A record half-year
AI startups raised $407 billion in the first half of 2026, per PitchBook. In all of 2025 the figure was $264.1 billion: six months brought more than the whole previous year. Both numbers come from one source, so you can compare them.
The whole world venture market raised $560.3 billion in the same half-year, according to KPMG. That is 2.2 times more than a year earlier.
For an AI buyer, growth of 2.2 times means developers have more money for product and sales. There will be more offers on the market, so compare them by the terms for leaving the contract, not only by the first-year price.
By quarter, KPMG gives $332.9 billion in the first quarter and $227.4 billion in the second. The second quarter came out smaller than the first: quarterly totals now depend on a few large rounds.
PitchBook and KPMG count in different ways, so you cannot divide $407 billion by $560.3 billion and call the result the AI share. We explain why sources disagree in the guide’s methodology.
The next period, the third quarter, is covered in the quarterly report. It uses Crunchbase data, which you also should not add to PitchBook numbers.
Two companies took half
About $217 billion of the $407 billion went to OpenAI and Anthropic, per PitchBook. Two companies got more than half of all the money the world’s AI startups raised in six months.
The largest deal of the second quarter, per KPMG: Anthropic raised $65 billion. Rounds, investors and valuations of both companies are on the pages on OpenAI and on Anthropic.
Why this matters to a company that does not invest in startups: many ready-made AI products run on the models of these two developers. Investor money decides how fast these models improve and how much access to them costs.
What follows from this. The AI model market rests on a few companies that raise tens of billions of dollars per round. When they change prices or access terms, every supplier that builds products on their models feels it at once.
For an investor, concentration means something else: the valuation of the whole sector depends on the fate of two companies. We explain how to read such risks and where signs of overheating begin on the page on the risks of AI investment.
For a client company, the conclusion is practical: find out which model each of your AI solutions runs on. If the answer is the same everywhere, all your AI depends on one supplier, even if you signed contracts with different vendors.
The largest deals of the second quarter
KPMG names four large deals in the second quarter of 2026:
1. Anthropic: $65 billion, a model developer.
2. Project Prometheus: $12 billion. The source does not describe the company’s product, so we give only the amount.
3. DeepSeek: $7.4 billion, a Chinese model developer.
4. Anduril: $5 billion, defense technology with AI.
The single Anthropic deal is larger than the next three combined. This is how a market looks when a few rounds decide the result of a quarter.
DeepSeek on the list shows that big money for models goes not only to American companies. Anduril shows a second direction: AI in defense, where the main customer is the state.
The size of a round says nothing about product quality. A round shows investor expectations, while a client needs something else: whether the solution works in its processes and what a switch to another supplier costs.
All deals with amounts, sources and check dates are in the deal tracker.
Russia over the same period
The Russian venture market totaled $55 million in the first half of 2026, according to the Moscow Venture Fund. That is about 48% less than a year earlier.
There were 54 deals with 33 investors. The largest deal of the half-year: almost $13 million for ArchiTech AI, a developer of corporate AI.
There are more deals than investors: 54 against 33. So some investors put money in several times, and the circle of those who fund Russian startups is narrow.
Who provided the money: private funds 67% ($37 million), business angels 16%, corporations 9%, state funds 8%.
The structure shows what Russian venture capital rests on: private funds gave two-thirds of the total, while corporations and state funds together gave less than a fifth.
For a company choosing a Russian AI supplier, the conclusion is this: many young developers have no cash reserve for years ahead. Before you sign, ask who the supplier’s investor is and how many months of work its money will cover.
World AI venture funding set a new record in six months, while the Russian venture market shrank by almost half. These are different markets: world records do not carry over to Russia directly.
We write in detail about money in Russian AI, corporations and state support in the section on Russia.
Conclusions for business
The half-year gives four conclusions for a company that decides how much to invest in AI and in whose AI.
1. AI money is concentrated: about $217 billion of $407 billion went to two companies. Your AI solutions most likely depend on one of them through a supplier.
2. World records do not carry over to Russia: the Russian venture market totaled $55 million in six months and shrank. Do not expect a Russian supplier to grow fast on fund money.
3. Set your company’s AI budget from the losses in your processes, not from news about rounds. The news shows where investor money goes, not where AI will save money for you.
4. Check a supplier before the contract: which model it runs on, where it stores your data and whether you can move to another supplier without losing your settings.
5. Do not mix sources. PitchBook, KPMG and Crunchbase give different totals for similar periods, so for decisions take one source and follow its trend.
The world market changes every quarter. To see the changes without reading the news every day, a monthly review is enough: large deals, reports, numbers for Russia and what they mean for business.
If your company needs not a market review but a decision on where to direct its AI budget, start with a list of processes and the losses in money for each.