Most of the data center news you read is about scale. Gigawatt campuses, billion euro investments, facilities the size of small towns. Reading it, you could conclude that the future of compute belongs entirely to a handful of giants, and that building anything smaller is a rounding error.
We are building something smaller. Here is why we think that still makes sense, and for whom.
Hyperscale is an answer, not the answer
Hyperscale facilities are extraordinary at what they do: enormous, standardized capacity for customers who need exactly that. But standardization cuts both ways. If your workload, your hardware, or your timeline does not fit the standard shape, you become a small fish in a very large pond. Your questions go into a queue. Your special requirements become a procurement exercise.
A meaningful share of compute buyers do not need a gigawatt. They need a few racks, or a few dozen, run well, by people they can reach.
Small means the operator knows your name
The practical difference between a small facility and a giant one is not the physics. Power is power, cooling is cooling. The difference is organizational. In a small facility, the person who answers your email has probably stood next to your hardware. There is no ticket routing system between you and someone who can act, because there is no room for one.
We have written this on our website and we mean it literally: we are small enough that you will actually talk to the people running it. That is not a compromise we are making until we get big. It is the product.
Focus beats generality at this scale
A small facility cannot be everything for everyone, so it should not try. Ours is built for compute heavy work: AI training, HPC, workloads that push hardware hard, plus colocation for companies that want a calm, well run home for their machines. Designing for a narrow purpose means every decision, from cooling to floor layout, serves that purpose instead of averaging across use cases.
The honest trade-offs
Small operators cannot offer thirty global regions or instant capacity in any city. If your architecture needs that, you need a hyperscaler, and you should use one. Many architectures do not. Training runs, research computing, rendering, national or EU bound workloads, and infrastructure that simply needs to live somewhere reliable and affordable: these run happily in one good facility in one good location.
The market seems to agree that there is room for both. The giants keep growing, and at the same time, focused regional facilities keep getting built, because some buyers keep choosing them.
We are betting on those buyers. If you are one of them, you know where to find us, and when you write, a human who works in the building will write back.
Read next
Why Finland for compute · What ISO 27001 actually means for your data
Lumilogic is building a compute focused data center in Riihimäki, Finland, with capacity coming online in early 2027.