Skip to the lesson

National Compute Strategies & Sovereign Programs

Glossary

AI gigafactory
The European term for very large public-private data-center-plus-compute complexes intended to give European researchers and firms access to frontier-scale training capacity on European soil under European governance, backed by EU and member-state funding. Europe's answer to importing nearly all its accelerators and cloud capacity from American firms.
Chokepoint
A narrow point in the supply chain controlled by one or very few suppliers — chip design firms, design software, or lithography equipment — through which a single decision can constrain everyone downstream. US leverage rests on controlling several chokepoints rather than on manufacturing chips itself.
Export controls
Government restrictions on the sale of advanced AI chips and the equipment to make them, imposed by the United States from October 2022 onward, initially targeting China and later expanding into a tiered global framework. The course's central force: they convert compute from a commodity into a geopolitically gated resource.
Foreign direct product rule
A legal mechanism that extends US export jurisdiction to products made outside the United States if they were built using American technology or software. It is how a chip fabricated in Taiwan with Dutch equipment can still fall under American export control, making the chokepoint legal as well as technological.
Negotiated dependency
The course's characterization of what most sovereign programs actually achieve: not independence from foreign compute suppliers but a deliberately structured arrangement — a deal, concession, or partnership — that secures enough access and enough local control to satisfy a nation's particular blend of motives on tolerable, durable terms.
Site-layer sovereignty
Control over the physical facility — land, power, buildings, and data residency — while the silicon and much of the software stack remain foreign. Often what a 'sovereign cloud' actually offers: sufficient for data-governance concerns, but not for protection against having access to compute severed at the chip supply.
Sovereign compute
A nation's ability to access and control AI computing capacity under its own law and authority. In practice a spectrum rather than a binary, measured by how many of four layers — silicon, site, stack, and people — a country actually controls versus depends on foreign suppliers for.
The four layers
The course's framework for decomposing compute sovereignty: silicon (the accelerator chips), the site (land, power, buildings, and network on national soil), the stack (firmware, drivers, orchestration software, and model weights), and the people (operators who can run the system independently). Almost no nation controls all four.