Impressive. A European assumes that "European governments" could release a model close to American frontier, and then argues his way back to natural slavery even in that world. "We wouldn't be able to make muh money". + China cope Euthanasia for the whole continent needed
Assume European governments release a good AI model that is only slightly worse than what is made by Anthropic or OpenAI.
What happens next?
In a new post, @pietergaricano and I describe the likely outcome: https://simongrimm.substack.com/p/is-there-a-market-for-less-capable
The model would provide temporary independence. But a model that is close to the frontier one month will be far behind it six months later. To stay a fixed distance behind the frontier, the model would need constant improving. To fund that improvement, the model would have to make money.
The market for artificial intelligence is not like a market for goods, but more like the market for talent.
In many industries, the best workers make much more than normal workers: the best football players and musicians make thousands of times the industry average; in AI research, the best engineers earn tens of millions of dollars.
Artificial intelligence is similar. Revenues are extremely concentrated in two companies, OpenAI and Anthropic. This is explained by imperfect substitution: new models can do tasks that no combination of old models can do (see GPT 5.6 solving an open Erdős problem). But new models can also do existing tasks up to a standard that weaker models can't, similar to how people can write books or compose music, but no committee could write War and Peace or come up with Bach’s cantatas.
As a result, people and companies often prefer better models. This is even true if a new model is slightly more expensive and only slightly better. Many things in business and life have nonlinear outcomes: a slightly better sales pitch can win a contract; a slightly better job application is the difference between getting and not getting a job. Across such tasks, a slightly better model is worth all of the upside it creates.
We do not think open-source models will be able to meaningfully undercut frontier companies. This is because they are worse at the most economically valuable tasks and will likely only take on easy or unimportant tasks. If cheaper fast-follower models take over such tasks, we might again end up in a world that looks like the market for knowledge work: a lot of work is routine and modestly paid, while the best work on high-value problems and make enormous returns.
Revenue concentration has enormous consequences for how governments should think about investing in frontier AI development.
Such investment is unlikely to be profitable. To justify investment, you might point to the national security benefits of such a model. But then we should be clear-eyed about the fact that it is governments who will pick up most of the tab. If you do a crash effort to build a frontier model, but end up two months behind the frontier, you will have spent hundreds of billions of dollars, without any market actors using your model. The same is true if you try to remain six months behind the frontier. Without financial returns (and without distillation, given that you'd need government funding), staying even a fixed distance behind the best models will be achievable only with exponentially growing subsidies.
Quite a bit of capital is being invested in companies that are developing or serving models behind the frontier: Mistral has raised at an €11.7 billion valuation, Baseten at $13 billion, Fireworks at $4 billion, and AMI Labs at €900 million. It is fine for private investors to take these risks. But European governments and those advising them should not assume there is a great opportunity being left on the table here. Instead, they should make life maximally easy for those private efforts.
