Big Ideas, Small Country of Finland

Every December, a16z publishes its list of Big Ideas, the things its partners think startups should build next year. I read this year's list twice. Once as a reader, and once as what I actually am: an engineer who has spent years helping large industrial organisations in Finland adopt AI. The second reading was more interesting, because almost every idea on the list has an invisible first line that says assume you are in America. Assume a market of 330 million people. Assume bottomless capital. Assume regulators who move slower than you do. Assume a hollowed-out industrial base waiting to be rebuilt.

Finland breaks every one of those assumptions. And the interesting thing is what happens when you break them. A few of the ideas die. Most get reshaped. A handful flip into advantages an American founder can't copy. Those are the ones I keep thinking about.

The factories never left

Take the industrial ideas first, because that's the world I work in. The American Dynamism essays are about rebuilding — bringing back the factories that left. Finland's factories never left. Our machine builders are still here, still world class, and their equipment has been instrumented for decades.

What I see from the inside is that these companies are not short of data. They're sitting on mountains of high-quality process telemetry that costs them almost nothing to produce — the exact material the rest of the world is now desperately trying to collect from scratch. The opportunity here isn't building new factories. It's making the living ones agentic: putting an AI layer on machines that already know everything about themselves.

And if you're wondering how a small AI team gets into a world like that: partnership with the incumbents, not competition. They own the customers and the steel. What they lack is exactly the thing you can bring.

Defence won't let you be casual

The same logic applies to defence, which has quietly become the most serious technology story in the Nordics. Finland is NATO's newest member, borders Russia, and last year took the overwhelming majority of Nordic defence technology funding. What makes this interesting for AI specifically is that defence is the one field where nobody gets to be casual about autonomy, reliability or trust. If you can make AI systems work under those constraints, you can make them work anywhere. The constraint is the training ground.

Regulation is the product

That points at the deeper pattern. The a16z list treats regulation as weather, something to route around. In Europe you can't route around it. AI Act, GDPR, DORA — the whole alphabet. Having spent real time on the compliance side of AI adoption, I used to see this purely as friction. I've changed my mind. A compliance layer you're forced to build at home becomes a product feature everywhere else. European customers increasingly require infrastructure and AI tooling that isn't American, that keeps data where it should be, that can prove what its agents did and why. Compliant by design isn't a tax if you can export it.

The cards nobody counts

And Finland holds some cards that surprise people. Nearly all our data centres run on renewable power, electricity is among the cheapest in Europe, and the climate does the cooling for free — which matters enormously in a world suddenly bottlenecked on compute. Our national health registries offer licensed data access no US state can match. Even the Finnish language — famously impossible — turns out to be a moat for voice and agent products: too small for the giants to prioritise, too hard to do casually, and full of clinics and municipalities that desperately need the help.

The shorter list

The pattern in all of this is the same. You don't win by running the American playbook in a country where its assumptions are false. You win by finding the ideas where the local constraint is secretly the global product. Strict privacy rules produce the privacy-preserving AI the world is starting to demand. A tiny language produces defensible niches. An old industrial base produces the data everyone else is trying to gather.

So when I read lists like a16z's now, I no longer ask which of the Big Ideas we could do too. I ask which of them we could do that Americans can't. That's a shorter list. It's also a much better one.