The new geography of compute
Energy, chips, networks and sovereignty are turning data centres into geopolitical infrastructure.
Compute is often discussed as a cloud abstraction. At frontier scale it is physical: power generation, cooling, construction, supply chains, fibre and political permission concentrated in particular places.
The shift
The surface-level story is speed: more output, shorter cycles, lower production cost. The structural story is that the location of scarcity moves. Once an old constraint becomes cheap, the system reorganises around what is still difficult to obtain—context, judgement, authority, attention or trust.
That distinction matters because markets routinely overvalue the newly abundant thing during a transition. Capability is visible. Integration is quiet. Yet integration is where a possibility becomes a repeated behaviour and, eventually, an institution.
Where the leverage moves
The geography matters because latency, regulation, resilience and access are not evenly distributed. Nations and companies are beginning to treat compute capacity as strategic inventory.
The practical response is not to chase every new capability. It is to identify which part of the value chain now moves faster, which part becomes newly defensible, and what kind of product can make that shift legible to a user. A good product is a theory of the new constraint made concrete.
The model to hold: capability diffuses; situated judgement compounds.
What I’m watching
Follow grid interconnection queues and long-term energy contracts as closely as chip announcements. The bottleneck is widening beyond silicon.
The most useful signals will not always arrive as announcements. They appear in changed pricing units, shrinking interfaces, new permission patterns and teams achieving outcomes that their organisational shape should not permit. Those are the moments when the future stops being a claim and starts becoming an operating fact.