AI market share snapshot
Social analysis shows OpenAI’s enterprise share shrinking from about 50% in 2023 to roughly 27% today, while Anthropic, Google and xAI have gained ground and hyperscalers now control about two‑thirds of AI compute capacity ( ). The same set of posts also notes AI‑native startups hitting rapid revenue milestones — some reaching $100M ARR quickly — and models approaching human performance on benchmark tasks ( ).
OpenAI’s early lead in business artificial intelligence has narrowed as companies spread spending across Anthropic, Google and other model vendors. (menlovc.com) Menlo Ventures said in its December 2025 enterprise survey that OpenAI’s share of enterprise model usage fell to 27% from 50% in 2023. The same report put Anthropic at 40% and Google at 21%, with the top three accounting for 88% of enterprise model usage. (menlovc.com) An earlier Menlo market update published on July 31, 2025 showed a slightly different cut of the same shift: Anthropic at 32%, OpenAI at 25% and Google at 20% in enterprise usage by mid-2025. Menlo said that update was based on a survey of 150 technical leaders across artificial intelligence startups and large enterprises. (menlovc.com) The market these companies are chasing is getting larger at the same time. Menlo said enterprise spending on large language models rose from $3.5 billion in late 2024 to $8.4 billion by mid-2025, then said total enterprise generative artificial intelligence investment reached $37 billion in 2025. (markets.financialcontent.com, menlovc.com) The infrastructure behind those models is concentrating too. Epoch AI said Amazon, Google, Meta, Microsoft and Oracle owned 67% of global artificial intelligence compute capacity in the fourth quarter of 2025, up from 60% in the first quarter of 2024. (epoch.ai) That concentration is showing up in supplier deals. Anthropic said on April 6, 2026 that it had signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation tensor processing unit capacity starting in 2027. (anthropic.com) The startup layer on top of those models is moving fast enough to reset software revenue benchmarks. Sacra reported that Cursor went from $1 million in annual recurring revenue to $100 million in 12 months, reaching that mark by early 2025. (sacra.com) Benchmark scores are also climbing, though they do not measure the same thing as real-world job performance. Epoch AI said its database tracks results across coding, reasoning and other tests, while LM Council’s April 2026 benchmark table showed leading systems scoring above 70% on Humanity’s Last Exam and posting triple-digit minute time horizons on software task evaluations. (epoch.ai, lmcouncil.ai) OpenAI is still pushing deeper into business customers as the field tightens. The Washington Post reported on April 16, 2026 that OpenAI executives said the company would introduce a new model aimed at “high-value professional work” as competition with Anthropic intensified. (washingtonpost.com)
