October 7, 2026

EU Defence Series: The Neo-Prime Time. Scaling Up Europe’s Defence AI Industry

Tevian Thomson Master’s student in international economic policy at Sciences Po Paris

Many past wars show that nations that successfully adopt and integrate new military technologies gain an advantage over those that fall behind in the race for innovation. The use of drones in Russia’s war against Ukraine illustrates this dynamic: both sides develop and deploy their technology at speed and scale to shift the trajectory of war in their favour. Artificial intelligence (AI) may represent the next such transition — yet Europe risks trailing despite its ambitions for strategic autonomy.

A new generation of AI-focused firms has emerged to deliver services for key defence functions — intelligence, targeting, command and control, and logistics — and to integrate weapon systems in its own right. Such ‘neo-primes’ are a relatively recent phenomenon and contrast with traditional large defence companies, known as primes, the main contractors that develop and integrate complex military systems for governments. ^{^[1]}

The US is home to several neo-primes upending the defence and broader technology sectors, with Palantir and Anduril among the most successful examples. European governments are uneasy about relying on foreign technology for security and defence, and several have begun to seriously question their partnerships with US companies. The Dutch State Secretary for Defence has called for a fully operational alternative to Palantir, while a UK parliamentary report names such dependence an “unacceptable point of weakness.”^[2] These concerns intensified with the June 2026 restrictions on foreign access to advanced US AI models such as Anthropic’s Claude Mythos 5 and Fable 5, which showed how frontier AI capabilities can fall subject to US regulation and become a strategic vulnerability for Europe.^[3] To stay abreast of AI development, Europe needs to encourage the creation and rapid growth of its own neo-primes.

Technology

The Draghi Report (2024) identified the tech sector as a main culprit in Europe’s productivity gap with China and the US.^[4] Technological sovereignty has since been linked to the ability to build independent defence start-ups and scale them into credible partners for the armed forces. To this end, Europe has sought to develop an indigenous AI technological base.

AI is often understood in terms of a three-layer stack: (1) the bottom infrastructure and data layer (data centres, chips, energy, storage, and processing); (2) the middle or model layer covering foundation models such as LLMs as well as specialised models; and (3) the top or application layer including the end-user applications built on those models, such as ChatGPT or AI systems to analyse battlefield sensor data.^[5]

European policymakers have focused on strengthening domestic capabilities across the AI stack and the wider digital domain.

The neo-primes mostly operate in the top layer but depend on the middle and bottom layers, meaning strategic autonomy requires control over, or uninterrupted access to, all three. For instance, American Anduril has partnered with OpenAI to develop operating software for Counter Unmanned Aircraft Systems (CUAS).^[6] Systems such as Anduril’s MENACE-T show how neo-primes seek vertical integration across the stack, controlling not only defence applications but also parts of the infrastructure needed to deploy capabilities to the battlefield.^[7]

Recognising that dependence can emerge at any layer, European policymakers have focused on strengthening domestic capabilities across the AI stack and the wider digital domain — an ambition reflected in the EuroStack initiative, which seeks to build European capabilities from infrastructure to applications.^[8] EuroStack aims to reduce Europe’s reliance on American cloud providers (around 70% of the European market), expand its semiconductor manufacturing capacity (around 9% of the global market), and support the development of European foundation AI models (around 20%).^[9]

Europe’s closest answers to Anduril and Palantir are German Helsing SE and French ChapsVision.^[10] Anduril is building the FQ-44 Fury; Helsing is developing the CA-1 Europa, integrating its Centaur AI software into an autonomous combat system. ChapsVision’s ArgonOS software processes large amounts of structured and unstructured data using AI tools to support analysis and decision-making.^[11] French and German security and intelligence agencies have chosen it as an alternative to Palantir.

In the name of European autonomy, Helsing entered a strategic partnership with French Mistral in February 2025, mirroring the Anduril-OpenAI deal two months earlier.^[12] The two firms aim to elevate European defence systems through Vision-Language-Action models, letting machines such as drones analyse sensor video, interpret it, and decide what action to take.^[13] That arrangement shows Europe can link model and application layers and that European AI firms are moving beyond software towards systems integration, a role traditionally occupied by primes.^[14] Whether such partnerships can translate into products with meaningful capabilities at scale remains contingent on funding.

Money

European capital markets have fundamentally rethought defence investment since Russia’s full-scale invasion of Ukraine in 2022. As the continent has recognised that it must take greater responsibility for its own security, environmental, social, and governance restrictions have been loosened, and the European Commission, in its White Paper on European Defence Readiness in March 2025, has made it clear that such spending is not excluded from sustainable finance frameworks.^[15] The European Investment Bank has also begun investing in defence, showing public institutional capital moving with the private sector.^[16] A similar signal comes from the EU’s Multiannual Financial Framework (MFF), which has emerged as a major stream of public funding through the European Defence Fund (EDF) and the European Defence Industrial Programme (EDIP).

As a result, defence now accounts for 10% of all European Venture Capital (VC) funding, up from less than 1% in 2020.^[17] Nevertheless, the scale gap remains: of the $12.3bn from VC funds raised by end-June 2026, $11.4bn went to the US, nearly half to Anduril.^[18] R&D funding in 2020-24 paints a similar picture: EU member states spent about $40bn, while the US government invested 12 times more, $520bn.

European capital markets have fundamentally rethought defence investment since Russia’s full-scale invasion of Ukraine.

Mergers and acquisitions (M&A), meanwhile, remain the only exit strategy for investors. M&A activity hit an all-time high in 2025 as defence companies sought new segments and products.^[19] Although M&A provides investors with a clear path to liquidity and can concentrate talent, bring scale, and improve efficiency in Europe’s fragmented defence sector, relying on acquisitions hampers the growth of independent neo-primes.^[20] A VC fund invests early for a minority share and seeks an exit once the company has grown, through an initial public offering (IPO), M&A, or a secondary sale.^[21] However, with defence IPOs in Europe at zero in 2025, acquisition remains the dominant exit path. This exposes start-ups to the ‘bear hug’ risk, whereby pressure to deliver returns on investment can push newer companies towards acquisitions by established primes that absorb their technologies into existing programmes rather than allowing them to scale into independent competitors.

Anduril, however, chose a different path, itself acquiring companies such as Area-I, Dive Technologies, and Klas to build a vertically integrated neo-prime capable of competing with primes like Lockheed Martin. This was made possible because Anduril could attract larger financing rounds while providing liquidity to earlier investors through secondary share sales. European firms hoping to follow this trajectory may face greater difficulties because of smaller, more fragmented capital markets, fewer investors willing to participate in large secondary sales, and defence technology’s weaker standing as an asset class. Firms such as Helsing may be pursuing a similar strategy, but the question is whether Europe’s financial ecosystem can consistently support companies until they reach neo-prime scale.

Industrial Ecosystem and Regulation

Even technological capabilities and access to capital are not enough to produce neo-primes: start-ups also need an ecosystem that lets them scale. The first issue is fragmentation: national systems insist on their own specifications, qualification processes, and contracting rules and favour domestic companies over foreign suppliers. A start-up seeking to sell to different governments enters separate markets, each with its own conditions, requirements, and certifications. Fragmentation also affects the day-to-day operations: a European drone manufacturer’s components cross national borders, where differing regulations and approval processes create delays and administrative burden that slow production.

This slows start-ups markedly, unlike in the US, where companies grow within a single market backed by a nearly $1tn Pentagon budget.^[22] McKinsey estimates that reducing fragmentation in Europe could unlock up to €45bn by 2030 in just four segments (advanced materials, defence electronics, dual-use mechanical components, and space assets components).^[23] The EU’s Defence Readiness Omnibus of June 2025 seeks to address this fragmentation of markets and supply chains, but it may take years for its full effect to be felt.^[24]

The second issue is European procurement systems designed around traditional defence programmes that involve years of qualification testing against detailed technical standards.^[25] The rise of software-defined systems challenges this model by shifting parts of defence innovation from decades-long production cycles towards continuous updates and weeks-long cycles. Bruegel’s analysis shows that 67–90% of military procurement in Poland, the UK, and Germany still goes to the top 10 global manufacturers, compared with 36% in the US. With its low tolerance for failure, this system disadvantages the 230 defence-tech start-ups founded across Europe since 2022 that could grow into neo-primes.

The question is whether Europe’s financial ecosystem can consistently support companies until they reach neo-prime scale.

Although Europe has been slower than the US, where initiatives such as the Small Business Innovation Research programme and the Defence Innovation Unit have created alternative pathways for start-ups to access defence procurement, the reality might be changing.^[26] Ukraine’s BRAVE1 framework, where frontline units are central to identifying and testing new technologies, shows how closer interaction between soldiers and start-ups can accelerate innovation, and the EU has adopted it by partnering with Kyiv.^[27] Europeans find it difficult to replicate this model in peacetime because they must comply with NATO standards and certification requirements and maintain planning stability. However, initiatives such as the AGILE pilot instrument in 2026, which aims to move technologies from laboratory development to operational deployment within months, indicate that Europe is attempting to shorten innovation cycles.^[28]

Finally, Europe’s regulatory environment creates additional barriers to accessing the data for AI development. With the emergence of generative AI in late 2022, the EU AI Act and the EU Copyright Directive allowed rights holders to opt their content out of AI model training, reducing the training data available for LLMs by 20–30% — a good outcome for creators, but not for AI start-ups.^[29] A GovAI study of 375 LLMs released over the past eight years found that EU data protection rules have stalled or blocked 11% of advanced model launches in Europe relative to the US.^[30] The Commission has proposed the Digital Omnibus package, which seeks to simplify parts of the AI Act, data protection regulation, and related legislation, including making it easier to use personal data for training.^[31]

At the foundational layers of the AI stack, the Commission’s impact assessment for the Cloud and AI Development Act identifies environmental permits and grid connection authorisations as factors delaying data centre construction.^[32] This shows how the broader European regulatory environment constrains frontier AI development within the EU, eventually limiting what is available for top-layer defence AI applications as a second- or third-order effect, despite military exemption from the EU AI Act’s provisions.

Conclusion

Ultimately, Europe’s challenge is not a lack of technological talent or entrepreneurial ambition, but a struggle to create and sustain the financial, market, regulatory, and infrastructural conditions that allow independent neo-primes to scale rather than remain dependent on foreign technology, capital, and political considerations. The rise of AI-centric defence and security neo-primes suggests that today’s competition will be decided not only by who develops the technology first, but also by who builds the financial and industrial ecosystem to turn AI innovation into transformative military capability.

Endnotes

[1] The Tech Buzz, “Defense Tech ‘Neoprimes’ Pull Billions as Silicon Valley Challenges Pentagon Giants,” TechBuzz, 4 October 2025, updated 22 June 2026; NATO Innovation Fund, “NIF Insights: Working with Defence Primes – Key Lessons for Startups and Scaleups,” 22 February 2026.

[2] Anna Desmarais, “Why Are European Governments Reevaluating Their Agreements with US Defence Tech Contractor Palantir?” Euronews, 5 June 2026.

[3] Gabriel Gavin, “Anthropic Suspends Access to Latest AI Models Following US Order,” Politico Europe, 13 June 2026.

[4] Mario Draghi, The Future of European Competitiveness: A Competitiveness Strategy for Europe (European Commission, 9 September 2024), 5.

[5] Cole Stryker, “What Is an AI Stack?” IBM Think, 29 November 2024.

[6] Mike Stone and Aishwarya Jain, “Defense Firm Anduril Partners with OpenAI to Use AI in National Security Missions,” Reuters, 4 December 2024.

[7] Anduril Industries, “Anduril Introduces Menace-T: A Compact, Field-Deployable C4 System Designed for the Tactical Edge,” Anduril Industries, 5 May 2025.

[8] Raluca Csernatoni, “The EU’s AI Power Play: Between Deregulation and Innovation,” Carnegie Europe, 20 May 2025.

[9] Francesca Bria, Paul Timmers, and Fausto Gernone, EuroStack: A European Alternative for Digital Sovereignty (Bertelsmann Stiftung, 2025), 11.

[10] Kyle Briggs, “Signal Brief: Helsing — Europe’s Answer to Anduril,” Defense Tech Signals, 21 October 2025.

[11] Anouk Schlung, “Germany’s spy agency picks French AI firm over Palantir,” Politico, 15 May 2026.

[12] Alexandre Piquard and Olivier Pinaud, “Accord entre Mistral et Helsing : «Dans l’IA, les Européens ont toutes leurs chances» [Agreement between Mistral and Helsing: “Europeans have every chance in AI”],” Le Monde, 10 February 2025.

[13] Piquard and Pinaud, “Accord entre Mistral et Helsing.”

[14] Helsing, “Helsing and Mistral Announce Strategic Partnership in Defence AI,” 10 February 2025.

[15] Brendan O’Brien and Will Holder, “European Defence M&A: The Investment Opportunity,” Bird & Bird, 14 May 2026.

[16] O’Brien and Holder, “European Defence M&A.”

[17] François Véron and Thomas Bodin, “Defense Tech: Who Owns Europe’s Next Arsenal?” Newfund Blog, 19 May 2026.

[18] Sylvia Pfeifer and Ivan Levingston, “Wars Trigger $12bn Venture Capital Rush into Defence Tech,” The Financial Times, 21 June 2026.

[19] Kai Nicol Schwarz and Anne Sraders, “Europe’s Best-Funded Defence Techs Eye Acquisitions as They Race to Commercialise,” Sifted, 13 October 2025.

[20] Bain & Company,Defense M&A Report 2026 (Bain & Company, 2026).

[21] Bob Zider, “How Venture Capital Works,” Harvard Business Review 76, no. 6 (November–December 1998): 131–139.

[22] Gene Frieda, “Europe’s Rearmament Problem Is Demand, Not ‘Help to Buy’,” Bruegel, 7 July 2026.

[23] Hugues Lavandier, Jakob Stöber, and Rafael Ocejo, “Opportunities Through Consolidation in the European Defense Industry,” McKinsey & Company, 12 February 2025.

[24] Véron and Bodin, “Defense Tech”; “Defence Readiness Omnibus,” EU Defence Industry, European Commission.

[25] Paula Soler and Peder Schaefer, “Europe’s Defense Startups Are Losing the Rearmament Race,” The Parliament Magazine, 29 May 2026.

[26] “Move Fast and Militarise Things,” The Economist 459, no. 9496 (April, 2026): 8–9.

[27] Benjamin Howe, “BRAVE1: Ukraine’s Defence-Tech Accelerator Explained,” DSEI Gateway, 5 June 2025.

[28] Véron and Bodin, “Defense Tech.”

[29] Bertin Martens, “Catch-up with the US or Prosper Below the Tech Frontier: An EU Artificial Intelligence Strategy,” Policy Brief 25/2024, Bruegel, October 2024.

[30] Egle Markeviciute, “Data Protection Rules Slow LLM Rollout in Europe, Study Says,” Euronews, 3 July 2026.

[31] Jennifer Rankin, “European Commission Accused of ‘Massive Rollback’ of Digital Protections,” The Guardian, 19 November 2025.

[32] European Commission, Impact Assessment Report Accompanying the Cloud and AI Development Act, SWD(2026) 502 final (European Commission, 2026).

Disclaimer: The views and opinions contained in this paper are solely of its author(s) and do not necessarily represent the official position of the International Centre for Defence and Security or any other organisation.