The AI model space is largely a duopoly, at least per Ramp’s data.
OpenAI and Anthropic together account for more than 95% of the token spend on Ramp. OpenAI has edged past Anthropic for the top spot. In third place is xAI, followed by Google at fourth place.
OpenAI is #1 with $13.7M in tracked spend, a 48.9% share. Anthropic drops to #2 at $13.1M and 46.8% — a gap of less than $600K on a combined $26.8M pool, or roughly two percentage points. This lines up with what’s been happening at the model level: OpenAI’s Astra has been outspending Anthropic’s Fable in raw enterprise dollars since just a couple of weeks after Astra’s launch.
After the top two, the drop-off is almost comedic. xAI holds a distant #3 with $633.9K (2.3%), and Google — despite owning one of the largest AI research organizations on the planet — sits at #4 with just $429.5K (1.5%). That’s a strikingly small number for a company with Gemini’s scale and distribution, though it tracks with Google’s broader slide down the AI lab rankings this year, where it’s fallen behind a swarm of smaller, faster-moving labs on pure model quality even as it keeps shipping.
Below that, the numbers stop being meaningful in any economic sense. Cursor — not a foundation lab at all, but a coding tool whose in-house models apparently get enough real use to register — takes #5 with $58K (0.2%). Moonshot, the maker of Kimi, follows at #6 with $48.7K. Z.ai clears #7 with $6.4K, ahead of DeepSeek at #8 with a mere $553.10 — a rounding error for a company that spent much of last year rattling Silicon Valley’s spending assumptions.
The shape of it is really a two-lab story wearing a ten-row table. OpenAI and Anthropic together account for 95.7% of everything Ramp is tracking, and the gap between #2 and #3 is roughly 20x — bigger than the gap between #1 and #2. Everyone else, xAI included, is competing for scraps left over by two companies that are also, not coincidentally, about to become part of the largest wave of AI-related IPOs in history.