Why companies buy instead of building
Large companies buy AI developers when time costs more than money. A team of your own needs hiring, data and months for a first version, while a company you buy already works with clients.
There are more such deals: about 305 in the first quarter of 2026 alone. There are usually three reasons to buy rather than build:
1. A team. Hiring a strong team one person at a time is slow, and with a company the buyer gets people who already work together.
2. A product and clients. Autodesk bought MaintainX for $3.6 billion and received its equipment maintenance software together with its clients.
3. A place in the chain. Nvidia offered $12.9 billion for Hugging Face: a chip maker reaches for the platform where developers get open models.
Deals worldwide
Deals of 2026 with verified amounts, according to Crunchbase and Solganick:
1. Nvidia and Hugging Face: an offer to buy for $12.9 billion, Q3.
2. AMD and World Labs: an intent to buy for $8.2 billion, Q3.
3. Stripe and OpenRouter: a purchase for $7.5 billion, Q3.
4. Autodesk and MaintainX: $3.6 billion, May.
5. Roche and PathAI: more than $750 million, May.
6. Anthropic and Stainless: more than $300 million, May.
Pay attention to the status. An “offer” and an “intent” are not yet a closed deal, and such figures cannot be taken as the price at which the market already buys.
The buyers here are not only AI giants: Roche works in pharmaceuticals, and Autodesk makes design software. An industry company buys AI for its own task, and this path suits a smaller company too, if the goal is specific.
The full list of mergers and acquisitions of 2026 is on a separate reference page.
Deals in Russia
In Russia, AI is funded mostly by the corporations themselves, not by venture funds. According to a study cited by ComNews, Russian IT corporations publicly directed $2.7 billion to AI in 2025.
Venture investment in AI companies in the same year was about $51 million, which is 32% of a $159 million venture market. The authors estimate the ratio of corporate to venture money at about 52:1.
Experts say the method needs checking, so we give these figures as an estimate, not as an exact count.
For comparison, in 2021 the whole venture market was $2.5 billion, and AI took $361 million of it.
What this means for a buyer of a stake: there are few public Russian deals with a price, and nothing to compare a valuation with. The price has to rest on your own check of the company, not on market examples.
Verified Russian deals with sources are collected in the deal tracker for Russia.
What to check before a deal
Before signing you need due diligence, a full check of the company: legal, financial and technical. In an AI company, the main risks are usually in the technical part.
1. Whose model. Its own model or a layer on top of someone else’s: if the product runs on another company’s model, that company sets its price and access.
2. Data rights. Which data the model was trained on, and whether the company has rights to it. Without rights, the buyer gets a risk of claims.
3. Dependence on people. If the product rests on two engineers, their leaving after the deal wipes out the purchase. Write the terms of their work after closing into the contract.
4. Revenue and clients. How many clients pay, and how many of them renew the contract.
5. The deal status on the market. Announcements such as Nvidia’s $12.9 billion offer for Hugging Face may not end in a purchase. Rely on closed deals, not on headlines.
How such a check works for an AI company is described on the page about AI due diligence.
How to value an AI company
The value of an AI company has three parts: revenue and its growth, the cost of repeating the product yourself, and the value for the buyer.
Global deals give a sense of scale, but not a price per dollar of revenue. Roche paid more than $750 million for PathAI, but without PathAI’s revenue in open data you cannot compare this price with your target.
In Russia, there are even fewer reference points: about $51 million of venture money in AI in 2025 against $2.7 billion of corporate money, by an estimate with a note on its method.
A practical order:
1. Calculate what it costs to build the same product with your own team, and in what time.
2. Calculate the effect of the product in your company over 3 years.
3. A stake price above the smaller of the two sums needs an explanation: what exactly you pay extra for.