A Durable Checklist for Programs Not Yet Announced

The specific programs in this course will age. New sovereign initiatives will launch, some named here will be restructured or quietly shelved, announced capacities will be revised, and the export-control regime itself will shift with each administration and each round of vendor workarounds. What should not age is the analytical apparatus. This final lesson consolidates the course into a durable method you can apply to any sovereign-compute program, including ones that do not yet exist.

Start with the three framing questions, which order everything else. First: which lane of the export-control regime does this nation occupy? Is it the author of controls, the primary target, a favored ally with broad access, or a middle-tier state whose access is licensed and conditional? The lane determines the menu of deals available and therefore bounds what the program can possibly be. A middle-tier state cannot simply buy unlimited frontier chips; a favored ally largely can; a target must attempt indigenization. Locate the nation on this map first, because it constrains every subsequent answer.

Second: which of the four layers — silicon, site, stack, people — is the program actually acquiring control over, and which is it leaving foreign? Resist the announcement's framing and trace each layer to its real source. Almost every program controls the site layer, since that is what capital and land buy. The meaningful variation is in the other three. Does the program invest in domestic or allied silicon, or import it entirely? Does it build local systems integration, operational software, and home-grown models, or license a foreign stack wholesale? Does it develop domestic operators who hold administrative control, or contract that out to firms reachable by another government's law? The pattern of answers is the program's true sovereignty profile, far more informative than any capacity figure.

Third: how many independent hands could sever the nation's access? Count them literally. A chip export license that another government could decline. A vendor that could be compelled to withhold updates or support. An operator subject to foreign legal reach. A single foundry at the far end of the supply chain, in a geopolitically exposed location, that everything ultimately depends on. Each is a hand on the switch. The more hands, and the more of them lie outside the nation's influence, the more brittle the sovereignty, regardless of how much capacity has been built. A program with magnificent facilities but four foreign hands on the switch is less sovereign than a modest one with two.

Layer onto these three questions the capital and vendor analysis from the previous lesson. Identify the financing pool — sovereign wealth, public budget, or private capital — and check it against the stated goal, watching for the mismatch that signals a program more marketing than substance. Map the cross-investment web to see whose interests are genuinely aligned and whose leverage is embedded. Trace the vendor and operator of each layer. Weigh the demand assumptions the financing rests on, and ask whether the capacity could prove overbuilt if the frontier or demand stalls.

With these habits, apply the course thesis as a predictive lens rather than a historical observation. Export controls reshape the global compute map — which means that when the controls change, the map will change, and you can anticipate the direction. Tighten access for a middle-tier group and expect those nations to accelerate accommodation deals, bring in more allied partners, and offer more security concessions to preserve access; expect some to hedge harder toward domestic capacity. Loosen access for an ally and expect private capital to rush in and sovereign motives to recede, since the supply-denial fear that justified state intervention eases. Tighten against a target further and expect deeper, costlier indigenization and a widening bifurcation between compute blocs. The controls are the independent variable; the programs are the dependent response. Watch the former to predict the latter.

Finally, hold the whole picture at the right altitude. The enduring reality is that genuine end-to-end compute independence is available to almost no nation on a relevant timescale, because the chokepoints in silicon design, fabrication, and lithography are real, narrow, and enormously expensive to reproduce. Every sovereign program is therefore, whatever its rhetoric, a negotiation of dependency rather than an escape from it. The serious analytical questions are always comparative and gradable: which layers, whose hands, how durable, at what cost, financed from where, and how exposed to a change in the rules. A country does not have or lack sovereign compute; it holds a specific, describable position on a spectrum of control, and that position was shaped decisively by lines drawn at the chokepoint. Master that framing and you can read any program — this year's or next decade's — discount its press release, and state in a sentence what kind of sovereignty is really on offer and how many hands could take it away.