Market data: OpenAI and Anthropic account for ~35% of enterprise AI deployments
- Menlo Ventures’ latest enterprise AI data shows a tight two-company market, with Anthropic leading enterprise LLM spend and OpenAI still dominating production penetration. (menlovc.com) - The clearest number is concentration: Anthropic held 40% of enterprise LLM spend in Menlo’s 2025 report, while 78% of CIOs told a16z they use OpenAI in production. (menlovc.com) - That matters because the same vendors now control model access, cloud capacity, and pricing leverage as enterprise AI budgets keep exploding. (menlovc.com)
Enterprise AI is starting to look less like an open market and more like a two-horse race. But the tricky part is that different datasets are measuring different things. Menlo Venture(menlovc.com)nAI at 27% and Google at 21%. A separate a16z CIO survey says OpenAI is still the most widely used in production, with 78% of enterprise CIOs using its models, while Anthropic has made the fastest gains. (menlovc.com) ### So who is actually ahead? Depends on the scoreboard. If you care about dollars flowing to model vendor(menlovc.com) somewhere in production, OpenAI is still the incumbent. Those are not contradictory. One company can be installed more widely, while another captures more of the newest or highest-value workloads. (menlovc.com) ### Why are the numbers so messy? Because “enterprise AI deployments” is not one clean metric. Some reports track spend. Some track usage share. Some track CIO adoption. And some focus only on found(menlovc.com)%, Anthropic 31%” should be treated carefully unless the underlying methodology is visible. The strongest public numbers I could verify point instead to Anthropic at 40% of enterprise LLM spend in late 2025, with OpenAI leading on production footprint. (menlovc.com) ### Why is Anthropic gaining so fast? Coding is a h(menlovc.com)illion in 2024, and a lot of the action sits in practical software use cases where model quality shows up immediately. Anthropic has been especially strong in coding and agent-style tasks, which are exactly the categories that spread fast inside engineering teams. (menlovc.com) ### Why does infrastructure matter so much? Because model competition is no longer just about benchmarks. It is about who can actually get enough chips and cloud to serve demand. Anthrop(menlovc.com)e end of 2025, and that it had signed for multiple gigawatts of next-generation TPU capacity with Google and Broadcom starting in 2027. That is not a normal vendor contract — it is industrial-scale supply locking. (anthropic.com) ### And Amazon’s role? Also enormous. On April 20, Amazon said it would invest up to another $25 billion in Anthropi(menlovc.com)han $100 billion on AWS over 10 years and to keep scaling on Trainium chips. So Anthropic is not just selling models into the enterprise. It is deeply tied to the cloud companies that host and finance the race. (cnbc.com) ### What does this change for buyers? Procurement gets more strategic. Enterprises are no longer just comparing model quality and price per token. They are asking harde(anthropic.com)ty rules, or gets pulled tighter into one cloud ecosystem? Multi-cloud availability helps, and Anthropic is leaning hard on that point, but concentration risk is still rising. (anthropic.com) ### Is this a bubble or real demand? The demand looks real. Menlo’s estimate of enterprise generative AI spend jumped to $37 billion in 2025, and more than half went(cnbc.com)omeday. The catch is that real demand can still produce a concentrated market with ugly lock-in dynamics. (menlovc.com) The bottom line is simple. Enterprise AI is consolidating around a few frontier vendors, and the contest is no longer just OpenAI versus Anthropic on model quality. It is also a fight over cloud pipes, chip access, contract leverage, and who gets to become the default layer inside big companies. (menlovc.com)
