The Siren Song of the Golden Share: The Troubled History of Government Equity in Technology
OpenAI’s proposal to hand the U.S. government a 5 percent equity stake has made a theoretical debate over the relationship between private corporations and the government into an immediately relevant policy question. Coupled with Sen. Bernie Sanders’ (I-Vt.) proposal that the federal government acquire a 50 percent ownership stake in the top American artificial intelligence (AI) firms—a radical idea that has received an ally in President Trump—we are now at a critical moment that could have serious implications for the future competitiveness of the United States in AI development.
A recent R Street analysis laid out a definitive case for why any form of government control over AI would represent a fundamental threat to American innovation, competitiveness, and freedom of expression, regardless of whether the approach was “hard” nationalization or “soft” equity arrangements. Due to the increasing prevalence of dialogue surrounding the U.S. government taking equity in private companies, it is important to understand there is a well-documented history that accompanies these efforts, with a consistent outcome. Government ownership of transformative technology companies reliably produces stagnation, cronyism, and competitive decline—none of which are acceptable when it comes to AI development.
The main mechanism discussed surrounding proposals for AI companies is the partial equity stake. Known as the “golden share,” it is proposed as a middle path between full nationalization and genuine market independence. The historical record could not be more clear: Even minority ownership, when paired with political influence over business decisions, functions as a powerful control mechanism regardless of the equity percentage. Governments with golden shares are not passive observers. They consistently intervene and in doing so elevate political calculation above commercial and technological judgment.
Advocates for government equity will likely invoke developmentalism, an economic theory that state-directed capital allocation can build strategic industries that markets alone would not produce. Whatever the merits of that argument might be in other contexts, developmentalism has no application to the development of frontier AI models, where the qualities needed for success are speed of iteration, competition for talent, and tolerance for risk. These qualities are notably and reliably suppressed when government ownership becomes entangled with private enterprise.
The historical record shows that stagnation, cronyism, and competitive decline are the rule rather than a pattern of exceptions. The Organisation for Economic Co-operation and Development (OECD)’s research on state-owned enterprises shows chronic underperformance, political interference, and governance failures that track with the degree of government ownership across dozens of countries and sectors. OECD’s 2024 corporate governance guidelines highlight these findings. A survey of European state-directed technology enterprises spanning failed industrial champions from Concorde to Minitel and a detailed analysis of the costs that AI sovereign wealth fund proposals would impose on American technology leadership reach the same conclusions. Three cases from advanced industrialized nations illustrate this phenomenon and should dissuade U.S. policymakers from pursuing even the softest of equity arrangements.
Deutsche Telekom and the Lost Decade of Broadband
Even though Germany privatized Deutsche Telekom in 1996, the federal government retained a substantial ownership stake. This partial state ownership status, which remains to this day, presents a textbook example of how this type of arrangement distorts incentives and delays the competitive dynamism necessary for technological progress.
Through the late 1990s and into the 2000s, Deutsche Telekom was buttressed by its privileged position and implicit government backing and leveraged this support to resist infrastructure competition. Rather than aggressively deploying broadband in order to compete with rivals, the company lobbied for regulatory arrangements that protected its legacy copper network. As a result, Germany—one of the world’s largest economies and a hub of engineering excellence—consistently trailed other European competitors in broadband deployment. To see German broadband stagnate while the competitive markets in Scandinavia and other European countries surged ahead was particularly jarring, as Germany had directly linked its economy to workplace digitization.
Germany’s broadband woes did not result from a lack of capital or engineering talent at Deutsche Telekom. Instead, government ownership produced a fundamental alteration of the company’s incentive structure. With state backing, Deutsche Telekom had fewer reasons to take risks, cannibalize its own infrastructure, or accept short-term losses in favor of long-term technological leadership and more reasons to cultivate political relationships that protected their existing revenue streams. This dynamic is reliably produced by partial government ownership of private companies.
Airbus and Valuing Politics Over Engineering
Founded in the early 1970s as a joint venture between European governments, the Airbus consortium stands as perhaps the most instructive long-term experiment in state-directed technology competition. The original rationale for the project was strategic: In order to compete with American dominance in the aerospace sector, Europe needed a commercial aviation champion to rally behind. Since private markets alone had not produced a challenger, a public effort was deemed necessary. This case is instructive to the AI argument, as many of the calls for nationalization are using similar arguments. National competitiveness is framed as being inextricably linked with strategic necessity, and the technology is too important to leave to markets alone.
The Airbus case has produced mixed results, making this case study more interesting than one of utter failure. The company was able to produce competitive aircraft and challenge Boeing’s duopoly. This outcome makes Airbus a likely example for AI nationalization advocates to point to as a success; however, the full picture shows that this case is more complicated than a macro-level outcome.
Government ownership produced chronic interference in business and engineering decisions. The 2006 A380 production crisis left the program more than two years behind schedule and cost the parent company, Eastern Air Defense Sector, an initial loss of $7 billion. This crisis was a direct result of the multi-national political structure imposed on the company by the government owners. When the crisis occurred and restructuring was required, a political battle broke out between France and Germany about which nation’s workers would bear the costs. Rather than making a commercial decision, the French president and German chancellor became locked into negotiations about how to distribute job cuts between plants in either nation. Airbus was upset about political interference from the governments, with the CEO calling it “a poison” for the company.
Regardless of the industry, once a government entity holds equity in a private company, the questions of which products to develop, which markets to serve, and which capabilities to prioritize are more likely to be answered in national capitals than in board rooms or by engineering teams.
Renault and the Weaponization of Golden Shares
While the Deutsche Telekom case demonstrates how government equity slows innovation and Airbus demonstrates how it supplants engineering and business judgment, the 2015 Renault episode demonstrates how government ownership can translate into direct, limitless political control over corporate decisions.
In April 2015, the French government, having held a significant stake in the automaker Renault for decades, used its ownership position to override the company’s board of directors and management on a fundamental question of corporate governance. Renault’s board had proposed that shareholders vote to preserve a one-share-one-vote structure, a decision broadly supported by the investor community. Rather than accept this business outcome, France’s economy minister spent approximately $1.3 billion in public funds to raise the government’s stake in Renault from 15 to almost 20 percent on a temporary basis. This purchase of shares, combined with the Florange Law granting double voting rights to long-term shareholders like the French state, allowed the government to block the shareholder resolution.
The economics were never in dispute in this scenario. Renault’s leadership and the broader investor community had decided that the one-share-one-vote governance model would best serve the company’s long-term interests. The French government overrode that judgment because it would reduce its influence over political objectives related to Renault, such as employment, industrial control, and balance of power within the partnership between Renault and Nissan. France had effectively weaponized its shareholder position to override a commercial decision it found politically inconvenient.
This is the starkest version of the golden share trap in action. When the government holds the roles of regulator and shareholder at the same time, the normal checks on constraining political interference in the affairs of private firms simply do not function.
The Trump administration has already demonstrated an alarming willingness to use equity stakes in this exact manner, invoking its special stake in the Nippon Steel-U.S. Steel partnership to prevent the company from closing an unprofitable plant in Illinois. This is nearly identical to the tactics France used against Renault’s board and shareholders in 2015. Now the question becomes “How will that inappropriate pressure be used to shape AI development?” Will model outputs, content policies, or deployment decisions bend to the whim of the executive branch? The same structure is in place, but with AI development and deployment occupying a large portion of the U.S. economy, the outcomes could be far more consequential than those seen in the steel and automotive sectors.
Why AI and Its Stakes Differ
The dimension of AI that makes nationalization proposals more dangerous than the historical precedents discussed above is the nature of the technology itself. Deutsche Telekom’s broadband delays cost Germany economic efficiency. Airbus’ production politics cost shareholders money. Renault’s governance battle cost investors and stripped managerial autonomy. These costly failures affected infrastructure, hardware, and manufacturing.
AI is an information technology. Its outputs are not tangible products like fiber connectivity, aircraft, or automobiles; instead, it produces the text, assessments, and recommendations that people increasingly rely upon to form beliefs and make decisions. When the government holds a golden share in an AI company and leverages the influence they historically have provided, it shapes the contours of public knowledge itself.
The history of government control over information and communications technologies—such as radio licensing, broadcast regulation, and the managed telecommunications monopolies of the 20th century—is a bipartisan chronicle of censorship, jawboning, and the subordination of editorial independence to government preference, which R Street has documented extensively. Golden shares in AI companies would create the same arrangement for that historical interference to repeat at a scale and speed that was simply impossible with earlier information technologies.
OpenAI has framed their 5 percent proposal as generosity, sharing the financial upside of AI with the American public. The historical record shows that this framing is a siren song that leads to stagnation, political capture, and the subordination of innovation to government preference. Deutsche Telekom, Airbus, and Renault all demonstrate the hazards of following that song. Odysseus survived by having himself lashed to the mast. The United States should simply sail a different course altogether.