3 Companies, 67% of AI's Funding, and the End of Cheap Inference
PitchBook says OpenAI, Anthropic, and xAI now take two-thirds of AI funding — and they're doubling API prices. Here's how Mistral, Google, and Alibaba are carving out lanes, and what it means for the tools you bet on.
The data point that should reframe every vendor announcement this week: 3 companies now absorb two-thirds of all AI venture funding, and those same companies just doubled their API prices. The land-grab phase is ending, and what each vendor does next is a better signal than any demo.
According to PitchBook's Q1 2026 AI VC Trends report, OpenAI, Anthropic, and xAI together took 67% of all AI funding for the quarter, leaving $83.5 billion to split across 1,543 other deals, as reported by The AI Insider. That concentration decides who can keep subsidizing inference and who has to charge for it now.
Developer Simon Willison argues April 2026 was the month OpenAI and Anthropic found product-market fit, pointing to roughly 2x API pricing on their newest models arriving alongside enterprise contracts that lock large customers into those rates (@simonw). The fact is that prices went up; the read is that the leaders finally have enough committed demand to trade growth for margin.
The challengers pick lanes
The vendors outside that top 3 are not fighting head-on. Mistral used its AI Now Summit at the Louvre to announce production deployments at Airbus, BMW, and EDF, alongside a coding agent called Vibe (Mistral AI). Alibaba claims its Qwen3.7-Max, released in May 2026, beats Gemini and ChatGPT on coding benchmarks and can run autonomously for up to 35 hours — vendor-reported figures, not independently verified (India Today). Google, rather than chase benchmarks, is staking out provenance: SynthID has now watermarked over 100 billion pieces of content and run 50+ million verifications inside Gemini, with OpenAI and ElevenLabs signing on (Google DeepMind).
What it means for you
For builders, budget for the 2x and assume cheap subsidized inference is over. Developer gravity is real, though — OpenAI's Codex climbed from about 100,000 NPM installs a day in January 2026 to more than 1 million a day now (@simonw), so the tools you standardize on this quarter will be expensive to unwind. For buyers, the one-model-for-everything pitch is fading: Mistral is positioning for regulated industrial work, Google for content authenticity, the frontier labs for raw capability. For anyone shipping media, the SynthID expansion is the quiet structural shift — when "Is this made with AI?" becomes a button in Search and Chrome, provenance stops being a feature and becomes default plumbing.
The bottom line: Treat 2026 as the year AI vendors stopped competing on everything and started competing on something specific — choose yours by the problem you actually have, not by the logo.
Sources
- Given the recent burst of activity around enterprise pricing and contracts, I think April 2026 was t
- We're taking on the hardest problems in the real world 🏗️🚚 🛫⚛️ Today at The AI Now Summit, held at
- Our Gemini for Science tools could help scientists unlock their next breakthrough. 🧬 https://t.co/EN
- @RihardJarc https://t.co/dZTlf1wexY shows @openai/codex installs through NPM were at ~100,000 a day
- New on the Engineering Blog: The access and permissions we grant agents should evolve with their cap
- AI Now Summit 2026 | Mistral AI <strong>Vibe runs on flagship Mistral models optimized for reasonin
- AI coding agents for enterprises | Mistral AI ● Deep project understanding that informs every compl
- Alibaba AI Qwen latest model beats Google Gemini and ChatGPT in coding - India Today Alibaba has la