London's push for a homegrown artificial intelligence champion faces an uncomfortable contradiction at its core.
The city wants greater sovereignty over its technology, meaning a domestically owned alternative to the American firms that dominate AI infrastructure and services.
Yet the ecosystem it would draw on is deeply entangled with those same firms.
The Financial Times, in its News Briefing podcast, described a renaissance in London's technology industry.
The capital already hosts the main overseas outposts of Google and Meta, the US technology giants.
It has also drawn well-funded American AI challengers, including OpenAI and Anthropic, the developers behind ChatGPT and Claude respectively.
That foreign presence is a large part of what makes London attractive, and it is also what a sovereignty drive is meant to counterbalance.
The tension is not fatal, but it is real.
A city whose AI credibility rests partly on being the preferred European base for American companies cannot easily present itself as their rival.
The talent picture shows the same double edge.
Much of London's local AI activity has been seeded by former staff of DeepMind, the AI research lab founded in the capital and now owned by Google.
Those alumni have brought expertise and investment into the city, spinning out new ventures and attracting funding.
That pipeline is a genuine strength, but it also traces back to an American-owned parent, underlining how far the local scene has grown within the orbit of the very companies it hopes to challenge.
The harder question is one of scale.
Competing at the frontier of AI demands enormous computing power, vast datasets and capital measured in billions of pounds.
The leading American developers are backed by some of the largest companies in the world and can spend accordingly.
A British champion would need comparable access to chips, data centres and long-term funding to stand a serious chance.
None of those inputs is easy to secure domestically, and several depend on supply chains the UK does not control.
There is also a strategic case for pressing ahead regardless.
Relying entirely on foreign providers for a technology expected to reshape economies and public services carries its own risks, from pricing power to data control.
A locally owned alternative, even a smaller one, would give the UK more say over the terms on which it uses AI.
The realistic path may therefore be less about matching the American giants head on and more about building capacity in areas where the UK already has depth, such as research talent and specific applications.
Whether that amounts to true sovereignty, or simply a more comfortable form of dependence, is the question London's ambition has yet to answer.
