Europe's build decade: energy, compute, industry

The next twenty years of European value creation run through a triangle: gigawatts, sovereign compute, and re-industrialised capacity. The capital thinks in decades; the execution happens in months; and the scarcest input is people fluent in all three corners at once. This is the thesis I've been converging on — and quietly building against — for fifteen months. I made the public version of the argument in a December 2025 op-ed; this is the operator's cut.

Start from what is no longer deniable. Europe has decided — through policy, through war on its border, through the arithmetic of AI — that it must build again: generation and grid, data-center capacity it actually controls, and the industrial base that turns electrons into things — with, beneath all three, the basic materials they are made of. Each corner of that triangle is being planned separately, by different people, with different vocabularies. That separation is the opportunity, because the physics and the finance refuse to stay separate.

The triangle is one system

A data center is an energy project wearing a technology valuation. An electrolyser-fed industrial cluster is a power-market position wearing a process plant. A grid connection queue is, functionally, the region's industrial policy. Whoever holds only one vocabulary misprices the others: technologists treat power as a line item until it becomes the binding constraint; financiers treat process physics as an appendix until a mass balance breaks a business plan; industrialists treat compute as overhead until their cost curve is set by someone else's tokens-per-megawatt-hour. I have spent the last fifteen months building models where these refuse to be separated — feasibility engines that run from thermodynamics to monthly debt service, compute economics expressed as chains of physical ceilings — and the recurring discovery is always the same: the value hides at the joints.

The tempo paradox

Infrastructure capital is patient by construction — pension money, family spheres, sovereign-adjacent funds, the investors who genuinely think in generations. But the projects themselves are sprints between long waits: years for a grid connection, then a permitting window that rewards weeks; a decade of asset life, decided in a financing round that closes in a quarter. The organizational form this rewards is unusual — deliberate at the capital layer, fast and instrumented at the execution layer — and it is rare precisely because the two tempos attract different personalities. I think the institutions that internalise both will define the era, and I notice the best ones already behave this way.

What AI actually changes here

Not the megawatts. What changes is the cost of rigor at speed. The feasibility work that took an owner's-engineer team a quarter can now be a living model in weeks — if, and only if, it is built to be checked: every number carrying its source, every limit computed rather than asserted, every gap stated with what would close it. I have now built such systems to bank-grade standards, and the lesson is that the technology is the easy half; the discipline is the differentiator. Diligence-grade speed is a weapon for whoever finances and builds this decade — the ability to evaluate ten sites, ten configurations, ten capital structures with the rigor that used to afford one.

Where I stand in it

Deliberately in the middle of the triangle. Banking taught me how infrastructure gets financed and what breaks credit committees. A manufacturing-software company taught me what factories actually do all day. The last fifteen months taught me to build the computational machinery myself. I don't claim the deepest expertise in any single corner — I claim the joints, where the mispricing lives. The work shows what that looks like in practice. I expect to spend the coming decades somewhere inside this triangle — it is where everything I have learned turns out to be useful, and where the Nordic hand of cards is strongest.

— Oscar · Stockholm, July 2026