Geopolitics and Development#
Contents
The claim I'm most confident about, and the one that gets least attention.
The development ladder is being kicked away at both rungs#
The post-1980 growth model runs: cheap labor → export manufacturing → services/BPO → domestic market.
Both middle rungs are under simultaneous attack:
- AI eats the BPO/IT-services rung - now, in this decade.
- Robotics eventually eats the manufacturing rung and enables reshoring - 2030s, per Robotics.
This has never happened before. Previous technology waves moved the ladder or raised the height of each rung; they did not remove two adjacent rungs within fifteen years of each other. Every country currently mid-climb is affected, and the ones furthest along are most exposed because they specialized hardest.
Three axes, not one#
Public discussion treats this as a single US–China race. It is at least three separate contests running on different clocks, with different winners, and they decouple:
| Axis | Chokepoint | Who leads | Decays by |
|---|---|---|---|
| Capability - best models | Talent, capital, accelerators | US, with China closing | Diffusion: 30–50%/yr |
| Capacity - deployed compute | Power, land, permitting | Contested; China and the Gulf structurally advantaged | Doesn't decay - it's built |
| Application - economic absorption | Institutions, firms, skills | Genuinely open | Slowest of the three |
The decoupling is the point. A country can lead on capability and lose on capacity, as Energy argues the US may. A country can buy capacity and fail to absorb it, which is the Gulf's risk. And absorption - the axis that actually determines living standards - correlates poorly with either of the other two.
Most analysis conflates all three into "who is winning," which is why most analysis is unhelpful.
Absorption is the slow axis for a reason that is not going to change. Capability is bought, capacity is built, and absorption has to be institutionalized - which means changing how firms are organized, what workers are trained for, what regulators certify, and what liability attaches to a decision made with machine assistance. Those are the inelastic complements seen from the national rather than the firm level, and they move on the timescale of professional generations rather than product cycles. The historical base rate is unambiguous: electrification took about four decades from viable to productivity-visible, and the delay was not the dynamos but the rebuilding of factories around them. Failure mode for this whole framing: if the current wave is absorbed mainly through software that firms already buy, rather than through reorganization, absorption could run far faster than the electrification analogy suggests and the third axis stops being the bottleneck. Watch whether measured productivity gains show up in firms that reorganized or in firms that merely subscribed.
Open-weight lag is the coupling constant#
Measured lag ~3–6 months (Epoch AI, early 2026) means capability axes stay closer than export-control rhetoric implies. Capacity and absorption still diverge. Policy that treats weights as containable fights the measured constant; policy that contests power, fabs, and institutions is aligned with the three-axis table. → bipolar, Game 1
Why the ladder is defended rather than replaced#
The pivot below is correct and it is also, in almost every country, politically unavailable, and it is worth being precise about the mechanism rather than blaming foresight.
The exposed sector is organized. It has employer associations, export-promotion agencies, dedicated ministries, tax regimes written for it, and a workforce concentrated in a handful of cities that vote. The beneficiaries of the pivot - future workers in domestic-demand sectors that do not yet exist, and firms not yet founded - have no association, no ministry, and no votes. This is the standard asymmetry of concentrated versus diffuse interests, and it reliably produces the same policy: subsidize the incumbent sector to defend the margin, which prolongs the specialization that created the exposure.
The second mechanism is that the arbitrage keeps paying while it dies. A wasting position is still a position, and the revenue it generates is the fiscal base that would have to fund the pivot. Countries therefore face the pivot decision at exactly the moment when the case for it looks weakest - the sector is still growing, the compression shows up first in headcount rather than revenue, and the crisis is legible only in retrospect. The signal a government would need in order to act early is a decoupling of sector revenue from sector employment, which is publicly reported and almost never treated as a macroeconomic warning. → India
The regions#
Each faces a different binding constraint. That is the organizing insight of this section, and it is why one page could not carry it.
| Binding constraint | The question | |
|---|---|---|
| United States and China | Each other, plus their own bottlenecks - permits for one, chips for the other | Does the leaky-bucket lead hold? |
| India | The exposed export sector is the growth model | Can 10M workers/year be absorbed elsewhere? |
| Europe | Regulatory capacity outrunning industrial capacity | Can it be a rule-setter without being a producer? |
| The Gulf | Nothing physical - capital and power are abundant | Can compute be converted into an economy? |
| The Global South | The complements: connectivity, credit, electricity, tenure | Does the free technology reach anyone? |
What working within it looks like#
The countries that come through this will be the ones that pivot early to:
- Domestic demand - growth that does not depend on selling labor cost to foreigners
- Energy production - the binding global constraint (Energy), and one where geography and permitting speed matter more than capital
- Data sovereignty - owning the flow data that Game 3 identifies as inelastic
Rather than defending the arbitrage, which is a wasting position and which political economy will make very tempting to defend.
The pivot is expensive and unpopular and has to begin before the crisis is legible. That is the hard part, and it is why I expect most countries not to make it.
One qualification, because the pessimism above can be overdone. A country does not have to run the pivot deliberately to end up with it. Domestic demand grows on its own as incomes rise, energy build-out is being driven by decarbonization commitments that were made for unrelated reasons, and data sovereignty is arriving through privacy politics rather than industrial strategy. Several countries will end up holding two of the three legs by accident. The argument here is not that the pivot requires unusual wisdom, only that the timing does: arriving at domestic-demand-led growth in the 2040s is a different outcome from arriving in the early 2030s, and the difference is one labor-force cohort that gets absorbed or does not.
Related: Energy on who can build capacity · Robotics · Game 2 - Nations · Assets