The list of rounds
Rounds of 2026 of $5 billion and more, from public sources, from largest to smallest:
1. OpenAI: $122 billion at a valuation of $852 billion, lead investors Amazon, Nvidia and SoftBank, by Bloomberg Law.
2. Anthropic: $65 billion at a valuation of $965 billion, lead investors Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital.
3. Project Prometheus: $12 billion in the second quarter, by KPMG data.
4. DeepSeek: $7.4 billion in the second quarter, by KPMG data.
5. Databricks: $5 billion in the third quarter, by Crunchbase.
6. Safe Superintelligence: $5 billion in the third quarter, by Crunchbase.
7. Anduril: $5 billion in the second quarter, defense AI, by KPMG data.
Five more companies got over $3 billion each in the third quarter: Crusoe, Moonshot AI, Mistral AI, Nscale and Kling AI. The source did not disclose the exact amounts, so they are not in the numbered list.
Five of the seven rounds on the list went to model developers: OpenAI, Anthropic, Project Prometheus, DeepSeek and Safe Superintelligence. Databricks works with company data, and Anduril makes defense systems.
All deals of 2026, including purchases and stock market listings, with the check date of each line, are in the tracker.
Who invests
In the OpenAI round, the money came from tech companies: Amazon with $50 billion, Nvidia with $30 billion and SoftBank with $30 billion.
SoftBank, a Japanese investment holding, gave OpenAI as much as Nvidia: $30 billion each. Amazon invested the most, $50 billion, but not the whole sum at once.
About $35 billion from Amazon depends on an IPO, the first public sale of OpenAI shares on a stock exchange, or on a technology goal. So part of the announced sum will come only if the company meets the conditions.
In the Anthropic round, funds were the lead investors: Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital. Google invested several billion dollars under an earlier commitment.
Two types of investors stand out. Corporations invest in model developers that buy chips and cloud compute from them. Funds buy a stake in the future growth of the valuation.
For a company choosing an AI supplier, this is a practical question. Large model developers have their own suppliers among their shareholders, and the terms of access to the models are linked to these contracts.
Why the rounds are so large
A model developer needs chips, data centers and electricity, and each new generation of models needs more of them. So round sizes are measured in billions, not millions.
In the third quarter of 2026, 27 companies got rounds of $1 billion or more, a record. In the second quarter there were 16 such companies, and in the first quarter 14.
These 27 companies took about a third of the world’s venture money for the quarter. The money goes to those who already raised a large round: the investor pays for scale and for a place in the queue for compute.
Why startups raise billions, and what this concentration means as a risk for the market, is covered on the page about billion-dollar rounds.
What happens with valuations
In May, Anthropic got a valuation of $965 billion and for the first time passed OpenAI, which its spring round valued at $852 billion. Neither company trades on a stock exchange yet.
A private company valued at about a trillion dollars means the investor buys a stake in future revenue, not current revenue. At such a price, even a small slowdown in growth changes the value of the stake a lot.
For some rounds on the list, the source did not disclose a valuation: Project Prometheus, DeepSeek, Databricks and Safe Superintelligence have none. You cannot compare companies by round size alone, because the same sum can mean a different stake.
When you read news about a round, look for three numbers: the amount, the valuation and the investors’ stake. Without the valuation, the amount only tells you how much money the company can spend.
Russia against this background
In Russia, AI and machine learning got $32 million of venture investment in 33 deals in the first 9 months of 2026. This is first place among industries by volume and by number of deals. The data is preliminary, as of 4 October 2026.
The largest Russian round of the year: ArchiTech AI, $13 million from UKM Group in a C+ round. This is orders of magnitude smaller than any round on the list above.
If you raise money for a Russian AI company, global valuations will not help in talks. The investor will ask about revenue, paying clients and rights to the data and the code.
For a Russian company, global sums are a guide not by size but by logic: big money goes to whoever has a product and paying clients. Russian rounds with amounts and sources are in the tracker of Russian deals.