Europe's Gigafactories and India's Different Hand
If the Gulf shows what unlimited capital buys and cannot buy, Europe and India show two other constraint profiles, and together the three sketch most of the strategic space available to nations that do not author export controls.
Europe's chief constraint is not money in absolute terms but speed, coordination, and sustained commitment. The European Union has world-class research, real capital, strong institutions, and a large internal market, yet it has chronically underinvested in frontier compute relative to the United States and China and has struggled with the slow, consensus-bound decision-making that twenty-seven member states impose. The bloc also carries a specific anxiety: it buys nearly all its AI accelerators from American firms and much of its cloud capacity from American hyperscalers, leaving it dependent on suppliers subject to another government's law and politics. That anxiety, sharpened by the broader realization that compute is strategic, produced the response most associated with Europe: a push to build large public-private 'AI gigafactories' — very large data-center-plus-compute complexes intended to give European researchers and companies access to frontier-scale training capacity on European soil, backed by substantial EU and member-state funding and coordinated at the bloc level.
Read through the four layers, the European gigafactory initiative is an attempt to secure the site and stack layers while continuing, at least initially, to import silicon. Europe is not, in the near term, trying to make its own frontier accelerators at the volumes required; it is trying to ensure that enough frontier compute physically exists in Europe, under European governance, that its AI ecosystem is not starved or wholly dependent on foreign clouds. Alongside the gigafactories sits a longer-term, more expensive bet on domestic fabrication capacity, supported by European semiconductor subsidies, aimed at reducing the silicon dependency over a much longer horizon. The honest assessment is that the gigafactory push addresses the immediate access-and-governance problem while the fabrication ambition, if it succeeds at all, addresses the deeper dependency only slowly and partially.
Europe's characteristic risk is execution. The strategy is sound on paper, but the bloc's history of announcing ambitious digital initiatives that underdeliver, combined with the speed at which the compute frontier moves, means the gigafactories must be built and filled quickly to matter — and quickly is not Europe's native tempo. The course's read is that Europe is buying insurance against being cut off or crowded out, which is rational, but that insurance only pays if the buildout keeps pace with a frontier that does not wait for consensus.
India presents a genuinely different hand. Its binding constraint is capital per capita: it cannot match Gulf sovereign wealth or American private investment on a per-project basis. But it holds three strong cards. First, scale — an enormous population, a vast pool of software and engineering talent, and a large domestic market for AI services. Second, a favorable geopolitical position: aligned enough with the United States to sit in a favorable lane of the control regime and attract chips and operators, while maintaining its own strategic autonomy. Third, a state willing to organize demand and subsidize access. India's national AI program reflects this hand: rather than attempting to out-spend anyone on raw capacity, it has focused on aggregating and subsidizing access to compute for its startups, researchers, and public institutions — effectively pooling national demand to secure GPUs and make them available at manageable cost — while nurturing domestic models, including Indian-language models, and a broad application ecosystem.
India's strategy, in the course's terms, leans on the stack and applications layers and on organized access rather than on owning the silicon or building the largest sites. It is a bet that value and a degree of sovereignty can be captured by controlling how compute is allocated and what is built on top of it, even while the chips themselves are imported. This is a shrewd allocation of a constrained budget: India spends where its money goes furthest — talent, software, demand aggregation — and imports what it cannot cheaply reproduce, using its geopolitical position to keep that import channel open.
Set the three side by side and the course thesis resolves cleanly. Each program is a distinct answer to the same environment. The Gulf answers with capital, buying site sovereignty and negotiating chip access through concessions. Europe answers with coordinated public investment, buying site-and-stack capacity as insurance while betting slowly on fabrication. India answers with organized demand and talent, buying allocation control and an application ecosystem while importing silicon on favorable terms. None achieves full independence, because the chokepoints forbid it. Each instead purchases a specific, defensible position on the dependency spectrum suited to its constraint — and each position was shaped, in the end, by lines drawn in the export-control regime. The map of sovereign compute is, at bottom, the negative image of the map of export controls.