I wrote this to track how AI is reshaping the balance of power — the fight for compute and chips, military applications, economic statecraft, alliances, and the struggle to govern any of it. It is my working map of the contest, not a prediction of where it ends.
AI has become the central variable in great-power competition. National power increasingly rests on a stack of compute, chips, data, energy, talent, and alliances — and whoever leads it gains military, economic, and diplomatic advantage. The contest is pushing the world from a single integrated system toward rival technology blocs. The question I keep weighing: a managed bipolar rivalry, a governed multipolar order, or a fragmented and dangerous race?
Compute and chips are the new strategic resources — like oil and steel in earlier eras. Leadership translates into military edge, economic leverage, and diplomatic weight.
The US–China competition for AI supremacy is the organizing axis. Export controls, supply-chain decoupling, and a "silicon curtain" are reshaping global alignments.
Technology denial and sovereignty drives pull the world toward rival blocs and a splinternet — even as shared risks demand cooperation no single power can provide.
What I keep coming back to: AI is doing to geopolitics what it does to everything — amplifying existing dynamics at speed. It sharpens the US–China rivalry, raises the stakes of the Taiwan chokepoint, accelerates the militarization of statecraft, and compresses decision time in crises. The defining danger, to me, is not any single capability but an unmanaged race: arms-race instability, miscalculation at machine speed, and a world split into hostile, non-interoperable blocs. The defining opportunity is strategic stability — guardrails, deterrence, and enough cooperation to keep competition from becoming catastrophe.
A small set of strategic inputs (compute, chips, energy, data, talent, capital) is converted into instruments of power — military, economic, informational, and standard-setting — wielded across the arenas of alliances, the Global South, and institutions, producing the balance of power and the type of world order. Feedback loops either consolidate blocs or diffuse power.
Two superpowers anchor the contest; a set of capable middle powers hedge and bargain; and the Global South is the prize both poles court. Where each actor sits — and how it moves — defines the emerging order.
| Power / Bloc | AI weight | Core strength | Key vulnerability | Strategic posture |
|---|---|---|---|---|
| United States | Pole | Frontier labs, chip design, capital, alliances, compute | Fab dependence (Taiwan), political volatility | Lead & deny: export controls + build the allied stack |
| China | Pole | Scale, data, manufacturing, state mobilization, applications | Advanced-chip access; equipment chokepoints | Self-reliance + indigenize; export infrastructure |
| European Union | Power | Market size, regulation, industry, research | Limited frontier labs & compute; fragmentation | Regulate & seek "strategic autonomy" |
| Taiwan | Pivotal | ~90% of leading-edge fabrication (TSMC) | Existential security exposure; the flashpoint | "Silicon shield"; deepen indispensability |
| Gulf (UAE, KSA) | Rising | Capital, energy, sovereign-compute build-out | Caught between US tech & China ties | Become compute hubs; hedge between poles |
| India | Rising | Talent, large market, digital public infrastructure | Compute & frontier-capability gaps | Strategic autonomy; multi-alignment |
| Japan / Korea | Power | Materials, equipment, memory, advanced manufacturing | Demographics; security dependence on US | Allied tech bloc; supply-chain nodes |
| Russia | Limited | Military application, info ops, resources | Sanctioned, chip-starved AI base | Asymmetric & disruptive; lean on China |
| Global South | Courted | Markets, minerals, demographics, swing votes | Compute, infrastructure & capability gaps | Non-alignment; extract concessions from both |
Compute is the strategic high ground of the AI era, and the supply chain that produces it is astonishingly concentrated. Control of its chokepoints — by a handful of firms and countries — has become the primary instrument of technology statecraft.
Since 2022, the US and allies have restricted advanced chips, equipment, and HBM to slow China — repeatedly tightened, and met by Chinese stockpiling and indigenization.
Concentrated leading-edge fabrication makes Taiwan the most consequential — and most dangerous — node in global geopolitics. A disruption there would be a global shock.
Efficient Chinese models in early 2025 showed capability can be reached with less compute than assumed — challenging the idea that controls alone secure a durable lead.
The US, EU, Japan, and others subsidize domestic and allied fabrication to reduce single-point dependence — slow, costly, and incomplete.
Massive data-center programs (US hyperscalers, Gulf sovereign compute) turn capital and energy into strategic capability — and new dependencies.
China's dominance in rare earths and key materials is a counter-lever — export curbs on gallium, germanium, and magnets answer chip controls.
The competition has moved from open globalization to active technology denial. As controls tighten and both poles build parallel stacks, the world drifts from one integrated system toward two — with the middle phases the most unstable.
Generative AI breaks out; the US launches sweeping chip controls; "AI as national security" becomes consensus. The race is declared.
Parallel stacks form. The 2025 US pivot exports the "American stack" to allies; China indigenizes; Gulf compute hubs rise. The silicon curtain thickens.
Autonomy and AI-enabled C2 mature. Crisis-stability and escalation risks rise. Pressure builds for arms-control measures — and for breakthroughs.
A bipolar or multipolar order settles. Either managed rivalry with guardrails — or a hardened, hostile two-bloc world. Path-dependent.
AI is moving from the lab into the battlefield — in surveillance, autonomy, cyber, and command. Its greatest strategic effect is on the speed of war: compressing decision cycles in ways that strengthen deterrence in some scenarios and dangerously raise escalation risk in others.
Ukraine became the proving ground for cheap autonomy and rapid software iteration — reshaping doctrine and the economics of force.
Lethal autonomous systems advance faster than the stalled UN talks meant to govern them — a widening law-and-norms gap.
AI compresses decision time, raising the risk of miscalculation at machine speed and inadvertent escalation in a crisis.
Programs to field attritable autonomous mass (e.g. "Replicator"-style efforts) and AI-enabled targeting are moving from pilot to scale.
AI redraws the map of what matters. The strategic terrain is no longer just oil and sea lanes — it is fabs, data centers, energy, minerals, and the cables that connect them. Controlling these assets is the new economic statecraft.
The interdependence paradox: the same global integration that creates shared prosperity also creates weaponizable chokepoints. Export controls, sanctions, and infrastructure deals turn economic ties into instruments of coercion — what scholars call "weaponized interdependence." Expect more strategic decoupling in critical layers, even as full separation remains too costly for either pole.
The poles are assembling rival ecosystems — and competing hardest for the uncommitted majority. The Global South is not a passive prize: it holds minerals, markets, demographics, and votes, and is learning to bargain across the divide.
India, the Gulf, Indonesia, Brazil, and others deliberately court both poles — buying US chips while keeping Chinese ties — to maximize options and avoid capture.
States build domestic compute, local-language models, and data residency to reduce dependence on either pole — a third force resisting bipolar capture.
Resource-rich and demographically young states hold real bargaining chips — minerals, markets, and the swing votes that shape global rules.
Unlike nuclear weapons, AI has no binding international regime — only a thin, fast-evolving layer of summits, soft norms, and bilateral dialogue. Closing the gap between the speed of the technology and the pace of governance is the era's central diplomatic challenge.
OECD/UNESCO principles, summit declarations, and voluntary commitments set expectations but lack enforcement.
The most realistic near-term wins: hotlines, incident protocols, and red lines (e.g. human control of nuclear use) to prevent inadvertent escalation.
Technology outruns diplomacy; rivalry blocks binding rules. Fragmentation, not a single global regime, is the base case.
Different actors hold different cards. The shared imperative, as I see it: compete without tipping into catastrophe — keeping the rivalry managed and stable rather than letting it bifurcate the world or spiral into conflict.
The signature dangers of the AI era are systemic — they cross borders and can escalate faster than diplomacy can respond. I map them by likelihood and severity to see where guardrails matter most.
Conflict or coercion over Taiwan would sever the leading-edge supply chain — a global economic and security catastrophe.
Pressure to field capability first erodes safety and verification, raising the odds of accidents and miscalculation.
AI-compressed decision cycles can drive inadvertent escalation faster than human crisis management can intervene.
Two non-interoperable tech blocs reduce cooperation on shared dangers and force the uncommitted to choose sides.
Powerful, cheaper models reach more states — and non-state actors — widening access to cyber, bio, and influence tools.
Deterrence, crisis hotlines, red lines, supply-chain resilience, verifiable measures, and sustained great-power dialogue.
Two axes define the outcome space: how power is distributed (concentrated/bipolar vs. distributed/multipolar) and how the order behaves (cooperative/governed vs. confrontational/anarchic). The world will not land cleanly in one box, but its center of gravity will tilt toward one of these four.
Signals I watch to know whether the world is heading toward managed stability or a dangerous race — and how the balance of power is shifting.
| Capability gap | Distance between the two poles' frontier systems |
| Decoupling intensity | Export-control scope, trade splits, parallel stacks |
| Taiwan risk index | Military tension & chip-supply concentration |
| Crisis-stability measures | Hotlines, red lines, confidence-building in place |
| Military-AI norms | Progress (or stall) on autonomous-weapons rules |
| Bloc alignment | How swing states are tilting over time |
| Governance cohesion | Binding commitments vs. fragmentation |
| Compute capacity | Domestic + accessible frontier-scale compute |
| Chip-supply resilience | Dependence on chokepoints; friend-shoring progress |
| Energy headroom | Power available to scale data centers |
| Talent & research base | Frontier publications, labs, net brain flows |
| Alliance depth | Integration into a tech-security bloc |
| Sovereign capability | Local models, data residency, autonomy |
| Defense-AI maturity | Fielded capability with human-control safeguards |
Leading indicator I watch: the gap between decoupling pressure and crisis-stability measures is the single best early signal of where the contest is heading. Rising decoupling with no matching guardrails = drift toward Cold War 2.0; managed decoupling with working red lines = a chance at détente.
Not commandments — reminders I re-read when the geopolitical noise gets loud.
Compute, chips, energy, talent, and alliances form the stack on which 21st-century power rests.
The US–China rivalry is the organizing axis; everyone else hedges, aligns, or bargains around it.
Control of the chip supply chain — and its chokepoints — is the primary instrument of statecraft.
~90% of leading-edge fabrication in one place is the most consequential and dangerous chokepoint on Earth.
Export controls, sanctions, and minerals leverage turn economic ties into tools of coercion.
Parallel stacks and a "silicon curtain" are forming — full separation is too costly, partial split is happening.
Compressed decision cycles can deter — or cause inadvertent escalation faster than humans can manage.
Only soft norms and bilateral understandings; the nuclear-control red line is the most important thread.
Minerals, markets, and swing votes give the uncommitted real bargaining power.
Managed competition with working guardrails — not a hostile two-bloc world or ungoverned chaos.
Annotated bibliography behind the great-power thesis, power-system map, player table, chip-supply-chain diagram, military escalation ladder, statecraft stack, governance timeline, grand-strategy stack, risk quadrant, and scenario matrix. Section tags (e.g. §06) show where each source is used. Diagrams and operating rhythms are my synthesis unless noted.
Scope. Synthesis of international-relations, semiconductor-industry, and defense-policy sources (May 2026). Hero-strip figures (e.g. ~90% leading-edge fabrication in Taiwan, Oct 2022 export controls, zero binding military-AI treaties) are directional claims drawn from the sources below — verify against latest BIS rules, fab capacity data, and UN/CCW proceedings before citing externally. Not legal, financial, or national-security advice.
Citations are numbered continuously [1]–[n] within this section.
Before you quote externally: The ~90% Taiwan/leading-edge fabrication figure compresses NSCAI (2021), CSET (2021), and industry estimates — often cited as “TSMC share” but more precisely “Taiwan’s share of advanced-node capacity.” Export-control details change frequently (BIS interim final rules 2023–25). Summit outcomes (Paris 2025, India next) are political and non-binding. Re-read primary sources and your jurisdiction’s compliance guidance before citing figures or making policy claims.