The Atlantic Gap: Defence Research Infrastructure in the US and Europe

The chart below plots annual budget against staff headcount for major US and European defence research institutions. Both axes are necessary: a small organisation with a large budget may contract out its engineering; a large workforce with a modest budget may lack the systems integration depth of a smaller, more focused laboratory.

The US institutions are not simply larger. They occupy a different region of the chart — high budget, large staff — because they are doing something structurally different: sustained systems engineering across decades, with certification authority, classified access, and procurement integration built in. The NNSA weapons laboratories alone (Sandia, Los Alamos, Livermore) employ 36,500 people on a combined $14.5 billion budget. AFRL, the Air Force Research Laboratory, runs a $9.5 billion core programme across eight technical directorates — larger by budget than any individual NNSA laboratory, and absent from most discussions of American defence research capability.

The European institutions cluster in the bottom left. France's CEA, shown as the full organisation, sits near the NNSA complex in budget terms but covers a much broader set of activities including civil nuclear. The UK's DSTL, AWE, and QinetiQ — the most substantive European defence lab cluster — together employ roughly 16,500 people but at a fraction of US scale and without the integration between institutions that characterises the American complex.

The five institutional categories below are analytically distinct. Collapsing them into a single "defence R&D" category produces misleading comparisons. A basic science institution (Fermilab, CERN) does different work from a mission-oriented weapons laboratory (Sandia, CEA DAM). An industry-applied research unit (Fraunhofer) is structured differently from a federally-directed engineering centre (AFRL). The table maps coverage across states.

The deepest analytical distinction is not the number of institutions or total spending. It is in how institutions connect research to procurement and operational deployment. The US national laboratory system occupies a stage in the weapons development pipeline that most European states have not built: sustained engineering development between research and production.

The existing EU and NATO mechanisms address different stages. DIANA identifies and funds early-stage technology in startups and deep-tech companies — it is an accelerator, not an engineering institution. The European Defence Fund provides project-based collaborative R&D; when a funded project ends, the engineering capacity assembled to execute it disperses. CCDCOE produces doctrine, legal frameworks, and exercises. None of these mechanisms creates or sustains the standing engineering layer shown below.

The DARPA model is frequently mischaracterised in European policy discussions. DARPA employs approximately 220 programme managers on a $4 billion budget and holds no laboratories. Programme managers rotate every three to five years and hold authority to fund research from concept through technology demonstration. DARPA is a funding mechanism with transition authority — it works because the standing engineering institutions (the national laboratories, service labs, and FFRDCs) exist to receive and develop what DARPA transitions. Replicating DARPA without the laboratory complex replicates only the funding mechanism.

The United Kingdom's decision to split the Defence Evaluation and Research Agency (DERA) in 2001 into DSTL (retained, 4,500 staff) and QinetiQ (privatised, now 6,000 UK staff) is instructive. Of QinetiQ's £328 million in R&D spending in FY2024, only £12.8 million — 3.9% — was internally funded. The remainder was government-contracted. Privatisation shifted defence engineering capacity from an institutional model to a contracting model. When contract flows change, capability disperses. The DERA split created a cautionary case for any European state considering equivalent restructuring.

The country-by-country analysis reveals four structural models, not a single "European" pattern. The variation is itself analytically significant: it means that proposals for EU-wide defence research policy must navigate a member state landscape with fundamentally different institutional starting points.

Per capita comparisons require methodological care. This chart makes a specific and limited claim: it compares the scale of identifiable public applied defence and security R&D institutions across selected states. The US figure covers NNSA weapons laboratories and DARPA only — a deliberately conservative subset that excludes US service laboratories (~$11 billion annually), FFRDCs, and DOE applied labs. Even this conservative subset substantially exceeds any European comparator. European figures are drawn from EDA-reported national defence R&D aggregates (2023 data), which include government contracts to industry and universities alongside standalone government institutes — meaning the US figure and European figures are not methodologically identical. The gap is directional, not precise.

Four institutional models operate across EU member states. France, Sweden, Finland, and the Netherlands maintain government lab models with direct mission linkage and procurement integration. Italy, Spain, Denmark, and Germany concentrate defence engineering in large industrial companies with government equity or procurement dependency. Poland, Greece, the Czech Republic, Romania, and Bulgaria are spending heavily but lack the engineering infrastructure to absorb that spending into domestic capability at scale. Belgium, the Baltic states, Austria, Portugal, and smaller member states have minimal or constrained research capacity.

The United States maintains the government lab model at a scale that places the European government lab states in the same institutional category but not the same order of magnitude. France's CEA DAM and ONERA combined represent a fraction of AFRL alone by staff and budget. Sweden's FOI at 1,500 staff compares by type with the Naval Research Laboratory at 3,000 staff, but no European state has an institution equivalent to the Army Research Laboratory, Johns Hopkins APL at 8,800 staff, or MITRE at 10,000 staff.

The structural gap is not that Europe has fewer laboratories. It is that the government lab model, where it exists, operates at substantially smaller scale than its American counterpart, and the majority of EU member states — including Germany, the largest economy — have not adopted the model at all. The existing EU and NATO coordination mechanisms (EDF, DIANA, CCDCOE) address early-stage innovation and doctrine. None addresses sustained engineering development, demonstration infrastructure, and mission continuity across programme cycles. This is the function that US national laboratories have performed since the Manhattan Project, and it is the function that Europe has not yet built at equivalent scale.

France and the United Kingdom demonstrate that the institutional model works in a European context. CEA DAM at roughly 4,500 staff represents perhaps one-eighth the combined scale of the three NNSA weapons laboratories. DSTL at 4,500 staff covers a far broader domain range than any single US laboratory of equivalent size, which reflects funding constraints rather than superior efficiency. The question is not whether the model is viable in Europe — it is — but why it has not been extended beyond the nuclear weapons context, and whether the current security environment provides sufficient political impetus to change that.