Anthropic (creator of Claude) has overtaken OpenAI (creator of ChatGPT) on reported revenue, with an annualised run rate of around $47 billion versus $25–33 billion for its rival. The most interesting part isn't the figure, but what it reveals: two radically opposed visions of how to build an AI company. One bets on scale and integration with power, the other on trust and independence.
In the AI race, OpenAI has long been seen as the undisputed leader — the one that popularised ChatGPT and defined the category. Yet by mid-2026, a quiet shift has occurred: Anthropic, long perceived as the cautious challenger, has overtaken OpenAI on revenue. Beyond the numbers, two philosophies of power are colliding. Let's break them down.
The shift in numbers
Anthropic is reported to have reached an annualised revenue run rate of around $47 billion, surpassing the $25–33 billion range attributed to OpenAI, as well as on enterprise subscriptions. At the same time, ChatGPT's monthly visits fell below a majority share of the generative AI market for the first time in May 2026 — a shift we analysed in our article on market share.
These figures are self-reported by the companies and should be taken with caution, but the trend is clear: Anthropic monetises better, particularly with enterprises, while OpenAI still dominates the harder-to-monetise consumer volume. The difference in model is striking and worth examining.
Two strategies toward US power
This is where things get truly interesting, because the two companies now embody opposing ways of navigating their relationship with the state.
OpenAI is betting on maximum integration. The company has offered to cede 5% of its equity to the US government — an extraordinary offer signalling that it sees its relationship with the state as existential. It has signed a deal with the Pentagon, voluntarily adopted a government coordination framework for the release of GPT-5.6, and publicly argues that its models should become the infrastructure of a US-led global economy. The bet: tying its commercial success to American strategic interests.
Anthropic is betting on independence and trust. The company has built its growth on enterprise trust, foregrounding its safety principles and regulatory cooperation rather than state equity participation. This stance has already put it in a standoff with the Pentagon, which we covered in our article on that conflict, where Anthropic refused certain military uses on principle. The bet: making reliability and ethics a commercial advantage.
Anthropic's success with enterprises highlights a counter-intuitive point: in a professional market, caution and predictability are worth their weight in gold. A company entrusting sensitive data to an AI wants a reliable, transparent partner that won't spring nasty surprises. Anthropic's reputation for seriousness — sometimes mocked as excessive prudence — has turned out to be exactly what the enterprise market was looking for. Trust isn't just a moral value; it's a commercial asset.
The IPO context
This shift comes at a pivotal moment. Anthropic is preparing its IPO for autumn 2026, and these solid revenue figures are a strong argument for investors. OpenAI's proposal to cede 5% to the state also raises a delicate governance question: can a regulator that holds a stake in the company it regulates really enforce rules impartially? Several observers have flagged this potential conflict.
Both approaches carry risks. OpenAI's integration with US power gives it resources and protection, but exposes it to becoming a political instrument and losing independence. Anthropic's independence earns it market trust, but has already put it in direct conflict with as powerful a client as the Pentagon. Neither is without danger.
What to take away
Anthropic overtaking OpenAI is more than a ranking anecdote. It validates, at least for now, the idea that you can succeed in AI by means other than scale and the race for size: through trust, reliability, and a measured relationship with power. Two models now coexist at the top, and the industry is watching which one proves more durable.
For the observer, it's a lesson that extends beyond AI: in any nascent industry, there is more than one way to win. The flashiest strategy isn't always the most solid. Sometimes it's discretion, consistency, and patiently built trust that ultimately prevail. The story isn't written yet, but as of mid-2026, it's the bet on discretion that's leading the race.