27 rounds in one quarter
Crunchbase counted 27 companies that raised $1 billion or more in a single round in the third quarter of 2026. In the second quarter there were 16 such companies, and in the first quarter 14.
The number of billion-dollar rounds grows every quarter of 2026. Yet the money does not spread across the market: about a third of all world venture money for the quarter went to these 27 companies.
Billion-dollar rounds go to companies of different kinds: model developers, data center builders and data platforms. They have one thing in common: they need money for compute and hardware, which cost more than any team.
The first half of the year shows the same picture: of the $407 billion AI startups raised, about $217 billion went to two companies. The market is ready to give large sums, but to a narrow circle.
The overall numbers of the quarter, including the AI share of all venture money, are in the results of the third quarter of 2026.
Who raised the money
The largest rounds of the third quarter: Databricks and Safe Superintelligence got $5 billion each. Crusoe, Moonshot AI, Mistral AI, Nscale and Kling AI raised more than $3 billion each, and Crunchbase did not disclose the exact amounts.
1. Model developers: Safe Superintelligence from the US, Mistral AI from France and Moonshot AI from China. Their money goes into model training and compute.
2. Infrastructure: Crusoe and Nscale build data centers and rent out compute capacity for AI. This is a bet that demand for compute will keep growing.
3. Data and apps: Databricks works with company data, and Kling AI makes models that create video.
For most of these rounds, the source did not disclose a valuation. So the round size does not tell you what the company is worth or what stake the investors got.
Two of the five companies with rounds above $3 billion, Crusoe and Nscale, sell compute, not models. How much money goes into data centers and chips is covered on the page about AI infrastructure.
The full list of 2026 rounds, including the first-half rounds of OpenAI and Anthropic, is on the page of the biggest rounds.
Why the money concentrates
By PitchBook, AI startups raised $407 billion in the first half of 2026, more than the $264.1 billion of all of 2025. About $217 billion of that went to two companies: OpenAI and Anthropic.
In half a year of 2026, AI startups got more than in all of 2025, and more than half of that sum went to two companies, by our calculation from PitchBook data.
The first reason for concentration is the cost of entry. A new-generation model needs billions in chips and data centers, and a small round will not pay for that work.
The second reason: investors follow scale. A large fund finds it easier to put billions into a few companies with clear revenue than to spread them over hundreds of small deals.
The third reason: corporations invest in their own clients. Chip makers and cloud companies give money to model developers, who return it by buying compute.
Risks of concentration
A report by the Roscongress Foundation, a Russian forum and event organizer, put the chance of an AI company market crash in 2026 above 50%. The reason: stock market growth rests on a narrow group of companies.
Since the start of 2024 the S&P 500 index has grown by about $7.5 trillion, and $4.9 trillion of that growth came from 17 AI-linked companies. If expectations for them are not met, not one industry but the whole index will drop.
The report also describes circular financing: chip makers invest in clients who buy their hardware. In this scheme, part of the supplier’s revenue is its own investment coming back.
For a company that buys AI services, the risk is concrete: if the market shrinks, some suppliers may close or raise prices. When you choose a supplier, also look at whose money it runs on.
Scenarios in which the AI market could shrink, and the signs of overheating, are covered on the page about bubble risks.
What this means for smaller companies
If 27 companies take about a third of the world’s venture money for the quarter, other startups must share a smaller part. A small AI company finds it easier to raise investment on revenue and client numbers than on the promise of a model.
If you run an AI startup yourself, the guide is not the size of other rounds but their logic: the investor pays for proven demand. Show revenue, repeat sales and the cost of winning a client before you talk about valuation.
For a company that buys AI solutions, concentration means dependence on a few model suppliers. If a supplier raises prices or changes access terms, it will hit all your solutions at once.
A practical step: build your solutions so the model can be replaced without redoing the process. Then a change of supplier stays a setting and does not turn into a new project.
For an investor from Russia, global billion-dollar rounds are a guide to valuations, not to deal size. The Russian market works with other sums, and a company has to be checked by its own revenue.