The Bifurcation Risk

What made war inevitable was the growth of Athenian power and the fear which this caused in Sparta.
— Thucydides, History of the Peloponnesian War

The argument so far treats the transition as a market process: cost curves fall, capability rises, institutional friction determines deployment speed. States appear only as sources of friction—permitting delays, regulatory constraints, liability frameworks that have not yet adapted.

This framing is incomplete. States are not merely friction. They are actors with preferences, capabilities, and time horizons that differ from markets. When a technology becomes strategic, states act strategically. And compute has become strategic.

The Fabrication Chokepoint

The chokepoint is not software. It is silicon.

TSMC and Samsung fabricate over eighty percent of leading-edge logic chips. As of late 2025, TSMC alone accounts for roughly ninety percent of the most advanced nodes—those below seven nanometers where frontier AI workloads run. The fabrication complex sits in Taiwan, ninety miles from a power that has not renounced the use of force to achieve unification. As Miller documents, China now spends more money each year importing chips than it spends on oil—the strategic significance is not lost on Beijing or Washington.(Miller 2022)

This concentration is not an accident of comparative advantage. It is the result of decades of industrial policy, capital deepening, and talent accumulation that cannot be replicated quickly. Intel's attempt to regain leading-edge manufacturing has consumed tens of billions of dollars and years of effort with results that remain uncertain. The CHIPS Act allocates substantial subsidies for domestic fabrication, but new fabs require four to five years to build and additional time to reach yield. The chokepoint persists through the period when agentic capabilities are expected to mature.

The United States has already acted on this understanding. The October 2022 export controls restricted advanced semiconductor equipment and chips to Chinese end-users. Subsequent rules tightened the definitions, closed loopholes, extended restrictions to third countries serving as transshipment points. The policy treats advanced compute as a strategic asset comparable to weapons technology. Denial is the objective, not merely delay.

China has responded with domestic investment at a scale that dwarfs American industrial policy. But manufacturing equipment remains the binding constraint. ASML's extreme ultraviolet lithography machines are produced in the Netherlands, subject to Dutch export licensing influenced by American preferences. As of late 2025, China has reportedly developed a prototype EUV machine, but commercial viability remains years away—ASML's CEO estimates China is 10-15 years behind in chip manufacturing technology. The machines cannot be commercially replicated on any timeline relevant to current strategic competition.

A reasonable criticism is that the market framework ignores geopolitics. Markets have difficulty incorporating actors whose objective functions are not denominated in returns. States maximize power, legitimacy, and survival—quantities that do not reduce to economic value. The omission was deliberate but not defensible. Integration is now required.

Three Scenarios

The structure admits three outcomes, each with different implications for the thesis.

Bifurcation without confrontation. China develops its own AI ecosystem on a delayed capability trajectory, using mature-node chips and domestically developed architectures. The V/C ordering holds in both blocs, though the specific values differ. Agents emerge in both economies, settling through different rails: Bitcoin and stablecoins in the Western bloc, digital yuan and permissioned ledgers in the Chinese bloc. The thesis applies, adjusted for jurisdiction. The market mechanisms operate in both contexts, with different parameter values and institutional constraints.

The Cold War space programs provide the parallel. The United States and Soviet Union developed separate launch vehicle architectures, separate orbital mechanics traditions, separate supply chains for propulsion and guidance. Neither ecosystem was clearly superior; each optimized for different constraints. The competition produced redundancy and inefficiency from a global perspective, but also resilience: neither bloc's capabilities depended on the other's cooperation. When the Soviet program collapsed in 1991, American aerospace did not collapse with it. The bifurcation was costly but stable.

Bifurcation with confrontation. A Taiwan scenario—blockade, invasion, or coercive unification—disrupts the fabrication chokepoint for years. Prices for leading-edge chips spike by multiples. Training runs become impossible for all but the best-provisioned actors. Inference costs rise instead of falling. The timeline compresses not toward capability deployment but toward strategic autarky. States requisition compute capacity for military and intelligence applications. The agentic economy is subordinated to the war economy.

The Suez Crisis of 1956 provides a partial parallel. When Egypt nationalized the canal, Western Europe lost its primary oil transit route overnight. The disruption was not merely commercial; it revealed how deeply European economies had become dependent on a chokepoint outside their control. The crisis accelerated the development of supertankers capable of circumnavigating Africa, pipeline infrastructure across Europe, and strategic petroleum reserves. The immediate shock was severe; the long-term response was diversification away from the chokepoint. A Taiwan scenario would trigger analogous responses—frantic investment in alternative fabrication, stockpiling of chips, and acceleration of domestic capacity—but the timeline for alternatives is measured in years, not months. The interim is the crisis.

The second scenario does not refute the thesis. It suspends it. The underlying physics—energy into computation into selected structure—remains true. But the institutional context within which that physics operates changes categorically. Market allocation gives way to state allocation. V/C ordering gives way to strategic priority. The energy floor becomes a floor on military compute rather than commercial inference.

State capture of apex compute. States may conclude that concentrated compute is too dangerous to leave in private hands—whether or not confrontation occurs. The arguments are available: AI systems can generate disinformation, design weapons, optimize surveillance, concentrate economic power. Regulatory capture may prove insufficient to align private incentives with public interest. Public utilities were created from less consequential technologies. The case for nationalization—or for licensing regimes so restrictive as to approximate nationalization—will find advocates.

The history of telecommunications provides the parallel. AT&T operated as a regulated monopoly from 1913 to 1984, its long-distance network treated as a natural monopoly too important to leave to market competition. The arrangement was stable for decades: private ownership with public oversight, guaranteed returns in exchange for universal service obligations. The system worked until it didn't. By the 1970s, the regulatory framework that had enabled AT&T's dominance was also preventing competition and innovation. The breakup came not because the original logic was wrong but because technology had shifted the terms. Compute may follow the same arc: consolidation justified by strategic importance, regulatory oversight hardening into control, eventual restructuring when the costs of the arrangement exceed its benefits. The timeline of such cycles is measured in decades.

If apex compute becomes a public utility, the thesis requires modification. The V/C ordering still describes which tasks automate first, but the deployment decision is no longer made by firms seeking returns. It is made by administrators seeking objectives that may diverge from economic efficiency. The liability sink becomes the state itself, absorbing consequence through sovereign immunity rather than insurance and indemnification.

Observable Signals

One asymmetry matters more than the others: the default, absent deliberate state action, is market allocation. Markets do not request permission to operate. States must actively intervene to stop them. The burden falls on the state, not on the market.

The geopolitical scenario is therefore not a prediction. It is a boundary condition—a set of circumstances under which the thesis requires modification.

The thesis as developed assumes a world where markets continue to function, where capital flows to highest-value uses, where deployment decisions are made by firms rather than strategic planners. If that assumption fails—through bifurcation, confrontation, or state capture—the specific predictions about deployment sequence, value capture, and settlement infrastructure require revision.

The falsification condition is observable. If leading-edge compute becomes a licensed strategic resource in major jurisdictions, with allocation determined by administrative process rather than price, then market mechanisms no longer govern deployment. The physics remains true. The economics no longer follows automatically from the physics. Politics intervenes.

Observable signals would indicate which scenario is unfolding.

Bifurcation without confrontation:

Divergent AI capability benchmarks between blocs, with neither achieving clear dominance
Separate settlement infrastructure development (Bitcoin/stablecoins in Western bloc; digital yuan and permissioned ledgers in Chinese bloc)
Continued chip trade restrictions without military escalation
Parallel standards bodies and certification regimes

Bifurcation with confrontation:

Taiwan Strait military activity or coercive measures short of invasion
TSMC production disruptions or evacuation of key personnel
Leading-edge chip prices spiking beyond 3× baseline
Emergency requisition of compute capacity for defense applications

State capture of apex compute:

Export control scope expanding beyond current targets to include inference hardware and model weights
Licensing requirements for training runs above specified compute thresholds, as some jurisdictions have already proposed
Mandatory government access to frontier model weights before or upon deployment
Nationalization of fabrication or datacenter infrastructure in response to perceived emergency
Compute allocation quotas or priority systems that subordinate commercial deployment to state-designated objectives

Some of these signals are already present in nascent form. Executive orders require reporting of large training runs. Allies coordinate on export restrictions. The EU mandates disclosure and conformity assessment for high-risk AI systems. China requires algorithm registration and security review.

The question is whether these interventions remain friction—slowing market-driven deployment without altering its direction—or become control, subordinating deployment to objectives other than economic return.

What Bitcoin Demonstrates and What It Does Not

Bitcoin appears in the first scenario as the Western bloc's settlement infrastructure, and its appearance requires an honest accounting. The temptation in a volume about energy, computation, and verification is to treat Bitcoin as the proof of concept for everything. It is not. What it demonstrates is specific, and the specificity is what gives it value as evidence.

What Bitcoin demonstrates. Verification without trust: a distributed system can reach consensus on the state of a ledger without any participant trusting any other, and the consensus has held under adversarial conditions for over fifteen years. Thermodynamic commitment: the Joule Standard described in Chapter 15 is not a metaphor — Bitcoin's proof of work literally converts energy expenditure into a costly history that cannot be cheaply rewritten. Unforgeable costliness: the cost of attack scales with the accumulated work, creating a security property that no institutional promise can match. Settlement finality: once confirmed, a transaction cannot be reversed by any party, including the system's designers. These are genuine achievements, and they are relevant to the trilogy's thesis because they demonstrate that the V/C ratio can be made favorable (low verification cost, high value of finality) in a specific domain.

What Bitcoin does not demonstrate. Governance: Bitcoin's governance is ad hoc, contentious, and slow — the block-size wars demonstrated that even a minimal protocol change can fracture the community, and the resolution was fork-based (Bitcoin Cash), not deliberative. Scalability: base-layer throughput remains limited, and the Lightning Network, while promising, introduces its own trust assumptions (channel counterparties, routing nodes, liquidity providers) that the base layer was designed to eliminate. Identity and liability: Bitcoin's pseudonymous design means that consequence cannot be routed to a responsible party — the liability sink problem of Chapter 24 is not solved but deliberately evaded. Mercy: the ledger is permanent by design; there is no mechanism for record expiry, sealed histories, or structured forgetting. The fourth equation has no Bitcoin answer. And energy efficiency: proof of work's energy consumption is not incidental but constitutive, which makes it unsuitable for domains where the verification cost must be low enough for universal access — exactly the requirement that Vol III's Right to Verify chapter demands.

The honest summary: Bitcoin is the existence proof for the Joule Standard (Chapter 15) and for verification without trust. It is not the existence proof for the coordination substrate (Vol II Appendix B), the receipt regime (Vol III), or the mercy architecture (Vol III Ch 13). The trilogy draws on Bitcoin where the evidence supports — thermodynamic commitment, unforgeable costliness, settlement finality — and does not extend the claims beyond their domain. The bifurcation scenarios above treat Bitcoin as settlement infrastructure, not as governance architecture, because that is what the evidence warrants.

What Remains Undetermined

Whether the transition proceeds through market allocation or state allocation is a political question, not an economic one. But the framework can say what follows from each answer.

If bifurcation proceeds without confrontation, the predictions hold with jurisdictional adjustment. The V/C ordering operates in both blocs. Agents emerge in both. Settlement infrastructure diverges but the logic is the same.

If confrontation disrupts the fabrication chokepoint, the timeline suspends until capacity rebuilds or strategic equilibrium is reestablished. The thesis is not wrong; it is postponed.

If state capture proceeds in major jurisdictions, the market mechanisms described here yield to administered allocation. The thesis is not wrong; it is superseded by a different logic.

The thesis is robust to the first scenario, suspended by the second, superseded by the third. Which scenario obtains is a question of politics and statecraft, not economics or thermodynamics.

The analysis has treated geopolitics as exogenous. This is a simplification, not a description. The transition itself creates the political stakes that determine which scenario unfolds. The state moves because the technology is consequential. The technology becomes consequential in ways that depend on how the state moves.

The interaction is reflexive. The outcome is undetermined.

The default is market allocation. States can end that default. The thesis holds unless they do.