From OpenAI to Sarvam, governments may want skin in the AI game

The idea would have sounded far-fetched even a year ago. Governments regulate companies, subsidise them and sometimes shield them from foreign competition. They do not usually ask for a seat on the table. Yet recent reports on OpenAi in the US and Sarvam AI in India suggest that governments could end up owning stakes in some of the most important artificial intelligence (AI) companies being built today. It could mark a larger shift in how states are beginning to view AI not merely as a technology sector but as a strategic national asset.

A changing mindset

OpenAI has discussed granting a 5% stake to the US government, the Financial Times reported. The proposal was part of broader conversations around growing scrutiny of AI firms in Washington and concerns over whether ordinary Americans would benefit from the enormous wealth that advanced AI could create. OpenAI reportedly suggested that other major US AI companies could consider similar arrangements as well.


The report emerged at a time when Washington has become increasingly assertive in its dealings with frontier AI firms. The administration has intervened in the release of advanced models such as of Claude AI, imposed restrictions on access to some cutting-edge systems and intensified scrutiny around national security risks and economic disruption. The proposed stake was reportedly framed as a way to align public interests with the gains generated by AI and reduce political resistance to the sector's rapid rise.

A week earlier, ET reported that the Indian government could end up holding a 1-2% stake in Bengaluru-based AI startup Sarvam. The proposed stake would arise from computing support provided under the IndiaAI Mission as part of Sarvam's ongoing $300 million funding round, which values the company at roughly $1.5 billion. Unlike a conventional equity investment, the government support is linked to compute infrastructure and other resources being extended through the IndiaAI programme. Reports suggested that the stake could be acquired through convertible instruments issued against this support.

Neither case represents a completed transaction. Yet the fact that such proposals are being openly discussed in two major democracies suggests that policymakers are beginning to think differently about ownership in the AI era.

Also Read | Sam Altman wants to give a 5% OpenAI stake to the US government: Report

AI begins to look like strategic infrastructure

The conventional argument for government involvement in technology has been straightforward. Governments fund research, build infrastructure and create favourable conditions for innovation. Ownership is usually left to founders and investors. AI is increasingly challenging that model. Frontier AI systems are becoming deeply intertwined with national security, public administration, defence, intelligence, healthcare and financial systems. Governments are not merely customers of these technologies. They are becoming dependent on them.

This creates a new problem. A country's most important strategic capability may be controlled by a private company whose priorities do not always align with national interests. The concern is no longer limited to who owns the technology. It extends to who controls access, who determines deployment policies and who makes decisions during moments of geopolitical tension.

The restrictions that governments, especially the US and China, have imposed on advanced chips, foundation models and cross-border AI transfers offer a glimpse of this reality. The US has used export controls to limit access to advanced semiconductors and AI capabilities. China has tightened oversight of advanced AI technologies and their transfer abroad. AI is increasingly being treated less like software and more like strategic infrastructure. Once a technology enters that category, governments often seek deeper involvement than regulation alone can provide.

Also Read | IndiaAI backing could make government a minority shareholder in Sarvam

Regulation versus ownership

At first glance, ownership may seem unnecessary. If governments can regulate AI companies, restrict exports, impose licensing requirements and control access to computing infrastructure, why buy shares at all? The answer lies in the difference between authority and influence. Regulation allows governments to stop companies from doing certain things. Ownership creates a mechanism for participation in long-term decision-making.

A minority stake does not give a government operational control. It does, however, provide information rights, governance access and a closer institutional relationship with companies that may eventually become central to national capabilities. Governments are not necessarily looking to nationalise AI. They may simply want a permanent seat close to the centre of gravity. The Sarvam and OpenAI discussions both point in this direction. In neither case is the government seeking control. The proposed stakes are small. What matters is not the size of the holding but what it represents.

The rise of sovereign AI

The phrase "AI sovereignty" has become increasingly common in policy circles over the past two years. Traditionally, technological sovereignty meant having domestic companies, domestic infrastructure and domestic expertise. AI expands the concept further. Countries now worry about who owns the models, where they are trained, where the data resides and who can restrict access to critical capabilities.

For India, this concern has helped drive initiatives such as the IndiaAI Mission and support for indigenous foundational models. Sarvam itself has emerged as one of the most prominent beneficiaries of this push. The company has been involved in efforts to develop Indian-language models and sovereign AI capabilities under government-backed programmes.

For the US, sovereignty concerns revolve around maintaining leadership in frontier AI while ensuring that the technology remains aligned with national priorities. Ownership stakes can be viewed as one more instrument in this broader sovereign AI toolkit. Alongside subsidies, compute infrastructure and export controls, equity becomes a way of institutionalising the relationship between the state and strategic AI firms.

Political derisking

There is another dimension that receives less attention. The largest AI companies face growing political risks. Concerns over job displacement, market concentration, misinformation and economic inequality are becoming central political issues. As AI systems become more powerful, governments are likely to demand greater accountability and a larger public share of the benefits.

From the perspective of AI companies, offering governments a stake could help create a political compact. If governments and citizens participate in the upside generated by AI, they may be more willing to support policies that enable its development.

This appears to be one of the ideas underlying the OpenAI proposal. Discussions around public ownership were connected to broader concerns about ensuring Americans share in AI-generated wealth. Some proposals have even drawn comparisons with the Alaska Permanent Fund, which distributes a portion of resource wealth to residents.

In India, the Sarvam proposal emerges from a different context but points toward a similar logic. The state is not merely funding innovation. It is potentially becoming a participant in the value that innovation creates.

The beginning of a larger trend?

It is still too early to conclude that government ownership of AI companies will become standard practice. Most governments are likely to remain focused on grants, procurement contracts, tax incentives and infrastructure support. Many will be reluctant to take direct equity positions in private firms. Yet a new direction is becoming clearer.

As AI becomes more central to economic growth and national power, governments are looking for mechanisms that go beyond traditional regulation. They want closer relationships with the companies building foundational models. They want greater visibility into strategic technologies. They want assurances that national interests will be protected if AI becomes as important as electricity, telecommunications or defence systems. The OpenAI and Sarvam discussions are noteworthy not because of the size of the proposed stakes. Five percent in one case and one to two percent in the other are not transformative holdings. What matters is that governments are beginning to contemplate ownership at all.

That shift suggests that the next phase of the AI race may not be defined solely by competition between companies. Increasingly, it may involve partnerships between states and the firms building the technologies that governments believe will shape economic power, security and geopolitical influence for decades to come.