Businesses Ditch ChatGPT for Cheaper Chinese AI Models as Microsoft Bets on Software

In June 2026, businesses began moving en masse from brand-name language models, such as those released by OpenAI, to Chinese alternatives. As the author of an ML digest on Habr points out, this isn't about fashion — it's a cold calculation aimed at cutting API costs. Microsoft, meanwhile, is trying to keep enterprise customers interested in AI through software, not just through the models themselves.

Why companies are leaving familiar brands

Savings on API calls are just the top layer of a deeper shift across the entire AI stack, the digest's author writes. A year or two ago, choosing OpenAI's models or another major Western provider was almost automatic: the brand guaranteed quality, and price was a secondary concern during experiments and pilot projects. By mid-2026, many companies had moved past the pilot stage and into large-scale production use of AI. At those volumes, the per-token price gap between a brand-name model and a cheaper alternative becomes a noticeable line item.

Over the past couple of years, Chinese labs have steadily improved the quality of their open and semi-open models while offering a significantly lower price per token than the flagship products of Western labs. For companies that track the unit economics of their AI features, that choice increasingly looks rational rather than risky.

What Microsoft is doing to keep customers

Tellingly, Microsoft is following the same logic — just from the other side. Instead of competing solely on the quality of its base model, the company is leaning on deep AI integration into its software. Copilot is built into Windows, Office, and a lineup of enterprise tools, which creates switching costs for customers that go beyond the price of a single API request. It's a retention strategy built on the ecosystem, not on the cheapest or most powerful model on the market.

• June 2026 — according to the digest's author, the month when this strategy shift became "the main driver of change in the industry"
• The reason for switching to Chinese models — savings on API costs in production use, not a change in tastes
• Microsoft's response — a bet on software and deep AI integration into its own products, not just on the model itself

What's behind the talk of an AI bubble

Over the past two years, the term "AI bubble" has been used to describe the gap between massive investment in models and infrastructure and the still not entirely clear return on that investment for end businesses. The digest's author suggests viewing the situation not as a question of faith in the technology, but as ordinary cost optimization. Once a company moves from an AI pilot to tens or hundreds of millions of model calls per month, saving a few cents per thousand tokens turns into a noticeable budget line. In that logic, choosing a model provider becomes a routine procurement decision, not a matter of brand loyalty.

This also explains why the author advises looking at "the entire AI stack from top to bottom" rather than just the price of a specific API call. Changes are happening simultaneously at the level of models, cloud infrastructure, and application software, and the winners will be those who can optimize all three layers at once, not just one of them.

What this means

The conversation about the "AI bubble" is increasingly shifting from "will anyone actually use these models" to a far more down-to-earth question: "which provider is cheaper at high volumes." For the market, this means stronger price competition between brand-name and Chinese models. For companies like Microsoft, it creates an incentive to make money not on the model itself, but on the software around it.