Russia Faces No American-Style SaaS Apocalypse, but Budgets Are Draining to AI Giants

While American investors write down Salesforce's market capitalization, the Russian IT market looks calm. The author of an analysis on Habr argues that the calm is deceptive: there will be no "SaaS apocalypse," but something less visible and more unpleasant is coming. Below are the main points of his argument. All figures come from his data.

Why the American scenario bypasses us

In the US, SaaS makes money on user counts. An AI agent replaces several of a client's employees, the number of licenses shrinks, and the vendor's revenue falls directly.

In Russia, much of the market works differently: money goes to implementation, customization, and support, not to "seats." So AI hits costs first and then, with a lag, the integrator's margin. The example is simple. A team used to sell 1000 hours of work; now an agent does the same task in 300. Either the contract price drops, or the margin does, or both. The client quickly notices that the work has become cheaper and starts haggling.

The backdrop is not rosy. In 2026 the market will grow to roughly 3.85 trillion rubles, but by only 4%. SaaS growth fell from 60% in 2024 to 24% in 2025, with a forecast of 15–20% for 2026–2027. The growth drivers have run out: critical systems have already been replaced, and the remaining companies are living out their days on old software.

On-premise is back, and data centers have become targets

According to the data cited, about 65% of companies prefer on-premise deployment and only 12% choose the cloud. Among organizations with more than 10,000 employees, 83% favor their own capacity. The market is not "moving to the cloud"; it is splitting into two segments with different economics.

In October, drones attacked three Yandex data centers over four days. Yandex restored 97.16% of client data in Yandex Cloud; the remaining 2.84% was lost irretrievably because clients had no backups on the platform. Ordinary load is growing too: Selectel repelled 88,618 DDoS attacks in the first half of the year, 45% more than a year earlier.

The practical takeaway is not "leave the cloud" but "spread the risk." Critical systems and backups should be kept in different environments and different regions. Businesses are already looking for backup sites in Kazakhstan and Belarus, because free cloud capacity inside Russia is about 90% occupied. The Ministry of Digital Development, for its part, has approved recommendations for protecting data centers from drones, with three criticality categories.

Money is moving to AI platforms

The analysis's most important thesis: traditional SaaS is not dying; its budget is flowing to AI tools. In the US, corporate software spending rose 58%, but this is a redistribution, not a market expansion. According to Ramp, in May 2026 Anthropic overtook OpenAI for the first time in share of corporate AI spending (34.4% versus 32.3%).

Russia takes part in this process in a gray form. Claude and ChatGPT are officially unavailable, and in October Claude blocked about 600 thousand Russian accounts. By August, the turnover of gray-market intermediaries using Chinese proxies reached 100–150 million rubles a month. Their prices are up to 70% below official ones, but the risks come with the discount: data leaks, sudden blocking, and substitution of a cheaper model, such as Qwen or MiniMax.

The alternative is GigaChat and YandexGPT. By some estimates, they trail the flagships by two to three years, but they offer payment in rubles, a contract with a Russian resident, and compliance with Federal Law 152-FZ. This is not a replacement in quality; it trades dependence for control.

What to do

• Project sellers: recalculate your economics for AI. Selling hours is getting harder; selling outcomes is easier.
• Product teams: features get copied within weeks, so defensibility rests on data, integrations, and the environment where agents work. This is how Bitrix24 is building its position.
• Open-source projects: "just open source" is ceasing to be a business. The market is shifting toward hybrid licenses such as Source Available.
• Everyone who has built AI into their processes: decide in advance what happens if the API disappears for hours or forever. The author calls AI the new single point of failure and demands that failover be built into AI usage policy.

There are limitations, too. Some conclusions are the author's interpretation, and the failover section in the original text breaks off. But the core logic is verifiable: look at your revenue structure, the geography of your backups, and where your AI access comes from.