The article discusses how market failures lead to underinvestment in critical AI applications like medical research, and proposes state intervention through mechanisms such as demand contracts and public compute to address this imbalance.
Before anyone gets annoyed by the title, this is not a post saying nationalise OpenAI. It is about one thing the market is visibly failing at right now and one thing socialists have always been right about, and they are the same thing. Some numbers first because otherwise this is just vibes. Amazon, Microsoft, Alphabet and Meta have guided somewhere between 720 and 745 billion dollars of capex for 2026. Nearly all of it AI infrastructure. The entire US federal R&D budget across every agency including defence was about 192 billion last year. NIH is 47 billion. So four companies are spending roughly 15x the world's biggest medical research funder on one technology in one year. AI companies took 61 percent of all global venture capital in 2025 per the OECD. In Q1 2026 it was around 80 percent per Crunchbase. Four rounds (OpenAI, Anthropic, xAI, Waymo) were 65 percent of every venture dollar on the planet that quarter. Now look at where that money does not go. The WHO's 2025 pipeline review found 90 antibacterial agents in clinical development, down from 97 in 2023. 15 are innovative. 5 work against a critical priority pathogen. This is the drug class where resistance is already killing people and the pipeline is shrinking. The thing i find interesting is that nobody is being greedy or stupid here. There is a 2015 paper in the American Economic Review (Budish, Roin and Williams) that looks at cancer clinical trials and shows private research systematically avoids projects with long commercialisation periods. Prevention and early stage trials take years longer to prove out than late stage trials so they get less money. Not a correlation, they identify it properly. The length of the feedback loop alone changes what gets funded. Same logic applies to AI. A writing assistant has a customer who pays next month and feedback in days. A diagnostic model for rural hospitals has a customer who cannot pay, a feedback loop in years and regulators on top. Capital is water, it runs downhill, and it will pick the writing assistant every time regardless of which one matters more. So here is where the socialists are right. Someone other than the market has to decide that certain problems get worked on. Every big example of this working was basically that. Apollo employed 400,000 people on a problem with no consumer market. The Human Genome Project cost 2.7 billion and dumped the data into the public domain and the sequencing cost curve fell off a cliff afterwards. There is a 2023 AER paper (Gross and Sampat) showing wartime R&D created tech clusters that were still producing companies in the 1970s. And here is where they are wrong. None of those needed the state to build the thing. It needed the state to be the customer. Pay for the outcome before it exists, let private teams compete, publish the result. The UK is already running this for antibiotics, an annual subscription for access to the drug instead of paying per dose, which is the exact fix for the antibiotic incentive problem. The five things that would actually move talent, in my view : guaranteed demand contracts, prizes paid on outcome, patient sovereign capital, public compute that university researchers can actually reserve, and for countries like India, stop building the fourth best chatbot and pick two problems where you have data nobody else has. Genuinely curious if people here think there is a version of this that does not end up as a prestige project or pork. The paper has a limitations section that i think is honest about that and everything but happy to be told i'm wrong. Disclosure : i wrote the paper this is based on. Not selling anything, it's free. Link in a comment so this post stands on its own.
The article argues that the real threat from AI is not the technology itself but capitalism's inability to distribute the wealth created by automation, and calls for universal basic income funded by AI productivity to prevent economic collapse.
Nathan Sanders and Bruce Schneier's essay in Tech Policy Press argues that AI fears often confuse technological flaws with capitalist structures, using medical AI to show how market incentives shape outcomes and contrasting different economic models.
A detailed critique of Bernie Sanders' proposed AI Wealth Fund, arguing that a one-time equity grab is the wrong mechanism, and offering an alternative that funds public AI benefits through data center chokepoints, infrastructure taxes, a Santiago Principles-aligned sovereign wealth fund, and dedicating a slice of compute to public institutions.
The article explores how AI's rapid advancement is concentrating wealth and diminishing labor's value, arguing that traditional economics cannot address the resulting inequality, and proposes that crypto networks based on community and redistribution may offer a solution in a post-AI future.
In this op-ed, Senator Bernie Sanders argues that because AI is built on collective human knowledge, the public should own half of major AI companies through a sovereign wealth fund, ensuring democratic control and that profits benefit society.