Why ‘defence innovation’ is not one market but three — and why treating it as one is an error about capability itself

For roughly a decade, the story that the defence-technology world has told about itself has had two characters. On one side stand the primes — the great legacy contractors, Lockheed Martin, RTX, Northrop Grumman, BAE Systems, whose names have been fixed to the largest weapons programmes for two generations. On the other stand the startups, the software-first insurgents, the venture-funded newcomers who promise to build faster, cheaper, and better than the incumbents they intend to displace. It is a satisfying story, with the shape of a David-and-Goliath contest, and like most satisfying stories it is a poor guide to the terrain.

The trouble is not that the story is false. In the United States, it is broadly true, and we will spend some time establishing just how true it has recently become. The trouble is that it is a description of one market — the American one — that has been mistaken for a description of a global phenomenon. When the primes-versus-startups frame is applied to Europe, it distorts more than it reveals, because the European problem is not incumbency but fragmentation. When it is applied to the states that Western analysts have taken to grouping under the acronym CRINK — China, Russia, Iran, and North Korea — it collapses entirely, because in a system built around the fusion of civil and military production the distinction between an incumbent prime and a commercial newcomer does not exist in the form the frame assumes.

This essay argues that there are not one but three structurally different defence-technology markets, that each is producing a different relationship between capital, the state, and capability, and that any analysis which treats ‘defence innovation’ as a single global category is not merely imprecise but wrong about the thing that matters most: the trajectory of military capability itself. The three markets are diverging, and they are diverging in ways that the incumbency frame cannot see. To read the field correctly, one has to abandon the two-character story and look instead at the structure beneath each market — at who allocates capital, who bears risk, and where the boundary between the civilian and the military economy actually falls.

I. The American market: the frame that is true

Begin with the United States, because it is the market the standard story describes, and because in the past two years the story has become more dramatically true than even its proponents expected. The rise of the American defence-technology startup is not a projection or a promise; it is now a matter of contract awards and capital flows of a scale that requires the incumbents to be taken seriously as challengeable.

Consider the anchor case. Anduril, founded in 2017 and for years dismissed by the establishment as a curiosity, reached a valuation of sixty-one billion dollars in a funding round in May 2026 — roughly double its valuation of eleven months earlier — on the back of disclosed revenue of some 2.2 billion dollars in 2025, itself a doubling of the prior year. More consequential than the private capital was the procurement signal beneath it. In March 2026 the US Army awarded Anduril a ten-year enterprise contract with a ceiling of up to twenty billion dollars, consolidating more than 120 separate procurement actions into a single framework — the largest such vehicle the Army had ever issued to a non-traditional contractor. Whatever one thinks of the company, that award is not a curiosity. It is the procurement system restructuring itself around a new entrant.

Anduril is the most visible case but not an isolated one. Palantir, the software company that pioneered the direct-sale model in defence, saw its market capitalisation cross three hundred billion dollars by mid-2026. A cohort of younger companies — Shield AI in autonomous flight, Saronic in autonomous vessels, Epirus in directed energy, Hermeus in hypersonics — has raised at valuations that would have been unthinkable five years ago. The sector as a whole absorbed on the order of forty-nine billion dollars in venture funding in 2025, roughly double the previous year’s total, and the pace through 2026 has run ahead of that again. These are the figures of a market in which new entrants have genuine procurement traction, not merely investor enthusiasm.

Epistemic status: the funding and contract figures are Confirmed from company disclosures, contract announcements, and market reporting current to mid-2026. Valuations of private companies are point-in-time and move quickly; the reader should treat specific figures as accurate to the date of writing rather than as fixed, and the managing editor should refresh them at publication.

The structural reason the American frame is true is a specific feature of the US acquisition system: the on-ramp. The Other Transaction Authority — a procurement pathway that bypasses the standard federal acquisition rules — allows the Department of Defense to contract with a non-traditional vendor in three to six months rather than the twelve to eighteen that a conventional contract requires. Together with the Small Business Innovation Research programme, the OTA gives a startup a route to early government revenue that de-risks the venture investment behind it. The American model works not because Silicon Valley is uniquely inventive but because the American state built a mechanism through which commercial speed can reach a government customer before the company runs out of money. The frame is true in the United States because the plumbing makes it true.

It is worth dwelling on a second case, because Anduril is now large enough that its trajectory could be read as exceptional rather than structural. Saronic, founded in 2022 by alumni of Anduril, took the same model into a different domain — autonomous surface vessels — and reached a valuation of around nine billion dollars by 2026 on the strength of a maritime-autonomy thesis and a small number of large early contracts. Its significance is not its size, which remains a fraction of the incumbents’, but its speed: a company founded during the pandemic reached multi-billion-dollar scale in a domain — naval shipbuilding — that has historically been the most incumbent-bound and capital-intensive of all. Saronic is evidence that the American on-ramp is not a one-company phenomenon confined to the aerial-autonomy niche where Anduril began, but a repeatable pathway that a founder can deliberately set out to travel.

The repeatability is the point. When alumni of one successful new entrant leave to found the next, choosing their domain in advance and raising capital against a known playbook, the market has stopped being a story about a single insurgent and become a system for producing insurgents. That system — venture capital that treats defence as an asset class, a procurement on-ramp that delivers early revenue, and a talent pool circulating between the new entrants — is what distinguishes the American market. None of its three components exists, in the same developed form, in either of the other two markets. That absence is why the American story does not travel.

This is the point at which most commentary stops, having established that the insurgents are winning, and generalises the conclusion to the rest of the world. That generalisation is the error this essay exists to correct. The American market has a feature — a fast, well-capitalised on-ramp from commercial development to government revenue — that the other two markets do not share, and the absence of that feature changes everything.

II. The European market: fragmentation, not incumbency

Turn to Europe, and the first thing to establish is that the incumbency frame does not describe the binding constraint. Europe has legacy primes — BAE Systems, Leonardo, Thales, Rheinmetall, Airbus Defence and Space — and it has a rising generation of well-funded startups. But the defining feature of the European defence-technology market is neither the strength of the incumbents nor the weakness of the newcomers. It is fragmentation: the fact that Europe does not buy defence as a single entity, and that its demand is split across twenty-seven national governments whose budgets, planning cycles, and industrial preferences are determined domestically.

The numbers make the point starkly. Even at the height of the post-2022 rearmament, collaborative procurement — Member States buying equipment jointly rather than nationally — has historically accounted for a small fraction of European defence spending; European Defence Agency data put collaborative equipment procurement at around eighteen per cent of the total. The consequence is duplication on a scale that has no American parallel: Europe operates far more distinct types of tanks, combat aircraft, and naval platforms than the United States, each with its own production line, sustainment chain, and national industrial constituency. The problem is not that Europe lacks innovators. It is that European demand is structured in a way that fractures the market a startup would need in order to scale.

And yet the European startup scene is, by any measure, thriving — which is the apparent paradox this section must resolve. Helsing, the Munich-based AI and strike-drone company founded only in 2021, reached an eighteen-billion-dollar valuation in a funding round in July 2026, having raised 1.8 billion dollars in Europe’s largest-ever defence-startup round. Quantum Systems, a German drone maker, roughly doubled its valuation to around eight billion dollars the same month. European defence, security, and resilience startups raised a record 8.7 billion dollars in venture funding across 2025, up more than half on the prior year and nearly four times the level of five years earlier. Germany’s share of total EU defence-tech fundraising rose from four per cent in 2019 to eleven per cent by 2024 — a structural shift toward homegrown champions.

Epistemic status: Confirmed. The Helsing and Quantum Systems figures are drawn from company statements and multiple independent reports of July 2026; the aggregate European funding figures are from the Dealroom–NATO Innovation Fund dataset. As with the American figures, valuations are point-in-time.

How can a market this fragmented produce companies this valuable? The answer is that the capital is flowing ahead of the demand structure, on a bet that the demand structure will change. Two things are driving it. The first is sovereignty: European governments and investors have concluded that dependence on American defence technology — subject to American export controls and American political conditions — is a strategic vulnerability, and there is a premium on building a native alternative. Helsing’s investors emphasise, pointedly, that the company remains roughly eighty per cent European-owned even as American capital participates; the ownership is part of the product. The second is the war in Ukraine, which has served as a live proving ground and a demand signal, pulling European systems into real operational use faster than any peacetime process would allow.

But the sovereignty premium and the wartime demand signal are not yet a substitute for a unified market, and this is the European market’s defining tension. Brussels has understood the problem for years and has built an expanding set of instruments to address it. The European Defence Fund, launched in 2021 with a budget of roughly eight billion euros for 2021–2027, exists explicitly to reduce fragmentation by co-financing collaborative, cross-border development — funding 224 projects with around four billion euros by late 2025, and allocating a further billion for 2026. Newer instruments — the European Defence Industry Programme, the SAFE facility, the ReArm Europe / Readiness 2030 initiative designed to mobilise up to eight hundred billion euros, and a proposed European Competitiveness Fund that would assign 131 billion euros to defence and space — represent an attempt to build, from the top down, the unified demand that the American market possesses by default.

This is the heart of the European story, and it is the opposite of the incumbency frame. The European question is not whether startups can beat primes. It is whether Brussels can assemble a coherent European market fast enough to give European startups something to scale into before the sovereignty premium that is currently funding them runs out of patience. The capital has arrived ahead of the market. Whether the market catches up is the single most important variable in European defence technology, and it is a question about procurement architecture and political will, not about the relative merits of incumbents and insurgents.

There is one further signal worth marking, because it suggests the demand structure may be beginning to move. In 2025 the French procurement agency contracted a startup, Harmattan, for an initial programme of a thousand loitering-munition units, delivered inside six months; the agency then increased the order fivefold. That is a European government routing procurement earlier in the cycle and rewarding delivery speed — the behaviour the American on-ramp institutionalises, appearing in Europe as a deliberate choice rather than a standing mechanism. One such case is not a trend. But it is the shape the European market would take if the reform succeeds, and it is worth watching for that reason.

Germany is worth isolating as a microcosm, because its trajectory shows both the promise and the constraint in concentrated form. The rise of Germany’s share of European defence-tech fundraising — from four per cent in 2019 to eleven per cent by 2024 — is not incidental; it reflects a deliberate national turn, catalysed by the post-2022 Zeitenwende and the special defence fund that accompanied it, toward treating defence technology as a domain of national industrial ambition rather than embarrassment. Helsing and Quantum Systems are both German, and both are, in effect, national champions in formation: companies that a European government would like to see scale, backed by capital that is being kept deliberately European in its ownership.

But Germany also shows the constraint. A national champion needs a market larger than its nation, and the German firms’ valuations already imply a customer base that only a unified European demand could supply. The German ecosystem has produced the companies; it cannot, by itself, produce the market they are valued against. This is the fragmentation problem viewed from the supply side rather than the demand side: even the strongest national ecosystem in Europe runs into the ceiling of a market that stops at national borders, and the German champions are, in a real sense, a bet that Brussels will remove that ceiling before the valuations have to be justified. The Nordic ecosystems — Sweden’s around Saab and its spinouts, Finland’s and Estonia’s dense clusters of dual-use firms — tell a parallel story at smaller scale: disproportionate innovation output constrained by the same fractured demand. The national ecosystems are real and rising. The market they need is still being assembled in Brussels.

III. The CRINK market: where the distinction dissolves

The third market is the one where the incumbency frame does not merely mislead but ceases to apply, and it requires the most care to describe accurately — both because the evidence is harder to verify and because the temptation to alarmism is strongest. The states that Western analysts group as CRINK do not, of course, form a single integrated market; Russia, Iran, and North Korea operate under sanctions regimes and industrial constraints that differ profoundly from China’s, and lumping them together can obscure as much as it clarifies. But they share one structural feature that distinguishes their defence-technology systems from both the American and the European: the deliberate erasure of the boundary between the civilian and the military economy. This section concentrates on China, because it is the most consequential case and the best documented, and treats the others as variations on the structural theme rather than as equivalents.

China’s organising strategy is what Beijing calls military-civil fusion. Elevated to a national strategy in 2017, it is a systematic programme to integrate the civilian commercial and research economy into the defence-industrial base, so that advances made in nominally civilian companies, universities, and laboratories flow to the People’s Liberation Army with minimal friction. The strategy explicitly draws inspiration from the American ability to leverage commercial innovation for military advantage — but where the American mechanism is a procurement on-ramp that a company chooses to use, the Chinese mechanism is a structural fusion that the state can compel.

The evidence that the fusion is producing results at the level of actual procurement — rather than merely policy documents — has recently become firmer. A study by Georgetown’s Center for Security and Emerging Technology, published in 2025, analysed 2,857 AI-related PLA contract award notices from 2023 and 2024 and identified 1,560 distinct organisations that had won at least one such contract. The finding is significant in a way that bears directly on this essay’s argument: while legacy state-owned defence conglomerates still lead AI-related military procurement, an emerging class of non-traditional commercial vendors and research institutions plays a consequential role. In the Chinese system, in other words, the ‘startup’ and the ‘prime’ are not competitors in a market. They are both inputs to a single state-directed pipeline, and the state has designed the pipeline so that it need not choose between them.

Epistemic status: Probable, with the CSET contract-dataset findings Confirmed as to their own methodology. The broader characterisation of military-civil fusion draws on CSET, the US State Department’s 2025 assessment, and the National Bureau of Asian Research; these are authoritative but necessarily work from partial visibility into a deliberately opaque system. The essay states what the open evidence supports and does not extrapolate to capability claims the evidence cannot bear.

The structural implications follow directly. China has been building the machinery of fusion around it: state-backed venture-capital funds — including a national guidance fund and regional funds each exceeding fifty billion yuan, launched in late 2025 — that channel patient capital toward early-stage ‘hard technology’ with dual-use application in aerospace, quantum, and integrated circuits. Where the American startup raises from Founders Fund and a16z and hopes for an OTA award, and the European startup raises from Dragoneer and Lightspeed and hopes Brussels builds a market, the Chinese dual-use firm is funded, directed, and absorbed by an integrated state apparatus for which the civil-military distinction is an obstacle to be removed rather than a boundary to be respected.

This is why the incumbency frame collapses in the CRINK market. To ask whether Chinese defence startups are displacing Chinese primes is to ask a question the system is designed to make meaningless. The relevant Western analytical response has not been to track that competition — there is none in the Western sense — but to attack the fusion mechanism itself: the US Department of Defense’s Section 1260H list of ‘Chinese military companies,’ the Commerce Department’s Entity List designations of firms tied to the PLA, and the State Department’s counter-fusion strategy are all attempts to raise the cost of the civil-military integration by cutting the fused entities off from American technology and capital. The West is not competing with CRINK startups. It is trying to break the machine that makes the startup-versus-prime question irrelevant.

A word of caution is essential here, because this is the section where an essay of this kind most easily overreaches. Military-civil fusion is a strategy, not a guaranteed outcome; China’s defence sector has a long history of inefficiency and corruption, and the fusion programme is in part an attempt to overcome those failings rather than proof that they have been overcome. The evidence establishes that the fusion is real, that it is producing procurement activity across a wide base of commercial and non-traditional vendors, and that it dissolves the analytical categories the West uses to understand its own market. It does not establish that the resulting capability exceeds, or will exceed, that of the United States. The honest claim is narrower and still important: the CRINK market is structured on a principle that makes the primes-versus-startups frame category-inappropriate, and Western analysis that carries the frame across the border will misread what it is looking at.

The rest of the CRINK grouping deserves more than the aggregating acronym allows, because the three other members diverge sharply from China and from one another. Russia’s defence-technology system is not a fusion model but a mobilisation model: a legacy state-industrial base, stripped of Western inputs by sanctions, sustaining wartime production through import substitution, reverse engineering, and third-country procurement routes. Its innovation is real but narrow, concentrated in the systems the war demands, and its startup layer is thin. Iran’s system is the most distinctive of the four — a genuinely indigenised asymmetric-capability base, built over four decades of sanctions, that has become an exporter of drones and missiles to the others; it is a fusion model of a different kind, in which the Revolutionary Guard’s economic empire blurs the civil-military line from the military side rather than the civilian one. North Korea’s is a command system in the strict sense, with no meaningful commercial layer at all, sustaining a narrow set of strategic programmes at ruinous cost to everything else.

What unites the four is not a shared market structure but a shared repudiation of the boundary the Western frame depends on — and, increasingly, a set of transfers among themselves that Western analysts are only beginning to map: Iranian drones in Russian service, North Korean munitions on European soil, Chinese dual-use components throughout. The CRINK ‘market’ is better understood as a set of structurally illiberal systems that are learning to supply one another under pressure, than as a single integrated bloc. But for the purposes of this essay’s argument, the aggregation holds at the level that matters: in none of the four does the primes-versus-startups frame describe the organising reality, and in all four the civil-military boundary that the frame assumes is, by design, absent.

IV. The case for a single market — and why it fails

Before drawing the threads together, honesty requires stating the strongest case against this essay’s thesis, because it is not a weak case. The single-market view holds that the three markets are converging, not diverging, and that the differences this essay emphasises are transitional rather than structural. Its argument runs as follows. Capital is global: the same American venture funds — Dragoneer, Lightspeed, General Catalyst, a16z — appear in the cap tables of American and European defence startups alike, and increasingly set the terms in both. Technology is global: the underlying advances in autonomy, AI, and sensing are the same across all three markets, developed in a shared research literature and diffusing across borders faster than any control regime can prevent. And the demand signal is global: the war in Ukraine has served as a single proving ground that is pulling American, European, and even Chinese systems toward a common set of battlefield-validated capabilities. On this view, there is one global defence-technology market with three regional access points, and the structural differences will erode as capital, technology, and demand continue to integrate.

This is a serious argument and it is partly right — which is precisely why it is dangerous. Capital, technology, and the demand signal are indeed increasingly global, and an analyst who ignored that would miss real convergence at the level of the technologies themselves. But the argument mistakes convergence in inputs for convergence in structure, and it is structure that determines capability trajectory. The same venture capital that flows into an American startup flows into a European one — but in the American case it meets a procurement on-ramp and in the European case it meets a fragmented market and a political bet, and those two destinations produce different companies from the same capital. The same autonomy research is available in Shenzhen and in California — but in one system it is absorbed through state-directed fusion and in the other through a competitive contract, and those two absorption mechanisms produce different rates and directions of fielding. Convergence in inputs against divergence in structure does not yield a single market. It yields three markets processing increasingly similar raw materials into increasingly different outcomes. The single-market view is the incumbency frame’s more sophisticated cousin, and it fails for the same reason: it looks at the inputs and misses the plumbing.

V. Three markets, three trajectories

Set the three side by side and the divergence is clear. The American market is a genuine contest between incumbents and insurgents, refereed by a procurement on-ramp that lets commercial speed reach a government customer — the frame is true, and the insurgents are, for now, ascendant. The European market is not a contest between incumbents and insurgents at all but a race between capital and market structure: the startups are funded, and the open question is whether Brussels can assemble a unified demand for them to scale into before the sovereignty premium exhausts itself. The CRINK market is not a contest in the Western sense at all, because its organising principle is the deliberate erasure of the boundary the Western frame depends on.

The reason this matters — the reason it is more than a taxonomic tidiness — is that capability trajectory follows market structure. An American startup’s trajectory is set by its ability to win contracts against incumbents through a fast acquisition pathway. A European startup’s trajectory is set by a political process in Brussels largely outside its control. A Chinese dual-use firm’s trajectory is set by the state’s capacity to direct capital and absorb output. These are three different engines, and they will produce different rates and directions of capability development. To forecast where military technology is going, one has to forecast the behaviour of three distinct systems, not one global market with regional accents.

This is also why the boomerang dynamic examined elsewhere in this volume bears on the argument. Western export controls and entity-listing are attempts to slow the CRINK market by denying it inputs; but a system built on forced indigenisation and state-directed capital is precisely the kind of system in which denial can accelerate domestic capability rather than suppress it, above a certain threshold of pre-existing capacity. The three-market structure and the boomerang are two views of the same underlying fact: that the instruments the West uses to understand and to act upon defence innovation were built for a market that only one of the three actually resembles.

The practical counsel that follows is modest and, we think, correct. Discard the single global ‘defence innovation’ narrative; it is an American story wearing a global costume. Read each market through its own structure — the on-ramp in the United States, the fragmentation-and-reform race in Europe, the fusion in China and its variants across the rest of CRINK. And treat the boundaries between the three as the most important analytical objects of all, because it is at those boundaries — export controls, ownership rules, entity lists, the sovereignty premium — that the divergence is being actively contested. The primes and the pretenders are real. But they are characters in one market’s story, and the field has three.

References

United States market — funding and procurement

Sacra. “Anduril revenue, valuation & funding.” Company analysis, 2026 (2025 revenue ~$2.2bn; March 2026 US Army 10-year enterprise IDIQ, ceiling up to $20bn; May 2026 Series H at $61bn).

Augment Market. “Anduril at $61B and Defense Tech’s 2026 Reset,” May 2026.

CNBC. “Silicon Valley Defense Tech Startups” (neoprimes and the legacy primes), October 2025.

ValueAdd VC. “Dual-Use Startups: How Defense Tech Investments Work,” and “Defense Tech Startups 2026,” 2026 (sector VC ~$49bn in 2025; OTA and SBIR on-ramp mechanics; Saronic ~$4bn+ valuation).

Landbase. “10 Fastest Growing Defense Tech Companies and Startups,” 2026 (sector and company valuation summary).

European market — funding, procurement and structure

Reuters / CNBC / Defense News. Coverage of Helsing’s $1.8bn Series E at an $18bn valuation, July 2026, and prior €600m Series D at ~€12bn (June 2025).

TechCrunch. “Daniel Ek-backed defense tech Helsing to raise $1.2B at $18B valuation,” May 2026.

Reuters. Quantum Systems $1.2bn round at ~$8bn valuation, July 2026.

Dealroom & NATO Innovation Fund. European Defence, Security and Resilience venture data (record $8.7bn in 2025; Germany’s share of EU defence-tech fundraising rising from 4% (2019) to 11% (2024)).

Swiss Aerospace Ventures. Analysis of the Harmattan / French DGA loitering-munition programme (initial 1,000 units delivered within six months; subsequent fivefold order), July 2026.

European Parliament / EPRS. Briefings on the European Defence Fund and European Defence Industry Programme (EDIP): collaborative equipment procurement ~18% of the total (EDA data); ReArm Europe / Readiness 2030 (up to €800bn).

European Commission, DG Defence Industry and Space. EDF Work Programmes 2025 and 2026 (~€1bn per year; 224 projects, ~€4bn committed by end 2025).

ECIPE. “Openness and Fragmentation in EU Defence Procurement,” Policy Brief, December 2025 (European Competitiveness Fund €131bn defence-and-space allocation under the new MFF).

CRINK market — military-civil fusion and counter-measures

Cole McFaul, Sam Bresnick and Daniel Chou. “Pulling Back the Curtain on China’s Military-Civil Fusion: How the PLA Mobilizes Civilian AI for Strategic Advantage.” Center for Security and Emerging Technology (CSET), September 2025 (2,857 AI-related PLA contract notices, 2023–24; 1,560 organisations).

The National Bureau of Asian Research (NBR). “Commercialized Militarization: China’s Military-Civil Fusion Strategy.”

US Department of State. “U.S. Technology in the Military-Civil Fusion Strategy,” August 2025.

US Department of Defense. Section 1260H list of “Chinese military companies” (NDAA FY2021), update of January 2025.

Foundation for Defense of Democracies (FDD). “Commerce Department Targets Chinese Firms Engaged in Military-Civil Fusion,” September 2025.

Foreign Policy. “China: Under Xi, PLA Adopts More Civilian Tech,” October 2025.