Peter Walker, Head of Insights at OpenRouter, posted OpenRouter’s chart of Anthropic and OpenAI shares of tokens and spend on 31 October, writing that “Opus 5.5 reversed the trend”. The piece takes the most recent week’s shares from the chart.
Short explainer · Anthropic draws 2.1x OpenAI’s spend per token
A vertical news Short summarising this story. Figures come from OpenRouter’s chart posted by its Head of Insights; our arithmetic is labelled.
Anthropic took 48% of the tokens but 66% of the spend between the two labs on OpenRouter in the most recent week of OpenRouter’s data, while OpenAI took 52% of the tokens and 34% of the spend. That puts Anthropic’s spend per token at about 2.1 times OpenAI’s, by our arithmetic.
The figures come from an OpenRouter-branded chart that Peter Walker, OpenRouter’s Head of Insights, posted on X on 31 October. The chart also says total tokens rose 224 times and total spend 81 times since the first week of 2026.
The key insight: As we read OpenRouter’s chart, Anthropic and OpenAI split the most recent week’s tokens close to evenly, but Anthropic took about two-thirds of the spend.
What OpenRouter’s Chart Shows
The chart, titled “Anthropic vs OpenAI so far in 2026”, plots each lab’s weekly “share of tokens vs share of spend” across OpenRouter users, from January to late September. Its footer credits “OpenRouter token & spend data” and openrouter.ai/data.
In the most recent week, OpenAI’s spend share was 18 points below its token share, and Anthropic’s was 18 points above, by our arithmetic. Anthropic took about 1.4 times its token share in spend; OpenAI took about 0.65 times.
Walker wrote that OpenAI “got back in the game with steady gains in spend share throughout the year”, that tokens “go kinda wild with the introduction of Luna + new Haiku”, and that “Opus 5.5 reversed the trend”.
Anthropic and OpenAI shares of tokens and of spend on OpenRouter in the most recent week of OpenRouter’s chart (two-lab shares). Spend-per-token ratios are our arithmetic.
Volume Up, Price Down
The chart’s totals rose 224 times in tokens and 81 times in spend since the first week of 2026. Spend grew about 36% as fast as tokens, so spend per token on those totals fell about 64%, by our arithmetic.
Two labs with close to equal token volume took very different spend. A lab’s revenue on a router depends on how many tokens it serves and on what each one costs.
The framework says the token “is simultaneously the unit of cognition the model produces, the unit of compute the data center serves, the unit of price the lab charges, and the unit of value the enterprise extracts.”
As we read OpenRouter’s chart, close to equal token shares carried very different spend shares for the two labs. The framework is the owner’s argument about token economics and does not discuss OpenRouter.
OpenRouter is one router, so its shares describe its own users, not the whole market. The chart shows Anthropic and OpenAI only, so the shares are of the two labs combined. It does not give token or spend totals in absolute terms, and it does not say how much of the spend reaches each lab after OpenRouter’s fees.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
In OpenRouter’s most recent week, Anthropic took 48% of the two labs’ tokens and 66% of their spend, about 2.1 times OpenAI’s spend per token, by our arithmetic. Since the first week of 2026, the chart’s tokens rose 224 times and its spend 81 times.
A note on sourcing. Shares and growth multiples come from an OpenRouter-branded chart posted on X by Peter Walker, OpenRouter’s Head of Insights, on 31 October 2026; the figures are as labelled on the chart. Ratios are our arithmetic. We did not contact OpenRouter, Anthropic or OpenAI.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
