Signs of overheating
The first sign: market growth rests on a narrow group of companies. By the Roscongress report, the S&P 500 has grown by about $7.5 trillion since the start of 2024, and $4.9 trillion of that growth came from 17 AI-linked companies.
The second sign: money for startups grows faster than before. By PitchBook data, as reported by Kommersant, AI startups raised $407 billion in the first half of 2026, against $264.1 billion in all of 2025.
The third sign: the money goes to two receivers. About $217 billion of the $407 billion went to OpenAI and Anthropic.
The fourth sign: the giants spend heavily on capacity. By the Forbes estimate, capital spending of Amazon, Microsoft, Alphabet and Meta will reach $700–725 billion together in 2026, against $410 billion in 2025, a rise of about 75%.
No single sign proves an AI bubble. Together they show that investor expectations run ahead of confirmed revenue, and you should keep this in mind for any AI investment.
Circular financing
The Roscongress report describes a separate risk: circular financing. Chip makers invest in clients who buy their hardware.
In such a chain, part of the demand is created by the seller’s own money. The supplier’s revenue grows while its clients raise rounds and falls when the rounds stop.
For an AI buyer in Russia this risk is distant, but its effects are close. If global model and compute suppliers cut spending, prices and terms for clients will change.
The practical lesson: if your solution runs on a foreign model, keep a plan to replace it. Plan the budget so that a rise in the model price does not wipe out the project’s effect.
How the giants invest in startups, and why the money comes back to them as orders, is covered on a separate reference page.
Limits: energy and data
Capital spending of $700–725 billion in 2026 goes into data centers and chips. By the Forbes estimate, Amazon will spend about $200 billion, Microsoft about $190 billion, Alphabet $180–190 billion and Meta $125–145 billion.
These investments have a physical limit. A data center without a grid connection does not work, and building capacity depends on energy as much as on money. Our sources do not count how much energy these sums need, so we give no figures here.
The second limit concerns data, and it is closer to your company. AI works with what sits in your systems, and a messy CRM or incomplete records in 1C, a common Russian accounting system, limit the effect more than the choice of model.
So a check of your own AI investment starts with data: where it sits, who is responsible for it and whether there is enough for the task.
Why companies invest anyway
Despite the talk of overheating, companies keep investing. In a McKinsey survey of 1,719 people from 97 countries in May–June 2026, the answers were:
1. 60% of organizations expect AI investment to grow next year.
2. 44% are scaling AI across the whole company.
3. 28% spend more than 10% of their IT budget on AI.
4. 40% of large companies are scaling AI agents.
The reason is that some companies already see the effect. 37% of respondents see at least some AI contribution to EBIT, earnings before interest and taxes.
There is also a defensive motive: 32% dropped the purchase of at least one product because they can build it themselves with agents. A company that does not invest in AI pays for what its competitors make themselves.
For a Russian company these numbers mean one thing: the question is not whether to invest in AI, but how to check each investment before the money is spent.
How to protect your AI investment
A stock market bubble and AI payback in your company are different questions. By McKinsey, only about 6% of organizations credit AI with at least 5% of EBIT, and what sets them apart is how they measure the effect.
Three steps protect your investment in your own AI:
1. A metric before the start. Each project gets one number it must change: response time to a client, the number of requests handled, the cost of a process.
2. A pilot before scale. One process first, and money for the next stage only after the pilot’s metric is confirmed.
3. A check before you invest. AI due diligence, a full check of the company, the data and the solution before a deal or a large project, shows what the project rests on and where it may not pay back.
How to calculate the effect and the payback period is covered on the page about AI investment payback. What the check includes is described on the AI due diligence page.
The full picture of a company’s investment in its own AI, from the budget to the choice between building and buying, is in the section on AI investment for a company.