OpenAI muốn chính phủ góp vốn, liệu có thể 'mua' được sự ổn định trong quản lý?

TL;DR

• According to Axios, OpenAI is in early discussions with the Trump administration, and the plan may involve ceding about 5% equity to a public fund.
• The disagreement lies in whether the government receives only financial returns or can intervene in corporate governance through voting rights and board seats.
• Related entities: OpenAI, Microsoft, Nvidia, Google, Meta.

According to Axios on July 2, OpenAI is in early discussions with the Trump administration. The plan may involve ceding about 5% equity to a vehicle similar to a sovereign wealth fund, allowing the American public to share in the gains from AI growth. No agreement has been reached, and there have been no official announcements from OpenAI or the White House.

This number is not insignificant. OpenAI announced on March 31 that it had completed a $122 billion funding round, giving it a post-money valuation of $852 billion. Based on this calculation, 5% corresponds to approximately $42.6 billion. This is not a symbolic donation but a financial stake large enough to change the terms of policy discussions.

What concerns investors is not whether the American public will soon receive AI dividends, but rather, as frontier models increasingly touch upon national security, job displacement, and social governance issues, whether regulators will evolve from external approvers into co-beneficiaries of the company's growth.

This is also the core of the disagreement between Sam Altman and U.S. Senator Bernie Sanders. OpenAI's approach is to exchange a relatively small economic stake for public sharing and political buffer. The proposal Sanders put forward in June is more aggressive, calling for a one-time 50% stock tax on large AI companies and allowing the government to influence corporate decisions through voting shares.

The 5% Discussion Stems from Pricing in Policy Risk

Frontier AI is no longer just a commercial product; it has entered a realm where the U.S. government believes preemptive intervention is necessary.

For ordinary investors, you can think of a model release like a "new drug launch." The company believes the product is ready, and the market is awaiting new features to drive subscriptions, enterprise clients, and ecosystem growth, but the government is concerned about safety testing, social impact, and national competition.

The pressure surrounding the release schedule for GPT-5.6 has already become apparent. The Associated Press reported that OpenAI, at the request of the Trump administration, restricted GPT-5.6 Sol, making it available only to approved customers. OpenAI's official statement was more tempered, describing it as a phased release requiring additional testing and coordination.

This cannot be framed as "the government blocked the product, so OpenAI exchanged equity for clearance." There is no publicly known transactional relationship between the two. However, the timeline shows that policy friction has already begun to impact product cadence, customer scope, and commercialization expectations.

OpenAI's high valuation is built on model leadership, rapid product iteration, and commercial expansion. If key model releases require waiting for administrative coordination, investors will factor in a policy discount. Conversely, if the company can incorporate the government into its revenue distribution structure, the market may reassess the magnitude of this discount.

Therefore, the market implication of the 5% plan is not a short-term boost for a specific version release. It is more like OpenAI's attempt to transform the regulatory relationship from an external constraint into a form of aligned interests. This change would affect the risk pricing for Microsoft, Nvidia, and the broader AI chain, but it is still only in the early stages of deliberation.

Altman and Sanders Are Battling Over Control

The concept of a sovereign wealth fund is not complicated. In simple terms, it means the government pools the proceeds from some public resource into an investment fund and then returns a portion to the public. OpenAI's April policy paper proposing a Public Wealth Fund was also centered on enabling citizens who do not invest in financial markets to share in the gains from AI growth.

Altman's vision views AI as a public asset that will generate enormous social returns. If leading labs contribute a small portion of equity to a public fund, ordinary people, even without buying OpenAI stock, could indirectly share in AI growth.

But equity is not the same as control. Equity can be purely an economic interest or can come with voting rights. The former is more like a dividend right, where the government receives returns but does not directly intervene in company decisions. The latter could potentially influence the board, major transactions, and corporate strategy.

Sanders' proposal is precisely aimed at control. His American AI Sovereign Wealth Fund Act centers on imposing a one-time 50% stock tax on large AI companies and placing the shares into a fund. His public explanation states the fund would be managed by an independent committee and would use voting shares to influence company decisions.

In a column on June 3, Sanders also explicitly stated that the government should acquire voting shares and have equal representation on the boards of the relevant companies. His logic is that AI will affect employment, wealth distribution, and public safety, and decisions cannot be left solely to a few tech companies.

Viewed this way, OpenAI's 5% discussion appears more like a proactive compromise. It acknowledges that the public should share in AI's dividends but seeks to avoid handing governance power directly to the government. For investors, the most critical distinction is the structure of the rights. If the 5% ultimately represents non-voting economic interests, it acts more like a cost for policy buffer. If it comes with governance rights, it represents a change in the company's control structure.

Government Stakeholding Changes the Risk Profile

The most optimistic interpretation is that once the government becomes an economic beneficiary, it will be more inclined to support the expansion of U.S. AI companies. If regulators also benefit from OpenAI's value appreciation, they might prioritize product launches, capital market paths, and global competitiveness.

But regulatory risk will not disappear. The government acting as both regulator and shareholder creates new conflicts of interest. It might relax oversight due to its ownership stake, or it might intervene more deeply in corporate decisions due to political objectives. Neither scenario aligns with pure market logic.

The idea that "government stakeholding" equals "nationalization" is also an oversimplification. Without voting rights, a 5% equity stake can hardly be termed government control. However, if the fund's design, voting arrangements, and board rights are not clarified, the market cannot treat it as a simple dividend tool either.

For related entities like Microsoft, Nvidia, Google, and Meta, the impact is not about short-term orders. A more reasonable understanding is that policy variables in the AI industry chain are becoming more prominent. In the past, investors primarily looked at computing power demand, model capabilities, cloud revenue, and capital expenditure. Now, they must also consider how leading labs handle the relationship between public returns, regulatory mandates, and national competition.

It's important to maintain a boundary here. Public reports suggest this is closer to discussions about OpenAI and Altman wanting leading labs to participate in a similar arrangement; it does not mean Anthropic, Google, or Meta have joined the negotiations. If the mechanism expands to include other companies, it could then become a new template for U.S. AI governance.

Equity Terms and Congressional Process Determine the Pricing

This discussion is still in its early stages, and the most important variables have yet to be determined. Whether the 5% is common stock, non-voting shares, or a fund interest with special terms will determine if it is a policy insurance or a governance gateway.

The congressional process is equally critical. A permanent AI sovereign wealth fund distributing returns to the national public would likely require legislative support. Without Congressional authorization, it will be difficult for the plan to evolve from discussions between the company and the government into a sustainable public financial mechanism.

Whether other AI labs follow suit will also affect market pricing. If only OpenAI participates, it looks more like a single company's political risk management. If more leading labs are incorporated into the same framework, it might become the cost of entry for U.S. frontier AI.

At this point, one cannot frame the 5% discussion as a done deal, nor equate it with the imminent clearance of GPT-5.6. It is more like an early signal: the valuation of frontier AI companies is expanding from model capabilities and computing power investment to encompass whether political pressure can be managed systematically. The real validation points are whether the equity has voting rights, whether the fund receives legislative support, and whether the company can still maintain its product release cadence.