partner later collects as commission. Net revenue accounting books only what the company actually keeps. Anthropic sells a meaningful share of Claude access through AWS Bedrock and Google Cloud's Vertex AI, and reports those sales gross — the full customer payment, including AWS's and Google's cut. OpenAI's ChatGPT and API revenue flows partly through its commercial arrangement with Microsoft, and OpenAI reports that revenue net of Microsoft's share. Neither treatment violates GAAP; large SaaS and cloud-marketplace businesses routinely choose one or the other depending on how much control they retain over pricing and the end customer relationship.

The practical effect: a business that is genuinely comparable in size to a net-basis competitor can post a headline ARR figure 20-30% higher purely from the accounting choice. In April 2026, OpenAI's Dresser memo argued this exact effect explained roughly $8 billion of Anthropic's then-reported $30 billion run rate — implying a real, net-comparable figure closer to $22 billion, below OpenAI's own $25 billion at the time. Anthropic has not issued a detailed public rebuttal of that specific $8 billion figure, and independent analysts have generally treated both the $8 billion adjustment and Anthropic's gross headline number as directionally informative rather than precise.

This isn't a novel problem in tech accounting — it's the same gross-vs-net question that dogged early cloud marketplaces and ride-sharing platforms a decade earlier. Uber, for years, disclosed both "gross bookings" and net revenue precisely because the two numbers told wildly different growth stories, and travel-booking sites like Expedia and Booking.com still report both merchant (net) and agency (gross) revenue separately for the same reason. What's unusual here is the stakes: two companies simultaneously worth a combined $1.8 trillion in private markets, both heading toward an IPO within months of each other, are using incompatible conventions for the single metric — annualized revenue — that public investors will lean on hardest to justify either valuation.

Revenue Mix: Why Anthropic's Enterprise Weighting Matters Beyond the Accounting Fight

Even setting the gross-vs-net dispute aside, the two companies' underlying businesses look different. OpenAI's revenue is still roughly 70% ChatGPT consumer subscriptions (Plus, Pro, Team, Business) across an estimated 20 million paid seats, with the remaining 30% split between API/developer usage and newer bets like Sora and a nascent advertising product. Anthropic's mix runs closer to 80% enterprise and API — Claude sold directly to companies for coding, customer support, and internal tooling — with consumer Claude.ai still a minority of revenue. Enterprise and API revenue is typically stickier and carries fewer refund/chargeback dynamics than consumer subscription revenue, which is part of why investors have been willing to price Anthropic's growth at a premium multiple even before resolving the accounting dispute.

The mix gap also explains why the two companies talk about growth so differently in investor conversations. Anthropic's enterprise contracts tend to be multi-year and usage-based, so a handful of large coding-tool and customer-support deployments can move the run rate by billions in a single quarter — which is part of why its ARR jumped from roughly $35.6 billion in April to $69.6 billion by June, a near-doubling in two months. OpenAI's consumer base is stickier in aggregate but grows more linearly, seat by seat, which is one reason its own six-month growth multiple of roughly 1.9x looks pedestrian next to Anthropic's 7.3x even before anyone touches the accounting question.

The IPO Stakes Behind the OpenAI-Anthropic Revenue Dispute

This isn't an academic disagreement — it's happening while both companies are on file. Anthropic filed a confidential S-1 on June 1, 2026, and OpenAI followed on June 8, seven days apart, meaning underwriters and institutional investors are being asked to price two competing IPOs off revenue figures built on incompatible accounting bases. A public listing forces audited GAAP financials, at which point the gross-vs-net distinction becomes visible line by line rather than argued over in a press-reported internal memo. Whichever company's true, apples-to-apples growth rate holds up better under that scrutiny stands to command the richer IPO multiple — and right now the market is pricing Anthropic slightly ahead, at a $965 billion valuation versus OpenAI's $852 billion, even with the accounting dispute unresolved.

For investors tracking exposure to either company through funds, secondaries, or vehicles like the Robinhood Ventures Fund, the accounting dispute is a live NAV risk: a headline ARR number that later gets restated on a comparable basis at IPO could move the implied multiple significantly in either direction. You can track how private AI company valuations are moving more broadly on our AI Valuations dashboard, and compare these multiples against public mega-cap tech names on our Big Tech Earnings dashboard.

Which Number Should Investors Actually Trust?

Neither headline number in isolation. Anthropic's $74.1 billion gross figure overstates the business relative to a net-basis peer by some real amount — OpenAI's own estimate, from April, was roughly $8 billion on a $30 billion base, or about 27%. Applying that same rough haircut to July's $74.1 billion would put a rough net-comparable Anthropic figure somewhere in the $50-55 billion range — still comfortably ahead of OpenAI's $41.3 billion, and still growing dramatically faster on a like-for-like basis. The honest read is that Anthropic is winning the growth race even after adjusting for accounting, but the specific multiple of that win (1.2x adjusted versus 1.8x headline) is genuinely uncertain until both companies publish audited, comparable financials as part of their IPO process.

The more durable signal isn't the accounting fight at all — it's growth rate and revenue quality. Anthropic's roughly 7.3x six-month growth and 80% enterprise mix would be remarkable under either accounting method; OpenAI's 1.9x growth and 70% consumer mix would look modest under either method too. Investors evaluating either company's pending IPO should weight the growth trajectory and customer mix over the single headline ARR figure, because the accounting dispute guarantees that number will keep moving even after both S-1s go effective.