OpenRouter users spent more on OpenAI's models than on Anthropic's in the week of September 7, the first time that happened since the week of February 26, 2024
OpenRouter users spent more on OpenAI models than on Anthropic models last week. This hasn't happened for more than 2.5 years
Context & Ripple Effects
OpenRouter’s data had already recorded lower-cost Chinese models overtaking US rivals in token consumption, while separate Ramp tracking showed Anthropic taking about 73% of spending among companies buying AI tools for the first time. Those measures pointed to a fragmented market rather than a settled two-company hierarchy.
Against that backdrop, OpenAI’s return to the top of OpenRouter spending is a meaningful reversal in a usage-and-spend dataset after a long Anthropic lead. It also narrows the contrast with OpenAI’s earlier reported revenue lead over Anthropic.
First-order effects
- OpenAI becomes the higher-spend model provider among OpenRouter users for the week of September 7, ending a run in which Anthropic had held that position since February 2024.
- Anthropic loses its spending lead in a visible model-marketplace measure, even as its strength among first-time corporate AI-tool buyers had been reported separately.
Second-order effects
- OpenAI and Anthropic face sharper pressure to defend spend share on OpenRouter, where users can shift allocations among competing proprietary and lower-cost models.
- For model buyers, the reversal adds another signal that spend allocation is changing across providers rather than following a stable vendor ranking.
Third-order effects
- If repeated, spending shifts on OpenRouter would reinforce model buyer power: providers must compete continuously on the capability-and-cost mix rather than rely on durable default status.
- The coexistence of OpenAI and Anthropic spending changes with substantial Chinese-model token use points toward a more contested inference market with several viable supplier tiers.
The trend: AI-model procurement is becoming more fluid, with buyers reallocating spend among frontier and lower-cost providers as performance and inference economics change.