Why a new generation of AI startups is choosing to build in London

For years, the playbook for ambitious tech founders was straightforward. If you wanted to build a global company, you went to San Francisco. But that assumption is starting to shift, particularly in AI.

A new generation of startups is choosing to build and scale from London, even as Silicon Valley continues to dominate funding and narrative. From our perspective at Fyxer, it is not a rejection of the US model, but points to a different vision and a new way of building.

Companies like Synthesia, now valued at around $4 billion, have actively chosen to build from the UK while expanding globally. It involves applying a different focus to the business as the AI landscape evolves.

“We moved from being research-led to building a product customers actually use, and then doubled down on being truly customer-focused,” says Victor Riparbelli. “For us, it’s about building the best product.”

The shift towards product development and a focus on real customer usage gives AI companies a lot more options. More and more, they are choosing not to compete at the model layer, where scale, capital, and infrastructure dominate, and instead focusing on owning the product experience and the customer relationship. If you are building into that layer, you do not need to be embedded in a specific ecosystem or tied to infrastructure. You can build from London and compete globally.

We see that opportunity starting to show up more clearly on the ground with the growth of tech companies in the UK. Of the 100 startups selected for VivaTech Top 100 Rising European Startups in 2026, 28 were based in the UK.

Talent is also a big part of it. More engineers want to work on products they can see and feel, not just small parts of a much larger system. They want ownership, speed, and the ability to see the impact of what they are building. You can see that in how companies are hiring.

This week, several London-based AI startups are hosting events to bring engineers into the room with founders. It reflects how these companies operate, close to the product and close to the user and offering autonomy and the potential to make a real mark on a product to the engineers that chose to work with them.

At the same time, exposure to the US still matters. As our CTO Matt Ffrench has pointed out internally, when he, my brother Rich, Fyxer’s CEO, and our first engineer spent time in a hacker house in San Francisco, the difference in mindset was obvious. US companies think at a much bigger scale. They were telling us to scratch out our targets and increase them 100 fold.

The self confidence and dreaming big in the US is inspiring. The best teams here are combining both the ambition and scale of the US, with a more product-focused, execution-driven approach.

We want to make the UK an absolute magnet for talent. There is real energy and optimism in the startup community, and that becomes a flywheel. It attracts more talent from around the world. That kind of belief did not exist in the same way a few years ago.

See Also

Keeping the human touch: using AI for growth

Now there are visible examples of companies choosing to stay, choosing to build, and choosing to aim much bigger.

At a macro level, the landscape is shifting too. “Although these companies are raising colossal amounts of money, it’s largely coming from the free cash flows of existing big tech companies,” Matt Clifford said in his talk at Fyxer recently. “It’s not like 2008, there isn’t this complex systemic risk.”

That dynamic is creating space for startups to experiment and grow. Large US companies are investing heavily in infrastructure, while startups can focus on application, product, and distribution.

This means you do not have to be in San Francisco to build something meaningful in AI. The question is no longer simply when to move, it is whether you need to at all. London is increasingly part of a new cohort of global AI hubs, producing companies that are not just scaling fast, but shaping how the next generation of AI businesses is built.