MMOOZZGG.com

AI investment

AI venture funding in the third quarter of 2026

6 min read2 sources

AI venture funding in Q3 2026 was $102 billion, 64% of all venture money in the world, according to Crunchbase. Below is a review of the quarter: billion-dollar rounds, stages, regions, acquisitions and stock market listings. At the end: what follows from this for companies in Russia.

On this page · 6 sections

The main numbers of the quarter

In Q3 2026, startups worldwide received $159 billion of venture money, and about 6,000 companies were funded. This is Crunchbase data from 5 October 2026.

Growth year on year is clear: Q3 2025 had $104 billion. Against the previous quarter the sum fell: in Q2 2026 startups received $212 billion.

Q1-Q3 2026 added up to $679 billion. One more quarter remains in the year, and the annual sum will keep growing.

Of the $159 billion, AI startups took $102 billion, or 64%. More than half of the world’s venture market in this quarter is AI.

It matters who got these $102 billion: companies that sell AI, not those that use it. Investor money goes into products that will later come to you by subscription or as part of a supplier’s solution.

The drop against Q2 does not necessarily mean a cooling. Quarterly sums now depend heavily on a few large deals: which rounds fell into Q2 we covered in the results of the first half.

Billion-dollar rounds

In Q3, 27 companies raised rounds of $1 billion or more. This is a record: Q2 had 16 such companies, and Q1 had 14.

These 27 companies received about a third of all world venture money in the quarter. The other roughly 6,000 startups shared the rest.

The largest rounds of the quarter: Databricks and Safe Superintelligence, $5 billion each. Crusoe, Moonshot AI, Mistral AI, Nscale and Kling AI raised more than $3 billion.

Three groups stand out among the recipients: model developers, data platforms and companies that build computing capacity.

For an investor, a billion-dollar round is a bet on one company with a high valuation. For a client, it is a signal: the company will grow and change its plans and terms, and its product will become part of the market where you buy AI.

How such rounds work and who invests in them we cover in our article on billion-dollar rounds.

Stages: late, early and seed

Late stages received $105 billion in Q3. These are rounds for companies that already have revenue and a product, and this is where the billion-dollar deals sit.

Early stages received $40.6 billion, 25% more than a year earlier. These are the first large rounds after a product has proven demand.

Seed rounds raised $13 billion. This is money for the first version of a product and the first clients.

New companies do get money, but several times less: $13 billion in seed rounds against $105 billion at late stages in the same quarter.

The ratio of stages shows where investors look. Late stages took $105 billion of the quarter’s $159 billion: most of the money goes to companies that are already large, not to new ideas.

The growth of early stages matters for buyers of AI solutions: in a year or two, some of these companies will come to market with ready products. The choice of suppliers will widen, but you will have to check them more carefully.

Regions

The US received $91 billion in Q3, about 57% of world venture money. The San Francisco area alone gave 24% of the world total.

The concentration in one region comes from where the developers of the largest models and their investors are based. The rest of the world shares less than half of the market.

The Crunchbase report gives Europe and Asia as separate rows. In our guide, we cover them on the pages on Europe and on China.

For a Russian company, this means a simple thing: most of the models that AI solutions run on are built in another jurisdiction. Not the supplier but another country’s regulator can change the terms of their use.

So when you choose a solution, look not only at the price but also at whether it can move to a model that runs on servers in Russia.

Acquisitions and IPOs

The large acquisitions of the quarter: Nvidia offered to buy Hugging Face for $12.9 billion, AMD announced its intent to buy World Labs for $8.2 billion, and Stripe bought OpenRouter for $7.5 billion.

Two of the three purchases have not closed yet. For Hugging Face the source gives Nvidia’s offer, and for World Labs AMD’s intent, so our guide presents them exactly that way, not as completed deals.

Stock market listings (IPOs, a first sale of shares): ChangXin Memory Technologies raised $8.6 billion at a valuation of $85.5 billion, and Bending Spoons $1.6 billion at a valuation of $18.4 billion.

Two more listings concern hardware and robots: Enflame raised $912 million at a valuation of $25.5 billion, and Unitree Robotics $905 million at a valuation of $9 billion.

Three of the quarter’s four listings went to makers of memory, chips and robots. Bending Spoons alone represents application software on the stock market this quarter.

Acquisitions change the market for clients. When a hardware maker or a payments company buys a platform, its prices, terms and product priorities change. All deals of the quarter with sources are in our deal tracker.

What it means for Russia

In Russia, AI and machine learning received $32 million of venture investment in 33 deals in the first 9 months of 2026, according to Venture Guide. That is first place among industries by both volume and number of deals. The data is preliminary, as of 4 October 2026.

The largest deal of the period: ArchiTech AI raised $13 million in a Series C+ round, with UKM Group as the investor.

The scale is not comparable to the world one, but the direction is the same: money goes to AI ahead of other industries. For a company, this means more Russian AI suppliers will appear, and you will need to check their maturity before the contract.

We do not convert the Russian figures into dollars: the sources use different exchange rates and methods. Compare the direction of the money, not the sums.

What to do now:

1. Do not compare your AI budget with world rounds: your processes and the losses in them set the benchmark.

2. Before you choose a supplier, check which model it runs on and where the data is stored.

3. Follow the market once a month, not every day: quarterly reports change the picture more than the news does.

Sources and fact-checking

Checked by the editors on 14 October 2026.

  1. Crunchbase News, Gené Teare, 5 October 2026 (news.crunchbase.com)
  2. ComNews, citing Venture Guide (Moscow Venture Fund), 5 October 2026 (comnews.ru)

Found a mistake? Tell us

Publications

Mike Mozg

Anthropic IPO raises $81.34 billion and makes its compute costs public

The Anthropic IPO raised $81.34 billion, and trading on Nasdaq began on 19 November 2026, the Russian business daily RBC reported. That beats SpaceX's base IPO, but it becomes a record only once the agreed $26 billion private placement closes. I don't read this as market news: Anthropic's public filing shows what compute costs a frontier AI lab. In my view, it shows where prices, limits and access

News6 min
Contact us