Fourteen main battle tank models. That is what Europe fields against America's one, alongside 23 howitzer variants to America's two and 16 submarine types to four in the US. The Kiel Institute put those numbers on the page in March, and they are the single most important thing to understand about European defence right now.
Because the story everyone tells is a different one. It is a story about money.
Here is the version you have heard: European defence-tech raised $8.7 billion in 2025, a 55% jump year over year — genuine acceleration. And yet Anduril, one American company, generated $2.1 billion in revenue in the same year. One firm's annual sales, nearly a quarter of an entire continent's venture pool.
The symmetry is seductive. It reads as a funding shortfall, a capital gap, a problem you solve by writing bigger cheques.
It is the comfortable misreading.
Because Europe is not short of defence money. Per the Kiel Institute, European defence spending stands at roughly $550 billion — the world's second largest after the US, ahead of China, roughly three times Russia's. In 2025 that was already about 60% of total US spending. Europe's contribution to NATO is substantial. The inputs are there.
What Europe gets for those inputs is fourteen tank models.
So the gap is not money in. It is capability out. Europe fragments its procurement along national lines, sustains an artisanal defence industrial base, and condemns itself to low volumes, high unit costs, and limited interchangeability. The Kiel authors are blunt: "Europe's capability gap is not financial, but real."
Europe's capability gap is not financial, but real. The continent fields fourteen tank models against America's one — the gap is capability out, not money in.
Which reframes the whole venture story. Anduril's $2.1 billion is not a funding-density problem. It is a production problem. Anduril ships autonomy stacks that get fielded. Most European defence capital sits in pre-revenue or Series A rounds. Europe is building venture portfolios; the US is building defence companies.
Look at where Germany's money actually went and the diagnosis sharpens. Around 95% of the Bundeswehr's €100 billion Special Fund was committed to traditional crewed platforms. Production capacity increases were slow and small.
And R&D? The share of the German defence budget going to research and development stagnates at around 2% — less than a fifth of the US share.
That is the trap in miniature. Even the money Europe does spend flows to legacy platforms, not next-generation capability or the industrial scale to field it. The €800 billion ReArm headline mostly buys ammunition, tanks, artillery, and the logistics to support them — the gaps two years of watching Ukraine exposed. AI and autonomy are a smaller line item bolted on top.
Now test the sovereignty plays against that reframe.
Start with the flagship signal. DTCP Defence — a newly launched €500 million fund dedicated exclusively to defence, security and resilience — led Six Robotics' €12 million seed for swarm autonomy software. The Oslo company works with the Norwegian Defence Research Establishment and builds hardware-agnostic drone swarming, positioned as a force multiplier in contested environments.
Its founder's pitch is the sovereignty thesis distilled: "Europe cannot afford to depend on others for the software that defines what its forces can do."
It is a coherent bet. Software-first, platform-agnostic, interoperable across systems — arguably the layer the Kiel authors point to when they flag Europe's neglect of autonomous systems, and, per the AIDEF blog, the "buy stacks, not platforms" logic that pushed Helsing to a reported $18 billion valuation. If Europe can't out-produce Anduril on airframes, own the decision loop instead.
Then there is the efficiency bet. Verkko Robotics' VOLTAIC claims 98% less energy than a traditional LLM, roughly 1/50th the cost per query, engineered to run on private hardware rather than hyperscale datacentres controlled by US providers.
The framing is explicitly sovereign. By default, queries to US AI providers move data outside EEA jurisdiction — a dependency contracts can manage but not eliminate. Von der Leyen's European Technological Sovereignty Package, unveiled 3 June, makes the same argument: "We cannot afford to depend on others for the technologies that keep our hospitals running, our energy grids stable and our services secure."
Efficiency reads as sovereignty here for the same reason it reads as discipline in the wider AI market — the gets-paid-now, less-datacentre pitch against the promises-payback-later capital pile. If you can't match the compute, change the economics.
And there is the access bet. Aire, the stealth lab from Meta and OpenAI alumni reportedly raising $400 million, will be based across the US, France and Israel. That is not hedging by accident — it is the Hong Kong argument in another register: sovereignty depends on accessing global networks while retaining control over IP, governance and vision. Industrial sovereignty, the argument goes, requires the opposite of isolation.
Three bets — leapfrog scale with efficiency, own the software layer, buy access across borders. Each is intelligent. Each treats the scale problem as something you can route around with innovation.
Here is why the routing may not work: none of these becomes sovereignty until venture rounds convert into fielded capacity. And that conversion is exactly what European procurement is built to prevent.
Follow the money's path. The European Defence Fund co-funds R&D through multinational consortia — three entities from three member states minimum — and explicitly does not fund procurement. It builds capability and relationships, not orders. That is the most misunderstood feature of the flagship instrument.
EDIP is the attempted bridge: procurement mechanisms that let a company invest in production capacity ahead of confirmed contracts. That bridge is precisely what separates a validated startup from a fielded one. Whether it carries real weight at scale is unproven.
Meanwhile fragmentation reasserts itself the moment stakes get high. FCAS — the Franco-German-Spanish sixth-generation fighter — burned €3.2 billion through 2025 for a first aircraft projected in 2045, before Belgium's defence minister declared "SCAF is dead" and Merz began questioning whether Berlin needs it at all.
The rival GCAP consortium — reportedly the UK-Italy-Japan group, per CES Intelligence — is building a competing sixth-gen fighter with a demonstrator reportedly due in 2027. So Europe is now funding two next-generation fighters in parallel, neither fully funded — the fourteen-tank pathology, live, at the frontier.
That is the tell. The moment cooperation is most needed, national interest fractures it. Bruegel's math is that integrated markets plus scaled procurement could halve unit costs — and that the existing incentives are simply too small to overcome home bias.
Contrast the American environment these bets compete against. The 2026 NDAA codifies multi-year munitions procurement and ties executive pay to production metrics. The Pentagon takes direct equity — 15% of MP Materials, a $1 billion investment in L3Harris to rebuild rocket-motor capacity. JPMorgan committed $1.5 trillion over ten years to security and resilience.
That is demand visibility five to ten years out, with federal capital de-risking capacity expansion. It is a machine built to convert money into fielded volume.
Europe's is built to convert money into R&D portfolios and national champions. DTCP's €500 million, VOLTAIC's efficiency, Aire's cross-border structure — these are rational hedges against a scale gap. But a hedge is not a fix.
The binding constraint was never the size of the venture pool. It is that the continent funds invention and fragments demand, then calls the resulting dependency a sovereignty problem — as if another fund would solve what fourteen tank models diagnose.
Sovereignty becomes narrative precisely where procurement fragmentation prevents scale. Until EDIP-style bridges and pooled demand actually turn a €12 million seed into fielded swarms, the seductive numbers will keep pointing at the wrong gap.