Jensen Huang's resignation from Nvidia is no longer just a corporate scandal. According to Izvestia, experts expect that by 2030 the company's share of the AI chip market will decrease from 90% to the current 50-60%. This is a good sign that one vendor has 9 cents of the market, it is no longer a market. Depending on a single provider is not only dangerous chips. With AI models, the logic is the same: the wider the choice, the better for those using these models. In Russia, this issue is discussed at the regulatory level: models are divided into sovereign and national bodies, and they discuss in which areas they can be used. The separation itself, according to experts, seems reasonable. National models can be created using open source components, so dozens of companies can develop them. What is important is that the model class should not decide in advance who will get access to the market. This problem is most acute in industries that offer the use of sovereign models mainly. If you force them to work only with them, the business will choose by the label, not by quality. Security and data protection requirements can be defined, but within these frameworks, companies should use what is more effective for solving their problems, whether it is a sovereign model, national or internal development on open components. Market participants also understand this. Yandex advocates for the preservation of competition and believes that industries should be able to "independently eliminate technological solutions for specific tasks". Avito believes that strict restrictions can slow down the development of technologies and affect the quality of AI products. MWS AI warns that with a limited choice of models for some tasks in critical industries, solutions that match the quality and cost may not be available.