The number one rule of real estate is location, location, location. For deep tech startups, that rule applies even more so.
Software companies have spent two decades dissolving the relationship between value creation and physical place. The internet turned location into a lifestyle choice. For deep tech, where you build is often dictated by the physics of what you’re building. A compound semiconductor startup needs proximity to epitaxy facilities and III-V wafer suppliers; perhaps five places globally have that supply chain. A solid-state battery startup needs dry rooms and electrode coating lines that exist at pilot scale in a handful of national facilities, not university labs. Of course, not all deep tech is infrastructure-bound. This binds less where software does the scaling e.g. AI, quantum software stacks, computational drug discovery. But for the founders whose work lives or dies by access to physical infrastructure, and that remains a large share of deep tech, geography is a constraint, not a preference.
Geography, however, doesn’t bind uniformly across a company’s lifecycle. It shifts.
At the early stage, the structural limit is lab access. You need to be near shared infrastructure (cleanrooms, synchrotron beamlines, pilot reactors) because you can’t afford to own it. This is why national labs, university research parks, and catapult centres exert such gravitational pull.
At the mid-stage, the constraint rotates to pilot manufacturing. The question is no longer “can I prove this works?” but “can I prove it works at production conditions?”
Late stage, it rotates again to customer proximity, regulators, and capital. Capital markets introduce their own form of geographic density, but that is a separate point and one worth unpacking on its own.
It’s natural for founders to build where their research began: near the lab, the co-founders, the university that incubated the science. In the early stages, that’s often exactly right. The risk isn’t the starting point; it’s treating it as permanent. The geography that produced the science and the geography that can scale it are often different places. Founders who recognise when the constraint has shifted move faster than those who stay out of habit.
The smartest founders don’t choose a single location. They plan a sequence - moving as the underlying substrate shifts from lab access to manufacturing depth to customer proximity - or they anchor themselves in the rare geography dense enough to carry them from lab to scale without forcing a move.
Some founders attempt to bypass this entirely, running split-stack models: R&D in one country, manufacturing in another, and sales in a third. They stitch together ecosystems rather than embedding in one. This doesn’t eliminate the bottlenecks. Rather it distributes them, with real costs in coordination overhead and IP fragmentation. But it reflects a maturing understanding of the practical truth that no single geography may serve every phase of a deep tech company’s lifecycle.
All of this assumes founders are choosing geographies based on physics and commercial logic. Increasingly, they are choosing among incentive regimes as well.
There is a second force reshaping the geography of deep tech: sovereignty. If you are a founder in 2026, you are almost certainly navigating a national strategy designed to anchor strategic capability domestically. Governments are engineering clusters through industrial policy - aligning capital, regulation, infrastructure, and demand in one coordinated wave - to make that possible.
This effort does not land on a blank slate. In geographies with legacy infrastructure, policy compounds decades of accumulated density: supplier networks, tacit knowledge, regulatory maturity, and university-industry loops. That depth reduces execution risk because the connective tissue already exists but legacy systems also reflect earlier technological eras and can carry inertia. Emerging clusters start from a different position: able to deliberately align infrastructure, regulation, and capital around the frontier from day one, unencumbered by inherited assumptions.
So what does this mean for founders?
The question is not “where should I build?” That framing is too static. The better question is: what is my dominant constraint for the next 24 months and where is that constraint already someone else’s routine problem?
If your science requires a synchrotron, you need to be near a synchrotron. If your next bottleneck is a manufacturing partner who understands your tolerances, you need to be near that partner. If your gating factor is regulatory approval within a survivable timeline, you need proximity to that regulator. The best geography is the one in which your hardest problem is already normalised; where the surrounding system has seen your problem before, where failure has been absorbed into institutional memory, and where iteration is part of the environment.
Every estate agent knows you’re not buying a house, you’re buying the neighbourhood. In deep tech, it’s the same. Before you sign the lease, look past the building and inspect the street.
-RF