• Monday Momentum
  • Posts
  • Anthropic Passed OpenAI by Ignoring the Customer Everyone Wanted

Anthropic Passed OpenAI by Ignoring the Customer Everyone Wanted

How the lab with a fraction of OpenAI's users just posted its first profit and took the revenue lead, and why the same week OpenAI started selling ads tells you which bet is working

Author

Justin Wright
August 24, 2026 • Est. Reading Time: 6 minutes

Happy Monday!

Anthropic's Q2 revenue came in at roughly 14 times the $787 million it posted in the same quarter a year earlier. (Source: CNBC)

For three years, the race between OpenAI and Anthropic looked like a rout. OpenAI had the household name, the billion users, the consumer mindshare, and the fundraising gravity of a small nation; Anthropic had a chatbot most of the public could not pick out of a lineup. Then the Q2 numbers landed, and the scoreboard flipped.

Anthropic reported preliminary Q2 revenue above $11.5 billion, roughly 14 times what it earned in the same quarter a year ago and more than double its own first quarter. It posted its first-ever positive operating income and, by July, its annualized run rate had climbed to $65 billion. For the first time, Anthropic surpassed OpenAI in quarterly revenue, and it did so while turning a profit that OpenAI, burning an estimated $14 billion this year, cannot yet approach.

OpenAI announced it will begin showing ads in ChatGPT across 31 European markets. The company built to disrupt Google is now adopting Google's business model. The lab that ignored the consumer everyone wanted just won the round, and the lab that won the consumer is now selling their attention to pay for them.

Anthropic posted Q2 revenue over $11.5 billion, its first operating profit, and a $65 billion annualized run rate, passing OpenAI in quarterly revenue for the first time. Roughly 80% of Anthropic's revenue comes from enterprise and API, led by Claude Code at an $8 billion run rate. OpenAI, with over a billion users, is projected to lose $14 billion this year on inference costs, and just launched ads in 31 European markets to monetize consumers it cannot cover at cost.

TL;DR

The Numbers Behind the Flip

A year ago, Anthropic earned $787 million in the quarter. This past quarter it earned more than $11.5 billion for a roughly 14x increase in twelve months. Its first quarter this year was $4.73 billion, so it more than doubled again in a single quarter. The annualized run rate went from $47 billion in May to $65 billion in July and, unlike almost every other company operating at this scale of growth, Anthropic did it while crossing into positive operating income.

OpenAI's trajectory is different. Its run rate sat near $25 billion from February through the spring, was passed by Anthropic in April, and has since climbed to roughly $40 billion by August. This is growth nonetheless, but not the vertical line Anthropic is drawing, and underneath it sits an estimated $14 billion in losses for the year. OpenAI is larger in users by a wide margin and now comparable in revenue, but it is spending far more than it earns to stay there.

Analysts working on Anthropic's confidential IPO, led by Morgan Stanley, Goldman Sachs, and JPMorgan, are now floating valuations above $2 trillion. Twelve months ago that number would have been absurd, but the Q2 print is what makes it arguable.

Two Bets, Placed Years Ago

The divergence did not happen this quarter. It was chosen years ago, and the companies simply arrived at the consequences of these choices.

Anthropic bet on the enterprise; roughly 80% of its revenue comes from API and business use rather than consumer subscriptions. Its breakout product is Claude Code, which went from a $500 million run rate last September to $8 billion by May, an eightfold climb in eight months. More than 300,000 business customers now account for the overwhelming majority of revenue. Anthropic never seriously contested the consumer chatbot war. It sold the pickaxes, especially to developers, and let someone else run the mining town.

OpenAI bet on the consumer. It built the fastest-growing consumer product in history, over a billion users and more than 50 million paying subscribers. That reach is a genuine asset, the largest audience any AI company has ever assembled, but it is also the source of the bleeding. Every free user costs inference dollars, and OpenAI has deliberately priced its models below the true cost of serving them to hold market share. Inference alone is projected at $14.1 billion this year. When your product is nearly free and your users number in the billions, scale is a bill that comes due.

Why Ads Are the Tell

OpenAI turning to advertising is a confession about the economics of consumer AI. When a company with a billion users decides to sell ads, it is because it cannot make those users pay enough to cover what they cost. Ads appear below ChatGPT responses for Free and Go users, labeled and, in Europe, non-personalized to satisfy GDPR.

OpenAI is monetizing attention because it cannot fully monetize usage, which is precisely the trap that produced the ad-supported web in the first place. The company that promised to reinvent how we find information is now inserting sponsored content into the answers, exactly as the search engine it aimed to replace has done for two decades.

Contrast that with how Anthropic makes its next dollar. It doesn’t need a billion people to see an ad. It needs a few hundred thousand businesses to run more Claude Code, sign larger API contracts, and deploy Claude deeper into their workflows. One model monetizes eyeballs while the other monetizes work. This quarter, monetizing work turned a profit and monetizing eyeballs required a new ad business to stem a $14 billion loss.

What This Means for Practitioners

For enterprise buyers, Anthropic's profitability is a signal about durability. A vendor that funds itself from the product you actually pay for is more stable than one subsidizing your usage with investor money and, now, ad revenue. That does not mean OpenAI is going anywhere, its balance sheet is enormous, but it does mean Anthropic's pricing is less likely to lurch as the economics catch up. Factor financial model, not just model quality, into long-term platform decisions.

For founders, the lesson is about which customer to chase. The consumer AI land grab produced the most famous product in tech and the largest losses. The enterprise path produced less fame and a profit. If you are building on AI, the boring buyer who pays for outcomes may be worth more than the viral user who costs you money on every request. Anthropic just demonstrated that at a $65 billion scale.

For anyone watching the IPOs, this quarter reframes both. Anthropic goes to market with a profit and a vertical growth curve. OpenAI goes with unmatched reach and an unresolved question about whether that reach can ever pay for itself. The market will price those stories very differently, and the ad launch is OpenAI trying to change its story before it has to tell it to public shareholders.

The Bottom Line

The obvious reading is that Anthropic won and OpenAI lost, but that reading is too simple. OpenAI still has the largest audience in AI, a run rate climbing toward $40 billion, enterprise revenue that just passed its consumer business, and a distribution advantage Anthropic cannot buy. A billion users is a foundation, and advertising against it could become one of the largest ad businesses in the world. The bet is not dead, it is just expensive.

Anthropic reached the revenue lead and the first profit by doing the unglamorous thing: selling to businesses who pay for value, and skipping the consumer war that made OpenAI famous. In an industry that spent three years measuring itself in users, the company that optimized for customers just moved ahead. The scoreboard everyone was watching turned out to be the wrong one.

In motion,
Justin Wright

If the company with a billion users has to sell ads to survive while the company with a few hundred thousand business customers turns a profit, is consumer scale actually an asset in AI, or is it a liability disguised as a moat until someone figures out how to make free users pay for themselves?

Food for Thought

Quick Hits

  • Etched, founded by three Harvard dropouts, doubled its valuation to $21 billion in under a month, raised $700 million led by Jane Street, shipped its first inference chips, and booked over $1 billion in contracts. (TechCrunch)

  • New Pew data shows 52% of Americans are now more concerned than excited about AI, up from 37% in 2021, with under-30s turning negative for the first time (55%) and 71% expecting AI to cut jobs. (Pew Research)

  • OpenAI's ChatGPT ads launch across 31 European markets on August 24, appearing below responses for Free and Go users, labeled and non-personalized to comply with GDPR. (OpenAI)

  • Google expanded its custom-chip partnership with Marvell, issuing warrants that could reach roughly $12.18 billion if fully exercised, as the custom silicon race against Nvidia accelerates. (Tech Startups)

If you haven’t listened to my podcast Mostly Humans: An AI and business podcast for everyone yet, new episodes drop every week!

Episodes can be found below - please like, subscribe, and comment!