X

Dear SoTA,

Earlier this month we announced our investment in OLIX, James Dacombe’s chip company. He started it in 2024.

OLIX builds a different piece of silicon for each stage of inference and connects them with light, so a model unrolls across a rack the way a car moves down an assembly line. The first chip, DX-1, does decode. It holds the model in fast on-chip memory, which means it needs no high-bandwidth memory and no advanced packaging. Those are also the two things the entire industry is currently queueing for.

It’s exactly the kind of company a fund like ours exists to back. So this next part is going to sound a little strange. I help run a fund with the word sovereign in the title, and my advice to founders is this: don’t just build for Britain. Build the thing you can’t stop thinking about.

Sovereignty matters. This fund is the government agreeing that it matters. But it’s my job to worry about it, not yours. That’s literally what SovAI is for: we back great founders so the country ends up with sovereign AI capability, without founders having to carry that weight themselves. So your job is to build the thing you can’t stop thinking about. Mine is to make sure enough of that happens here.

Some great companies here do put ‘sovereign’ in their branding, and that’s fine. For them the word is a spec, like on-prem or air-gapped. If you sell to defence departments, they need their data on servers the UK controls, and the flag is doing real work. But notice that even those companies lead with benchmarks. The decks that worry me are the other kind, where the flag is doing all the work, and the word British is sitting in the space where the word best should be.

Somebody does have to think about this, and James certainly does. His holding company, Bletchley Industries, published a letter in March that puts it well: “Rebuilding sovereign industrial capacity is not only an economic opportunity but a practical necessity, and it is work that requires patient capital and companies capable of coordinating complex manufacturing and engineering at scale.” All true. But look at who’s saying it and when.

That’s a holding company talking, and James can write it because he spent the seven years before it building the things it describes.

That’s the order it happens in. You do the work first, and you earn the right to talk about it later. Since I’m Dutch I’m contractually obliged to mention ASML, which makes 90% of the world’s advanced lithography machines and is the reason every country that wants cutting-edge chips ends up knocking on a Dutch door. People tell that story as a strategy. I don’t think it was, at least not at the start. What actually happened is that a few hundred engineers stayed interested in one narrow problem for much longer than anyone thought reasonable, and the geopolitics showed up at the end.

Everyone knows startups take years, though nobody really tells you what the years are made of, and that is because a lot of it is not that glamorous. It’s often months where nothing visible happens, and the only reason to keep going is that the problem is still interesting. A sovereign strategy won’t get you through those weeks. Patriotism probably won’t either. And neither will a supportive minister, and we have great ones. (As I write this, our Kanishka Narayan has just become the first AI minister to sit in Cabinet, which is great news for the country.)

What gets you through is an unwavering obsession. James started CoMind as a teenager because people in his family had strokes and he wanted to see inside a brain in real time. That’s an obsession.

He’s not the only one. We are building an incredible portfolio which, alongside OLIX, now counts Callosum, Ineffable Labs, Isomorphic, and CuspAI. All of these founders are unapologetically obsessed with what they’re building. Their firms are committed to Britain too, and that’s fantastic. But the quality of their ideas transcends borders. That doesn’t mean Britain is irrelevant. Quite the opposite. The reason funds like ours exist is to help ensure that when world-class companies emerge, more of the talent, expertise, intellectual property and economic value stays here. Britain benefits most when the companies shaping the future are built and scaled from Britain.

Founders can afford to mostly ignore sovereignty and remain obsessively focused on their great ideas. The question of sovereignty is ultimately one for government, funds like ours, and the occasional patient holding company. It’s a property of a whole system: where the capital sits, where the talent stays, which layers of the stack somebody here actually owns. Full independence is off the menu for everyone, superpowers included. Someone asked me about this recently and I said that the AI supply chain is the greatest group project the world has ever seen. What a country can get is leverage, some place in the stack the rest of the world has to route through.

Share

So, no, don’t just build for Britain. Build the thing you can’t stop thinking about. The founders I admire most would be obsessed with the same problem wherever they were born. What matters to us is that they choose to build, grow and create value here in Britain. If that’s you, and you’re building here, then come and talk to us.

Yours,

Joséphine Kant

Author biography:

Joséphine Kant is the Head of Ventures at the UK’s Sovereign AI Fund, a £500m government fund to back British AI companies. Her background spans early-stage startups, big tech, and venture investing.

Share

Write to the Society for Technological Advancement on letters@ilikethefuture.com.