Fusion and the Narrowing of State Imagination

After decades of incremental progress, fusion sits at an inflection point. The remaining uncertainties strain the justificatory frameworks of modern systems. China’s response reveals the constraints shaping contemporary state imagination. Fusion has long occupied a peculiar place in the technological imagination: at once inevitable and perpetually deferred. It promises not merely abundance but a transformation in the material basis of modern life. For decades, this promise licensed a certain intellectual excess — a willingness to pursue multiple paths, to tolerate uncertainty, to entertain solutions not yet fully legible to existing institutions. That posture is beginning to change. At several points in modern history, frontier technologies have exposed not only technical limits but institutional ones. They reveal how states handle uncertainty: whether it is treated as a domain to be explored or a condition to be managed. Fusion now occupies such a position. After decades of research and renewed investment, it sits at an apparent inflection point because the remaining uncertainties no longer align comfortably with the ways modern systems justify and large expenditure and commitments.

Fusion is routinely described as a race, a metaphor both convenient and misleading. Races presume a defined track, a known finish line and an agreement on what constitutes victory. Fusion offers none of these. Its significance lies in its resistance to linearisation. It forces institutions to confront situations in which timelines cannot be credibly promised, progress cannot be cleanly staged, and failure must precede success at scale. In this sense, fusion functions less as an energy project than as a test of how states behave when uncertainty cannot be deferred. The state of play in China In China fusion has been elevated to a position of long-term institutional importance. This is evident in the scale and persistence of its physical infrastructure. Beijing operates multiple magnetic confinement facilities, including the Experimental Advanced Superconducting Tokamak (EAST) in Hefei and the HL-2M tokamak in Chengdu, which achieved first plasma in 2020. In November 2025, the Chinese Academy of Sciences publicly unveiled the BEST Research Plan, explicitly linking the Burning Plasma Experimental Superconducting Tokamak to burning plasma regimes relevant to the next stage of fusion research. But these facilities share a defining feature: they are all tokamaks. China does conduct non-tokamak fusion research but the contrast in scale and institutional framing is pronounced. The Chinese First Quasi-axisymmetric Stellarator (CFQS) is a joint project with Japan's National Institute for Fusion Science, constructed by Southwest Jiaotong University. Public technical descriptions characterise it as a compact research device — parameters consistent with exploratory experimentation rather than national-scale infrastructure. There is no publicly announced Chinese stellarator or alternative-confinement facility comparable in scale or institutional status to EAST, HL-2M, or BEST. The difference is not one of existence but elevation

Large, state-backed systems require a particular kind of intelligibility. They must be governable in advance, decomposable into subsystems, and capable of being justified within planning frameworks that demand measurable progress. Tokamaks, for all their complexity, meet these requirements better than most alternatives. Their parameters are understood, their scaling laws are familiar, and their failures are, to some extent, predictable. This does not make them optimal, it makes them administratively intelligible.

Such an approach carries clear advantages. Tokamak research is the most mature branch of magnetic confinement fusion. Its physics is comparatively well characterised, its diagnostic techniques standardised, and its failure modes familiar. Concentrating resources here reduces epistemic risk, facilitates international collaboration, and produces results that are readily digestible to domestic planners and foreign peers.

But it also narrows the space of futures treated as serious candidates. By committing national-scale infrastructure primarily to tokamaks, China effectively accepts the dominant assumptions of contemporary fusion research: that tokamak confinement represents the leading path forward; that remaining obstacles are largely matters of scale, materials, and integration; and that progress will be cumulative rather than discontinuous. This position is defensible, but it carries major constraints.

Fusion history offers multiple examples of paradigms that persisted long after limitations became evident. In such cases, the problem was not scientific error but institutional inertia. Once facilities, funding streams and career incentives align around a particular approach, deviation becomes costly regardless of merit.

China's fusion programme displays the characteristics of a system optimized for coherence. Large facilities anchor research agendas. Multi-year planning cycles reward predictable advancement. Participation in international frameworks reinforces consensus alignment. Within this structure, failure is acceptable when it can be framed as incremental learning within the paradigm. It is less clearly rewarded when it implies the paradigm may be inadequate.

The result is a programme that appears expansive externally but constrained internally. What fusion reveals China could be doing instead If China's current fusion programme reflects an effort to manage uncertainty, it also clarifies what an alternative posture would entail. The distinction is not between ambition and neglect, but containment and exploration.

China's institutional system possesses features well suited to frontier domains characterised by long timelines and uncertain outcomes. It can sustain projects across political cycles, absorb repeated failure without immediate collapse, and mobilize large, multidisciplinary efforts without requiring near-term commercial validation. Fusion is a problem that would, in principle, benefit from such capacities.

A programme designed around exploration rather than convergence would look very different. Rather than organising around a single dominant paradigm, it would treat fusion as a space to be systematically mapped. Progress would be measured less by proximity to a reactor design than the closure of unknowns.

Such a programme would enforce divergence by design. Multiple confinement concepts — tokamaks, stellarators, compact tori, magneto-inertial systems — would be pursued at comparable institutional weight, not as symbolic hedges but as vehicles for generating decisive information. Designs would be selected for the clarity of the questions they pose and the diagnostic value of their failure.

Failure, in this framework, would be an explicit output. Projects would be launched with defined termination criteria. Teams would be rewarded for demonstrating non-viability under specified conditions. Negative results would be treated as cumulative assets rather than quietly absorbed.

This posture would also reorder fusion's internal priorities. Instead of plasma performance dominating early-stage evaluation, first-order attention would be given to materials survivability under neutron flux, tritium handling, heat extraction, remote maintenance, and component replacement — the factors that determine whether fusion can exist outside controlled laboratory environments. China's own CFETR planning documents, which foreground tritium breeding ratios and neutron irradiation effects, implicitly acknowledge this reality.

None of this implies abandoning tokamak research, it just means deferring convergence until the design space has been narrowed by evidence rather than habit. China has the structural capacity to attempt such an approach so the fact it does not is significant. Fusion, AI, and the management of uncertainty Fusion is not the only domain in which this posture appears. A similar logic shapes China's approach to artificial intelligence, where development is consistently paired with control. Official discourse emphasises application, governance and risk management alongside innovation, while deployment in regulated sectors is treated as the primary objective. Systems are evaluated less as speculative frontiers than as components to be embedded within existing administrative and economic structures.

In Western benchmark-driven narratives, Chinese AI is often framed as constrained or lagging at the frontier. Domestically, however, optimization for deployability and governability reflects rational adaptation to regulatory and procurement realities — rewarding reliability and integration over unpredictable gains.

The analogy has limits. AI is fast-moving and diffuse; fusion is slow and irreducibly physical. But the recurrence of the same posture across such different domains suggests something more than coincidence: a preference for futures that can be specified in advance over those that must be discovered. Fusion as a problem of late modernity

It is tempting to interpret fusion's tardiness as a technical failure or a byproduct of geopolitical competition. Neither explanation is sufficient. In reality the remaining obstacles are not primarily scientific, or financial, but institutional.

The United States and China approach fusion from different starting points but converge in how uncertainty is handled. The United States runs this through markets, China through planning hierarchies — but both are ultimately systems for converting uncertainty into something that can be managed, reported on and defended. The mechanisms differ but the effect is similar. Both systems narrow the range of futures treated as plausible and favour paths that can be defended in advance.

Fusion exposes this convergence because it resists accommodation. It cannot be rushed into profitability without distortion, nor folded cleanly into planning without prematurely fixing assumptions. It requires exploration before optimisation and failure before refinement. What it reveals is not a lack of resources but a narrowing of acceptable uncertainty. Abundance has not expanded imagination; it has increased the cost of deviation. Scale has not enabled risk; it has disciplined it.

China's case is particularly revealing because it possesses the necessary capacity. Few systems are as capable of absorbing failure, sustaining parallel exploration, or treating negative knowledge as a strategic asset. That China nonetheless gravitates toward orthodoxy in fusion suggests the constraint is internal rather than imposed.

Fusion, then, is not stalled at the edge of a breakthrough. It is stalled at the edge of a choice between epistemic postures. Whether to continue privileging futures that can be specified in advance, or to reopen space for those that can only be discovered through sustained engagement with the unknown.

That choice is rarely articulated. More often it is deferred, managed or obscured by narratives of progress. Fusion cuts through those narratives precisely because it offers no shortcut. It may ultimately tell us less about the future of energy than about how modern states have come to relate to the unknown — not with imagination but administration.