OpenAI and Anthropic, two prominent artificial intelligence development companies, are facing market headwinds that could challenge their long-term financial viability, leading some observers to suggest their potential nationalization if they fail as private enterprises. Both companies were founded by AI developers who expressed concerns about the unchecked development of AI by large corporations, aiming to steer the technology toward outcomes beneficial to humanity. However, they are now seen by some as having succumbed to market pressures, prioritizing investor value over public interest.
Just weeks ago, both OpenAI and Anthropic filed for initial public offerings (IPOs), with initial discussions revolving around potential trillion-dollar valuations. This generated concerns about the concentration of wealth, with proposals emerging for the federal government to acquire a stake in these companies for a US sovereign wealth fund or to distribute their revenues as a taxpayer dividend. More recently, however, the market sentiment has shifted, with reports of public opposition to AI data centers and a decline in the stock value of major AI chip manufacturer Nvidia. SpaceX, another tech and AI giant, also saw its stock price drop shortly after its IPO.
Questions are now being raised about the sustainable profitability of leading AI labs like OpenAI and Anthropic. The economics of frontier AI models present significant challenges: they are expensive to train and rapidly depreciate as newer models emerge, creating a narrow window for profit extraction. Enterprise clients are also becoming more efficient in minimizing their AI token usage. Furthermore, the models themselves are increasingly commoditized, with similar performance across leading offerings, which depresses prices.
A significant competitive factor is the rise of open-source and Chinese competitors, which are only a few months behind the leading labs in capability and offer models for free that OpenAI and Anthropic sell. Many of these free and open-source models can be run locally, on private clouds, high-end servers, or even personal devices, calling into question the substantial capital investments made by companies in data centers.
Despite these financial challenges, the organizations themselves are recognized for their value. They employ highly talented AI scientists and engineers who continuously innovate, driving significant global interest and usage of their products. The ongoing growth in usage suggests that a large number of people would be negatively impacted if these companies ceased to exist. The core issue, according to some, is not the people or the products, but the private, for-profit economic model under which AI is currently being developed.
If the market determines that these companies cannot generate a growing financial return for shareholders, they could face collapse. Alternative structures proposed include returning OpenAI to its non-profit origins, reorganizing both companies as university research centers, or, more broadly, establishing public ownership and operation of their product-oriented capabilities.
The proposal for nationalization suggests separating the companies into product innovation and compute operations. The innovation function could be publicly managed, similar to national labs, with more rigorous congressional oversight than the venture capital funding they have recently received. The US has a history of successful national labs that have driven innovations in various fields, and frontier AI development is seen as a current gap in the existing $200 billion R&D portfolio managed by Congress.
AI operations could be managed as a commodity resource, akin to public utilities for electricity or water, with local or regional ownership, nationwide distribution, and strict regulation balancing fee extraction with infrastructure investment. The US also has experience managing national, regional, and state supercomputing centers. Other countries, including Switzerland, Spain, Singapore, Germany, and Australia, already operate public AI labs and provide public access to supercomputing centers for general AI model use.
The benefits of public ownership are seen as clear: democratic oversight could make important AI models open, transparent, and responsive to public demands rather than private shareholders. These models could be aligned with democratic values, avoiding advertiser influence and training only on appropriately licensed data. The focus could shift to maximizing AI's usefulness to society rather than pursuing the goal of artificial general intelligence. Public ownership could also foster scientific cooperation over corporate competition, potentially reducing the resource and environmental costs associated with continuous, large-scale model training runs driven by investor hype.
For the companies and their employees, this shift would mean a return to their original mission of developing AI safely in the public interest, aligning with their theoretical governance structures that prioritize mission over profit. While not advocating for executive or investor golden parachutes, compensation packages for employees would be aligned with civil service standards, with those seeking higher pay potentially moving to any remaining private labs. While the long-term sustainability of these companies as private firms is questioned, the exact timeline for such a scenario remains uncertain.
