The Rise of European Defence-AI: How Helsing, BAE, Thales, and Saab Are Rewriting the Prime Stack

Helsing and the Talent Asymmetry

Founded in 2021 by Gundbert Scherf and Niklas Kuehne, Helsing was explicit from the start about its ambition: to become the primary AI software layer for European defence. The company has grown to more than 500 employees across Munich, London, and Paris, almost all of them machine learning engineers or defence-domain specialists. That headcount figure is modest by the standards of a Thales or a BAE Systems, but the ratio of ML researchers to total staff is orders of magnitude higher. Helsing is not trying to build a fighter jet; it is trying to own the intelligence layer that makes the jet effective.

The company's first major programme win, a contract to provide AI-enabled electronic warfare software for the Eurofighter Typhoon, announced in 2023, demonstrated that European air forces were prepared to source their most sensitive software from a firm that did not exist three years earlier. That willingness to break from incumbents would have been almost unthinkable before the invasion of Ukraine changed the political calculus around procurement speed. Helsing has since expanded its portfolio to include sensor fusion software, autonomous systems decision support, and what the company describes as sovereign AI infrastructure for government clients.

The talent asymmetry is stark. BAE Systems employs tens of thousands of engineers across its global programmes, but the company's own 2024 annual report acknowledged that digital and AI skills remain a strategic priority for recruitment. Thales, in its investor briefings, has been candid that competition for machine learning specialists in France and the United Kingdom is intense and that retention is an ongoing challenge. Neither company is standing still, but both are fighting against a structural disadvantage: the most talented AI engineers in Europe have more options than ever, and writing targeting algorithms for a legacy prime is not always the most attractive of them.

BAE Systems: Buying Time With Partnerships

BAE Systems' response to the AI-native insurgents has been pragmatic rather than transformational. The company has invested heavily in its own AI and autonomy capabilities through its Air sector, most visibly in the context of the UK's Tempest programme, now rebranded as the Global Combat Air Programme. BAE is the lead systems integrator for GCAP, which means it retains the relationship with the UK Ministry of Defence that matters most for the next generation of combat aviation. But the company is increasingly reliant on partnerships and acquisitions to fill the gap between its own AI capabilities and what the programme demands.

BAE's acquisition strategy has been telling. The company purchased AI and autonomy specialist firms in both the United Kingdom and Australia over the past four years, and its Digital Intelligence division, headquartered in Guildford, has grown to become one of the larger cyber and AI units in the European defence sector. The division works across intelligence community contracts, data analytics, and electronic warfare software. It is capable and well-funded. But it operates within the cultural and process constraints of a company that has been building physical platforms for decades, and those constraints are not trivial. Programme governance structures designed for hardware development do not always accelerate software iteration.

The more interesting question for BAE is whether its role as prime systems integrator on GCAP becomes a choke point or a springboard. If the company can position itself as the trusted integrator that pulls Helsing's algorithms, Leonardo's sensors, and its own electronic warfare systems into a coherent, certified whole, it retains enormous value. If Helsing or a successor firm eventually develops the certification track record and the sovereign infrastructure credentials to deal directly with governments, BAE's integration premium shrinks considerably.

Thales and the Cortex Bet

Thales has taken a more architecturally deliberate approach. The French prime has invested significantly in its Cortex AI platform, which it describes as a common AI infrastructure layer across its defence, aerospace, and security business lines. The pitch to customers is integration without vendor lock-in: a platform that can ingest data from heterogeneous sensors, run inference workloads at the edge, and push outputs to decision-support interfaces, all within a certified and sovereign environment.

Cortex is not a single product; it is a family of capabilities that Thales has assembled through a combination of internal R and D, acquisitions, and partnerships with European cloud providers. The company has been explicit in investor day presentations that it views AI not as a feature to be added to existing products but as a platform business in its own right. Thales Chief Executive Officer Patrice Caine has framed the company's digital transformation as existential rather than optional, a framing that is increasingly common among European primes but that Thales has backed with more concrete capital allocation than most.

The weakness in the Cortex story is certification and deployment speed. Building a common AI platform across the diverse regulatory and security frameworks of France, the United Kingdom, Germany, and their respective allied partners is genuinely hard. Each sovereign customer has distinct data sovereignty requirements, distinct certification processes for safety-critical AI, and distinct procurement rules. Thales has the relationships and the cleared facilities to navigate those requirements. It does not always have the software engineering velocity to meet them at the pace that operational commanders now demand.

Saab and the Platform Lock-In Play

Sweden's Saab has pursued a different strategy again. Rather than racing to build a general-purpose AI platform, Saab has focused on embedding AI capabilities deeply into its own platforms: the Gripen fighter, the GlobalEye airborne surveillance system, and the ground-based combat management systems that have found buyers across Nordic Europe and beyond. The logic is platform lock-in: if Saab's AI is inseparable from the platform, customers who buy the platform also buy the AI, and the upgrade cycles that follow keep Saab in the programme for decades.

This approach has a strong commercial logic in markets where Saab is already the incumbent. Sweden's own defence forces, and several NATO partners that operate Gripen, represent a captive base for AI-enabled upgrades. Saab has also benefited from the surge in European defence spending following Sweden's NATO accession in 2024, with procurement budgets for ground-based air defence and electronic warfare expanding across the Nordic and Baltic theatres.

The vulnerability in Saab's position is that platform lock-in works until a sovereign customer decides that the AI layer is more critical than the platform and starts demanding open architectures. That pressure is already visible in UK Ministry of Defence procurement guidance, which increasingly emphasises open standards and modular architectures for defence software. If Saab's AI is too tightly coupled to its own hardware, it may find that customers start specifying requirements in ways that force decoupling, at which point the competitive dynamics look more like those facing BAE and Thales.

The European Defence Fund and the Capital Question

Underlying all of these competitive dynamics is a significant injection of public capital. The European Defence Fund, administered by the European Commission, has disbursed hundreds of millions of euros into collaborative defence R and D across EU member states since it reached full operational capacity. A substantial portion of those disbursements has flowed into dual-use AI research, autonomous systems development, and sovereign AI infrastructure projects. The Fund's collaborative structure, which requires participants from at least three EU member states, has created an incentive for exactly the kind of coopetition between primes and AI-native firms that the market is now beginning to produce.

Leonardo, the Italian prime, has been one of the more active participants in EDF-funded collaborative projects, frequently appearing alongside Thales and smaller AI-specialist firms in consortium bids. Saab, as a non-EU NATO member following Sweden's accession, faces a more complex relationship with EDF funding structures, which creates a subtle but real competitive asymmetry with its EU-headquartered rivals on programmes with EDF backing.

The capital question matters for Helsing specifically because building and operating sovereign AI infrastructure, the kind that can run classified workloads at scale inside government networks, requires data centre investment and security certification expenditure that is genuinely expensive. Helsing raised significant venture capital in its early rounds, including from prominent European and transatlantic investors. But as the company moves from software licensing toward sovereign infrastructure contracts, its capital requirements will grow, and the question of whether it partners with, or is eventually absorbed by, a prime contractor becomes increasingly live.

Coopetition: The Only Viable Endgame

The trajectory of all four companies points toward the same structural outcome: coopetition. The legacy primes have the contract relationships, the cleared facilities, the certification track records, and the political relationships with procurement ministries that AI-native firms cannot replicate quickly. Helsing, and firms like it, have the engineering talent, the software velocity, and the architectural thinking that primes cannot easily hire their way into acquiring. Neither side can win cleanly without the other.

The coopetition model is already visible in embryonic form. Helsing's Eurofighter software work puts it in a supply chain relationship with the Eurofighter consortium, which includes BAE Systems. Thales has structured Cortex explicitly as a platform that third-party AI developers can build upon. These are not yet deep partnerships, and the commercial tensions remain significant, but the direction of travel is clear.

What will determine which companies capture the most value in this new structure is not who writes the best algorithm today, but who controls the data infrastructure, the certification standards, and the programme integration role tomorrow. The primes have a structural advantage in the latter two. Helsing and its peers have an advantage in the first. The European defence AI market is large enough, and growing fast enough, that both can win. The question is the terms on which they share the stack.

The scale of the shift underway in European defence AI becomes clearer when set against the concrete financial and operational data now emerging from the companies and institutions involved. The figures below frame the competitive landscape and the capital flows that are reshaping it.