Sber and Yandex Debate AI Models: What Russian Industry Actually Needs

The recent public discussion between Sber and Yandex once again demonstrated how the conversation about artificial intelligence in Russia is concentrated around the largest technology corporations. Herman Gref stated that Yandex is fine-tuning the Chinese Qwen model, while Yandex disagreed with this assessment, emphasizing that the company continues to develop its own foundational models, maintains a full development cycle, and does not depend on external solutions.

This discussion is important from the perspective of technological sovereignty, data control, computing infrastructure, and the ability of Russian companies to independently create core technologies. At the same time, however, it reveals a certain distortion in the perception of AI development. One might get the impression that the future of Russian AI is determined exclusively by competition among a few major companies and their foundational models.

For an industrial enterprise, this dispute is of secondary importance. A plant manager primarily needs to understand whether artificial intelligence can help reduce defects, prevent equipment failures, minimize downtime, improve safety, optimize resource consumption, or accelerate work with technical documentation. Which model underlies the solution certainly matters in terms of security, cost, and vendor dependence, but a foundational model alone does not generate economic impact. Results emerge only when the technology is properly integrated into a specific production process.

The development of Russian artificial intelligence cannot be built solely around a few universal platforms. The country needs not only large models but also hundreds of applied solutions, industry-specific developers, engineering teams, integrators, and small IT companies. Such a company may not have the resources to create its own foundational model, but it may deeply understand a particular technological process, specific equipment, or the specifics of a given industry.

It is often small, specialized developers who are able to offer the most precise solution to an industrial enterprise. However, they face significantly greater difficulty in gaining access to customers, production data, testing sites, computing resources, and existing support measures. As a result, large corporations discuss technological leadership among themselves, while small IT companies continue to search manually for clients and offer them already developed products.

This is where one of the main gaps in the Russian AI market emerges. An enterprise often does not yet know which specific task should be addressed with AI. It may lack prepared data, an internal technology team, and an understanding of where to begin. The IT company, for its part, is interested in selling its own product. As a result, the dialogue often begins not with an analysis of the production problem but with a presentation of the capabilities of a ready-made solution.

The developer explains what the system can do, and the enterprise tries to find a place for it within its processes. A pilot project may then be launched that appears technologically interesting but does not solve a truly significant problem. The expected effect is not achieved, and the enterprise becomes disappointed in artificial intelligence. Yet the cause of failure may not have been the technology itself, but an incorrectly formulated task, unprepared data, or an attempt to adapt production to an already existing IT product.

The correct logic should be the reverse: first identify the enterprise’s real problem, analyze the production process and the state of the data, then consider several possible solutions, conduct an independent assessment, launch a pilot, and measure the results. Only then should a decision on scaling be made.

From Fragmented Support Measures to a Clear Roadmap

Russia already has a significant number of support measures for IT companies, digitalization, robotization, automation, and the implementation of domestic software. Tax incentives for accredited IT companies remain in place, along with investment tax instruments, grant programs, preferential financing, and subsidies. In 2026, in particular, a separate selection process was conducted for manufacturing enterprises to reimburse costs associated with production robotization.

The problem lies not so much in the absence of support as in the fact that neither enterprises nor developers often see the system as a whole.

An industrial enterprise seeking to implement a digital solution does not always understand where to turn. One institution supports software development, another the purchase of equipment, a third robotization, a fourth the implementation of a specific category of domestic software. Each program has its own deadlines, criteria, co-financing requirements, lists of eligible expenses, and application procedures.

It may be difficult for an enterprise to determine whether support is available for preliminary diagnostics, data preparation, and a pilot project, or whether only the purchase of a ready-made product is financed. It is not always clear whether the solution must be included in the Unified Register of Russian Software, whether adaptation, integration, and staff training costs are eligible, and whether the relevant program operates continuously or only within a short competitive window.

A similar problem exists for IT companies. Developers, especially small and regional ones, often do not know through which institution they can reach an industrial customer, gain access to a testing site, or secure funding for a pilot implementation. As a result, they act as they are accustomed to: independently searching for clients and attempting to sell a ready-made solution.

As a consequence, enterprises and developers operate almost in parallel worlds. The enterprise may not be able to formulate a technological request. The developer seeks to promote the product it already has. Support institutions offer various programs, but market participants find it difficult to determine which one matches a specific task. The necessary synergy between a production problem, technological competence, and a financial mechanism does not arise.

The more the system depends on individual competitive selections, complex applications, and discretionary expert decisions, the higher the risk of opacity and potential misuse. This does not mean that the competitive principle itself is wrong. Grants and subsidies are necessary for fundamental research, high-risk developments, critically important technologies, and initial industrial testbeds. But the mass adoption of AI requires clearer and continuously operating rules.

An enterprise that invests its own funds in technological diagnostics, data preparation, solution adaptation and integration, specialist training, and conducting an industrial pilot should clearly understand in advance what tax benefit, investment deduction, preferential financing, or compensation mechanism it can expect.

Global practice shows that tax incentives can serve as a broad-based tool for supporting research and technological development. For example, in the Netherlands, the WBSO mechanism allows enterprises to reduce labor costs and other expenses related to research and development. In 2025, nearly 19,000 companies used it, with 97% of beneficiaries belonging to small and medium-sized businesses. Artificial intelligence was among the fastest-growing areas of supported activity.

Russia should study such experience while preserving its own development institutions. Functions need to be reasonably divided. Clear tax and investment incentives can be used for the mass support of implementation, while direct government funding should remain for strategic, scientific, and the most high-risk projects.

At the same time, support for domestic technologies should not lead to the isolation of the Russian market. Technological sovereignty does not mean rejecting global scientific experience, open technologies, or international cooperation, but rather the ability to independently define development goals, control critical data and infrastructure, choose suppliers, and maintain the capacity to develop one’s own solutions.

Support for domestic software should also not result in the creation of a closed market centered around a few largest companies. For an industrial enterprise, what matters is not only the product’s origin or brand recognition, but the solution’s ability to actually solve the assigned task, ensure security, deliver economic impact, and provide ongoing support.

The Initiative of the Association and the CCI RF Working Group on AI

The Association of Innovative Solutions and Artificial Intelligence “Regions of the 21st Century,” together with the Working Group on Artificial Intelligence under the CCI RF Council for Information Technology and the Digital Economy, has put forward an initiative to establish regional AI competence centers within the chambers of commerce and industry of the constituent entities of the Russian Federation.

This initiative was announced at the first meeting of the Working Group on Artificial Intelligence under the CCI RF Council for Information Technology and the Digital Economy and received support from the participants. We intend to work consistently toward its practical implementation.

Regional chambers of commerce and industry have a strong understanding of the structure of the local economy and interact with industrial enterprises, the business community, government authorities, and development institutions. Therefore, they can serve as clear entry points for enterprises interested in implementing artificial intelligence but unsure where to begin.

An enterprise should be able to approach such a center not with a request to purchase a specific software product, but with a description of a production or management problem. It should then receive comprehensive support: assistance in formulating the task, initial process diagnostics, assessment of data readiness, independent evaluation, selection of several possible technological solutions, and information about available government support measures.

Regional competence centers should become a connecting link between industrial enterprises, Russian IT companies, industry experts, research organizations, development institutions, and government authorities. For small and regional developers, this would create a clear and equal mechanism for accessing real production challenges. For enterprises, it would provide the opportunity to choose not between promotional presentations from individual vendors, but among several solutions that have undergone preliminary expert evaluation.

Such a center should not become yet another software catalog and should not promote the interests of any particular technology company. Its task is to build a clear route for the enterprise from a production problem to an implemented solution: problem identification, diagnostics, expert review, technology selection, identification of support measures, pilot implementation, impact assessment, and scaling.

In the long term, the work of regional centers will make it possible to create an objective map of real challenges and barriers to AI implementation in industry. On this basis, existing support measures can be improved, shortages of domestic solutions identified, requests formulated for developers, and regulatory changes proposed.

The development of artificial intelligence in Russia should not be reduced to a dispute between a few major corporations over whose foundational model is more domestic. Nor should it begin with the selection of a predetermined supplier or software product.

Only in this case will artificial intelligence become not an object of corporate competition or another form of technological reporting, but a practical instrument for the development of Russian industry, regions, and a competitive domestic IT ecosystem.