The main facts of 2026 in five numbers
Startups worldwide received $159 billion of venture moneyventure capital. Venture capital is money that funds and private investors put into young companies in exchange for a share, expecting a few of them to grow very large. in the third quarter of 2026, by Crunchbase, against $104 billion a year earlier. Of that, $102 billion went to AI startups, 64% of the quarter: more than half of the world’s venture market now goes to one topic.
AI startups raised $407 billion in the first half of 2026, by PitchBook data, against $264.1 billion in all of 2025. The Stanford AI Index put corporate AI investment in 2025 at $581.69 billion, of which $344.66 billion was private investment and $214.44 billion was company purchases.
Capital spendingcapex vs opex. Capex vs opex describes two kinds of company spending. Capital expenditure (capex) is money a company spends on long-term assets such as buildings, servers and equipment, not on the current costs of the month. of Amazon, Microsoft, Alphabet and Meta will reach $700–725 billion in 2026, by the Forbes estimate. In 2025 these four companies spent $410 billion.
The conclusion is simple: AI money gathers in a few companies and in infrastructure. For a company in Russia, this is a reason not to chase other people’s rounds but to count where AI will save money in its own business. All the numbers on the topic, with sources, are in the AI investment statistics.
Who invests: giants, funds, states
Giants. In 2026 Amazon puts about $200 billion into capital spending, Microsoft about $190 billion, Alphabet $180–190 billion and Meta $125–145 billion. This means data centersAI data center. AI data centers are buildings with thousands of servers on graphics cards and other accelerators, with strong power supply and cooling. Neural networks are trained and run there. and the hardware that models run on.
Funds. In the first half of 2026, about $217 billion of the $407 billion went to two companies, OpenAI and Anthropic. Countries. By the Stanford AI Index, in 2025 the US took $285.88 billion of AI investment, Europe $20.92 billion and China $12.41 billion.
States build their own compute capacity and models so they do not depend on outside suppliers. What it means for you: a few companies set the prices of compute and models, so put a backup supplier into your AI plan from day one.
Meta here means Meta Platforms Inc., the owner of Facebook, Instagram and WhatsApp.
Large deals and what they mean for suppliers
The large deals of recent months show where the money goes: into models, data platforms and defense projects. Databricks and Safe Superintelligenceartificial superintelligence. Artificial superintelligence (ASI) is a hypothetical AI that surpasses any human in almost every field: science, invention, strategy and communication. raised $5 billion each in the third quarter of 2026, by Crunchbase. Project Prometheus got $12 billion and Anduril Industries $5 billion in the second quarter, by KPMG data.
Each deal sits in the tracker with its amount, source and check date, and the tracker is updated once a month. If your supplier builds its product on a model from one of these companies, its price and terms depend on their next round. Ask your supplier which model the solution runs on and what will replace it if the terms change.
Russia: corporations against venture money
In Russia, AI money comes mainly from large corporations, not from funds. By a Kama Flow study, Russian IT corporations publicly put $2.7 billion into AI in 2025. Venture investment in AI companies for the same year was about $51 million, a ratio of about 52 to 1.
Experts say the study’s method needs checking, so this is an order of magnitude, not an exact measure. The lesson for a company: there are few independent Russian AI startups, the choice of suppliers is narrow, and you should assess dependence on one supplier before you sign a contract.
Deals, funds and state support in Russia are covered in the review of AI investment in Russia.
What a company should do: invest in its own AI
For most companies, the question is not whether to buy a stake in someone else’s startup but how much to invest in AI inside their own business. In 2026 McKinsey surveyed 1,719 people from 97 countries: 37% of organizations see at least some AI contribution to operating profit (EBIT).
The amount alone gives no result: the return goes to those who pick processes with measurable losses. 32% of organizations dropped the purchase of at least one product because they can build it themselves with AI agentsAI agent. By definition, an AI agent is a neural network that does not just answer but carries out a task step by step on its own: it looks for data, uses programs and checks the result., so the question of whether to build or buy is decided again for each process.
The order of steps is simple. List the processes where money is lost. Estimate the losses of each in money and pick one for a pilot with a metric in the contract. Decide what to build in-house and what to buy ready-made, and only then approve the budget for the year. How to calculate the amount and the payback is covered in the articles about investing in your own AI and about payback.