AI Battle: Should the State Subsidize Domestic AI Models

May 15, 2026 · 8 min read

Question: Should the state subsidize domestic AI models?

Perplexity

Yes, the state should subsidize domestic AI models, but not "just anything," but selectively – where there is a strategic effect: public services, critical infrastructure, data security, industry, and education. The logic here is simple: developing a competitive AI model is a very expensive race, and without initial support, the market usually shifts towards ready-made foreign solutions because they are already mature, cheaper to implement, and have a faster return on investment. This is precisely why the discussed bill around AI in Russia includes state support measures for "sovereign," "national," and "trusted" models: the state is trying not only to regulate the market but to create an internal technological loop that can be used without dependence on external suppliers. In practice, this can work if subsidies are tied to measurable results – model quality, security, localization, and implementation in specific industries.

But there is a flip side: subsidizing without strict criteria means risking money and entrenching inefficiency. Developing a model on par with foreign leaders requires hundreds of billions of rubles, and even market participants admit that "from scratch" without open-source and global cooperation, it is almost always too expensive and does not guarantee comparable quality. A bad scenario is when budget money goes to "national" solutions that are formally domestic but in fact lose in quality, and businesses and government agencies are still forced to use stronger external technologies. Therefore, a reasonable policy is not to hand out subsidies to all developers, but to finance infrastructure, data, computing power, pilot projects in the public sector, and applied models where the effect is measurable. For example, subsidies may be justified for AI in medicine, document management, industrial diagnostics, or cybersecurity, but not as a perpetual subsidy for the mere fact of being "domestic."

Sources:

Gemini

The question of the necessity of state subsidization of domestic AI models is a subject of active discussion, with arguments both for and against. Proponents of subsidization point to the strategic importance of AI development for national security and technological sovereignty. They emphasize that AI has already become a foundational technology affecting all spheres of life, from the economy to defense. State support can stimulate mass adoption of AI in critical sectors such as industry, agriculture, transport, and healthcare, which, in turn, can lead to significant cost savings for organizations. Furthermore, subsidies can help domestic developers compete in the global market and contribute to the creation of "sovereign" AI models that meet national interests and standards. An example of such support is the mandatory implementation of AI when companies receive state subsidies, which is already practiced in Russia.

On the other hand, critics of state intervention note that an economy with "distorted incentives" may require specific support tools, but subsidies do not always guarantee effectiveness. There is a risk that without proper control and transparency, subsidies can become a tool for "budget skimming." Additionally, excessive reliance on state funds can slow down innovation and lead to the creation of uncompetitive products that lag behind global counterparts. Some experts believe that businesses themselves should be interested in implementing AI, and the state should create conditions for free competition and market development, rather than directly financing individual developments. It is also important to consider that even with significant state support, a lack of access to global technologies, service support, and a stable ecosystem can be a serious obstacle to the development of domestic AI.

There are various approaches to AI support in global practice. Some countries, like the USA and China, focus on mass adoption and service development, while the EU concentrates on regulation and infrastructure investment. Russia, according to its strategy, views AI as a tool for security and technological sovereignty, integrating it into state programs and defense projects. At the same time, as experts note, the Russian AI market is showing growth despite sanctions, which may indicate the presence of internal potential. The state actively stimulates AI adoption, including through the conditions for receiving state subsidies, expecting significant economic effects in the future.

Sources:

AI in Public Administration: Global Experience - 24KZ
Government to Mandate AI Implementation for Companies Receiving Subsidies
Authorities to Mandate Business AI Implementation for 13 Subsidies
How Different Countries Use AI: Case Studies in Key Sectors - Kazinform
How AI Legislation is Formed in Different Countries | ComNews
National AI Development Strategies until 2030 - IT-World
Can AI Govern a State - VELIKOROSSЪ
Artificial Intelligence: A New Mandatory Requirement for Receiving 13 Subsidies - JSC ASP
AI Application by Companies to Become Mandatory for Receiving Federal Budget Subsidies - BUSINESS Online
Does AI Deliver Expected Economic Effects - Big Data Association
What Should the Russian AI Model Be Based On? - Habr
Computational Dead End. Why Russian AI Lacks Capacity - ComNews
And a People's Phenomenon: What Business Thinks About the Sovereign AI Bill | Forbes.ru
How AI Affects the Economy and Business: 6 Arguments "For" and "Against"
Artificial Intelligence in Russia: How Investments and State Support are Changing the Landscape - Journal

ChatGPT

State subsidization of domestic AI models is an issue that requires a balanced approach. On the one hand, supporting national developments stimulates technological sovereignty and reduces dependence on foreign technologies, which is particularly important in conditions of global instability and sanctions. China can serve as an example, investing significant funds in the development of its own AI platforms, which allows for the development of a local industry and the creation of high-tech jobs. Furthermore, state investments can accelerate the adoption of technologies in strategically important sectors such as healthcare, education, and industry.

On the other hand, subsidization should not turn into a subsidy for inefficient projects or support technologies that lack market demand. It is important that state support stimulates competition and innovation, rather than leading to monopolization and inefficient budget spending. In countries with a developed AI sector, state programs are often aimed at creating an ecosystem of startups and research centers, which increases the overall level of expertise and promotes market development. South Korea can be considered an example, where the state actively finances research initiatives and creates conditions for the commercialization of developments.

Thus, state subsidization of domestic AI models is justified if it is competently organized and aimed at creating conditions for the long-term development of the industry, increasing technological independence, and competitiveness. Otherwise, there is a risk of inefficient resource utilization and stagnation of innovative potential.

Sources:

Analysis of AI development strategies in China, Nature
State policy in the field of AI in South Korea, Ministry of Science and ICT of the Republic of Korea

Grok

The state should subsidize domestic AI models, but only in a targeted manner— focusing on strategic areas like national security, critical infrastructure, and independence from foreign vendors. This approach works globally: China, through massive subsidies (as part of "Made in China 2025"), has propelled companies like Baidu and Alibaba to leadership in certain AI metrics, surpassing the US in 66 out of 74 advanced technologies. In the US, the CHIPS Act and $3.3 billion in federal AI R&D investments for 2025 have secured the leadership of OpenAI and Anthropic, despite private capital. Russia is also on the right track: from 2023 to 2026, 3.8 billion rubles have been allocated for AI research centers, grants up to 50 million rubles, and subsidies of 3 billion for developers, which has already yielded projects like Stratagonia for public administration.

Opponents of subsidies appeal to the market: AI is "drowning in private capital," and money would be better spent elsewhere, as in the case of the EU, where despite efforts, the bloc lags behind the US and China in investments in "winner-takes-all" AI. In Russia, 22% of AI costs are covered by the state, with the rest by business, which minimizes risks./iskusstvennyij-intellekt-v-rossii-kak-investiczii-i-gosudarstvennaya-podderzhka-menyayut-landshaft) However, without state support, the risks are higher: sanctions and monopolies of Big Tech (Nvidia, OpenAI) will make domestic models hostages, as is currently the case with chips.

Ultimately, subsidies are not charity but an investment in sovereignty: they accelerate breakthroughs where the market is slow due to high barriers (data, computing). An example is Russian grants for sugar beet evaluation by "Mallen Systems," where AI is paying off in the agricultural sector. The key is transparency and KPIs to avoid failures like those in Europe.

Sources:

National Strategy for AI Development in Russia (tadviser.ru)
AI Support Measures in Russia (ai.gov.ru)
CSIS: Federal R&D Funding for AI (csis.org)
Atlantic Council: DeepSeek and AI governance (atlanticcouncil.org)
The Hill: AI doesn't need subsidies (thehill.com)

Claude

I will allow myself to find up-to-date information on state support for AI and existing practices. It's a controversial issue, but it becomes clearer with arguments from both sides. Here is an argued response:

For State Subsidies

There are compelling reasons for state support of AI. Firstly, it's a matter of sovereignty and technological security. Russia's AI development strategy envisages the priority use of Russian generative AI models in public administration and critical information infrastructure. Relying solely on foreign models (OpenAI, Google, Claude) creates dependence on geopolitical conditions and export restrictions.

Secondly, state support accelerates fundamental research that the market initially does not fund. This is noticeable globally: governments of many countries are developing national AI development strategies that include practical steps for implementing new solutions in various economic sectors. In Russia, 26.49 billion rubles are allocated for the program's implementation during 2025–2027, which allows for the creation of competence centers and personnel training. Over 100 universities have launched AI specialist training programs, more than 17.6 thousand specialists have been trained, and 36 bachelor's and 86 master's programs are in operation.

Against Unconditional Subsidies

But here's a counterargument: large-scale state support for AI can be ineffective if there are no concrete results. History has shown that direct funding without performance requirements leads to losses. The Ministry of Industry and Trade attempted to sue a company for 500 million rubles due to delays in launching domestic microprocessors, for which state subsidies were allocated (the Arbitration Court denied the ministry's claim).

Conditional support with obligations works more effectively. The subsidy recipient must invest their own funds of at least 30% of the project cost and undertake an obligation to implement the developed solutions in an amount at least twice the subsidy amount. Such a mechanism motivates developers to create real, commercially viable solutions.

Conclusion: The state should subsidize AI, but not just by handing out money – but by creating a system of incentives through grants, tax breaks, and co-financing requirements. Without this, state investments risk becoming investments in image rather than in real technologies.

Sources:

National AI Development Strategy (TA Adviser)
Ministry of Industry and Trade on AI Project Subsidies (Kommersant)
Federal Project "Artificial Intelligence" (nationalprojects.ru)

Sources:

Blog