If the Markets Reject OpenAI and Anthropic, the US Should Nationalize Them
This essay was written with Nathan E. Sanders, and originally appeared in The Guardian . OpenAI, and then Anthropic , were each formed by AI developers who feared unrestrained corporate AI development—specifically, that companies like Google and Meta would steer the technology towards deleterious, maybe even catastrophically unsafe, outcomes for society. Their founders proclaimed that their new…
When the stock markets turn away from OpenAI and Anthropic, it might be time for the United States to take a closer look at nationalizing these companies. Both were formed by AI developers who had concerns about large corporations using their technologies in ways that could harm society. While their founders believed their labs could be trusted to develop AI for the benefit of humanity, they eventually became entangled in the same market incentives that led to the rise of tech giants like Google and Meta.
Over the past few weeks, OpenAI and Anthropic have both filed for IPOs and have been met with speculation about their trillion-dollar valuations. However, there are growing worries about these companies concentrating wealth globally.
Some observers have suggested that the federal government should seize a portion of these companies' stock to create a sovereign wealth fund or distribute their revenues to taxpayers. Currently, the headlines revolve around public backlash to AI datacenters and the slumping stock of AI chip giant Nvidia. SpaceX's newly minted stock price even dropped dramatically just weeks after its IPO.
Even the leaders of ChatGPT and Claude, the AI systems behind OpenAI and Anthropic, are facing headwinds as they struggle to generate massive equity assets that were once presumed.
The market's assessment of these companies may soon determine their financial worth. The economics of the big AI labs suggest that there may not be a substantial return on investment, as the training of frontier AI models is costly and their value depreciates quickly when newer models emerge. Enterprise clients are becoming more cautious with AI token usage, and the models themselves are becoming commodities, with similar performance and pricing.
Open-source and Chinese competitors are also catching up quickly and offering free models that are comparable to those sold by Anthropic and OpenAI. Even accounting for the training costs, the unit economics of AI as it is currently conceived may not be sustainable.
While OpenAI and Anthropic have talented AI scientists and engineers, their products are doing a lot of good in the world. The question is whether the system that supports them is valuable as a market equity. If the market determines that a financial return for shareholders is not possible, the companies may collapse. Perhaps private, for-profit models are not the best way to develop AI.
OpenAI could return to its private non-profit roots, while Anthropic's founders rejected such a model. Alternatively, the organizations could be reorganized as research centers at universities, returning AI research to academia where it belongs. However, the best outcome for society might be to establish public ownership and operation of the product-oriented capabilities of OpenAI and Anthropic.
To achieve this, these companies could be divided into two parts: product innovation and compute operations. The innovation function could be managed publicly, similar to national labs in the US. Congress could provide more oversight than the venture capital these labs have recently received. The US has a long history of successful national labs that have produced world-shaping innovations in various fields, including spaceflight, telecommunications, and nuclear power.
Congress already manages a $200 billion R&D portfolio, which includes frontier AI development, a significant gap in current funding.
The compute operations could be managed like public utilities, with local or regional ownership and nationwide distribution. Strict regulation would be necessary to balance fee extraction from ratepayers with capital investment for infrastructure. Although AI datacenters are not the same as power or water treatment plants, the US has experience managing national, regional, and state supercomputing centers.
Other countries, such as Switzerland, Spain, and Singapore, already operate public AI labs, and Germany and Australia have national supercomputing centers accessible to the public. The benefits of public ownership and operation of AI would be clear, with democratic oversight ensuring these technologies serve the public interest.
Written by urgent.news from Schneier on Security's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
