Since last Friday, the ban on Anthropic’s Mythos and Fable models has fueled debates on European digital sovereignty. Reactions are multiplying around the same concern: what would happen if the United States one day decided to cut off access to certain artificial intelligence technologies?
If the question is legitimate, it perhaps misses the point, because the most significant event in this sequence is not that Europe discovers its dependence on American laboratories, it is that American laboratories discover their own dependence on Washington.
Because if OpenAI, Anthropic or Google DeepMind are today perceived as the new centers of power in the global economy, capable of directing capital flows, redesigning industrial value chains and transforming entire sectors of economic activity, Washington’s decision reminds us of a truth that Silicon Valley had ended up forgetting: when a technology becomes a power issue, the last word no longer belongs to the engineers.
In Evian, during the G7 summit, Dario Amodei, Sam Altman and Demis Hassabis performed an unprecedented scenography, they not only defended global access to artificial intelligence, but above all the fundamental condition on which their valuations are based, namely the certainty that no one can, overnight, close the markets on which their growth depends.
Anthropic is probably just the first domino.
The major event is not that a laboratory was sanctioned, but that Washington showed that it was ready to intervene directly in the global dissemination of the most advanced models. From then on, the subject goes far beyond Anthropic, it concerns OpenAI, it concerns Google DeepMind, it concerns xAI, and tomorrow any laboratory whose growth is based on a hypothesis that has become almost invisible as it seemed so acquired: the free global circulation of models.
For three years, investors have valued artificial intelligence as a global industry. Models cross borders at the speed of light. Users number in the hundreds of millions. Addressable markets are measured on a global scale.
And for companies valued at several hundred billion dollars, this risk upsets their financial models, and not at the best time…
Why Amodei, Altman and Hassabis now speak with one voice
De facto, the G7 summit offered an unusual scene, Dario Amodei, Sam Altman and Demis Hassabis defended a remarkably convergent message. Cooperation between democracies, common standards, international coordination, sharing of cybersecurity capabilities: the leaders of the three most influential laboratories in the world have suddenly adopted the same discourse.
This union of circumstances is not trivial, anthropic and OpenAI are competing for the same customers, the same researchers and the same investments. DeepMind pursues the same technological ambition. However, against Washington, the rivalries seem to fade momentarily.
Their economic model cannot tolerate a kill switch, future AI revenues rely on a global market. The colossal investments made in data centers, GPUs and training infrastructures assume that the models can be distributed on a global scale. The day this assumption becomes uncertain, the financial equations begin to change.
Everyone will be able to understand that behind the calls for international cooperation lies a much more concrete concern: preserving the conditions which justify the current valuations of the industry.
Hyperscalers have more to lose
Another angle largely absent from current discussions is that the most exposed players are the hyperscalers.
AI is now the primary growth driver for cloud infrastructure. Microsoft is building data centers at unprecedented speed to power OpenAI. Amazon is increasing its investments in the AI capabilities of its cloud. Google justifies an increasing share of its infrastructure spending through model training and inference.
Behind each model lies a much larger value chain, in which hundreds of billions of dollars are committed based on a belief: global demand for AI will continue to grow without major impediment.
But if models become subject to regimes of authorization, control or geopolitical restriction, this entire mechanics could be affected. Laboratories sell models, while hyperscalers sell the highway on which these models travel.
The economic impact could be catastrophic for an entire ecosystem, which more than a bubble, resembles
Silicon Valley discovers the risk it itself helped create
The irony is worth noting that for more than a decade, major US technology companies have supported the idea that certain technologies should be protected for reasons of national security.
Huawei has been gradually excluded from many Western markets. Exports of advanced semiconductors to China have been restricted. Critical infrastructure has become a strategic subject, export controls have multiplied, with the aim of preserving American technological advance.
Today, the same mechanisms are imposed on the American champions themselves. Silicon Valley is discovering that it is no longer completely in control of its destiny and that Donald Trump is ready to increase the pressure.
