The shift in four numbers
The two-layer AI economy
The dominant narrative around artificial intelligence has been a story of American scale — trillion-dollar foundation-model companies, hyperscaler capex cycles, and a compute infrastructure race that demands the kind of capital only a handful of US platforms can deploy. That narrative is correct but incomplete. It describes the foundation layer. It misses the application layer — and the application layer is where most of the profit will be made.
The AI value chain has two distinct competitive battlegrounds. The foundation model layer — large language models and the compute infrastructure beneath them — is capital-dominated and consolidating around a small number of US players. The application layer — the software that makes AI useful for specific industries, workflows, and outcomes — is where scale is still up for grabs. And Europe is competing here.
Europe is not losing the AI race. It is competing in the portion of that race where most of the long-run commercial value will be generated — and where its specific structural characteristics, from deep industrial expertise to privacy-sensitive enterprise customer base to a concentration of domain-expert talent, provide genuine competitive advantages. European AI application companies raised 66 cents for every dollar raised by US counterparts in 2025 — a ratio that would have been unthinkable a decade ago.
The US is winning the race to build the most powerful models. Europe is competing to win the race to make those models useful — and that is where most of the profit will be made.
| Layer | What it is | Capital profile | Frontrunner |
|---|---|---|---|
| Compute & Infrastructure | GPU clusters, data centres, power, specialist semiconductors | Hundreds of billions; consolidating to a few suppliers | US-dominated |
| Foundation Models | Training and maintaining frontier general-reasoning AI systems | Billions per year sustained compute spend | US leads; Europe via open-source & sovereign initiatives |
| Application Layer | Software that translates AI capability into specific industrial, professional, or consumer value | Domain expertise & customer access matter more than raw capital | Europe competing — and leading in many verticals |
The capital contrast between the two competitive battlegrounds is stark. Winning at the foundation-model layer requires sustained annual compute spend that exceeds the GDP of many mid-size economies. Winning at the application layer requires product intuition, customer access, and the ability to translate AI capability into demonstrable workflow improvement — skills that do not require a US zip code or a $100 billion balance sheet.
Europe's venture ecosystem — from peripheral to competitive
The structural change in European technology venture over the past decade is not marginal — it is a transformation. The ecosystem that was once a tenth of the US market in terms of capital formation and company creation has become a genuine competitive force at the application layer of the AI economy.
The pace of value creation at the application layer is genuinely without precedent. The combination of powerful foundation models, accessible API infrastructure, and increasingly capable agentic frameworks has enabled a generation of application companies to achieve enterprise-grade product maturity in timeframes that compress the previous software development cycle by an order of magnitude.
The US accounts for approximately 97% of global AI deal value in H1 2025, concentrated in the small number of frontier-model companies that require sustained billion-dollar capital infusions to remain at the cutting edge of capability development.
Despite representing only 2% of total AI deal value, Europe accounted for 23% of deal volume in H1 2025 — a sharp divergence that reflects a high-velocity ecosystem of application-layer companies raising smaller, more frequent rounds as they prove product-market fit and scale revenue at speed.
The value–volume divergence reveals something important about how the European AI ecosystem is actually developing. European companies are not raising the trophy rounds that dominate headlines — those belong to US foundation-model companies. What they are doing is building at a frequency and velocity that reflects a maturing ecosystem of application builders who understand how to create commercial value from AI capability without requiring the capital scale of a hyperscaler.
The traditional constraints that held European tech back — limited access to capital, a shortage of technical talent, fragmented domestic markets — have not disappeared. But they have loosened materially. Capital is more accessible than at any prior point in European technology history. Talent, including senior technical talent from the US, is increasingly willing to build in European cities. And the AI application layer, by its nature, rewards domain expertise and customer intimacy over raw compute scale — European strengths that the LLM race did not.
Why Europe wins at the application layer — the structural advantages
Europe's emergence as a competitive force in AI application development is not accidental. It reflects a set of structural characteristics that happen to align well with the specific demands of the application layer — and that are less relevant, or actually disadvantageous, in the foundation-model race.
| Advantage | How it manifests | Why it matters at the application layer |
|---|---|---|
| Deep industrial & professional domain knowledge | Concentration of manufacturing, financial services, healthcare, and industrial expertise creates pools of founders who understand the specific pain points AI can solve | Application value comes from correctly identifying the workflow, not from model capability alone. Domain experts build better applications. |
| Enterprise customer relationships & trust | European software companies have long-standing relationships with large enterprises that provide distribution advantages and feedback loops unavailable to new entrants | Enterprise application adoption requires trust, implementation support, and compliance assurance — all of which incumbents hold. |
| GDPR & privacy-first culture | Strict data governance creates demand for AI applications with verifiable privacy controls, on-premise deployment, and data residency compliance | Creates a protected addressable market for European vendors who can offer compliant-by-default solutions that US providers struggle to match at equivalent speed. |
| Maturing venture ecosystem | Deep-tech VC in Europe has grown from a nascent industry into a sophisticated ecosystem with funds specialising in industrial AI, defence tech, climate tech, and health AI | Sector-specialist investors accelerate the process of matching capital to the right domain-specific application opportunities. |
| Talent attraction reversing | A growing number of senior technical and commercial leaders who previously relocated to the US are returning to build in European cities; European companies are also competing for US talent | Application development requires full-stack product talent, not just AI research capability — a talent profile Europe is more competitive at attracting and retaining. |
The competitive logic is straightforward: AI application value is created at the intersection of AI capability and domain expertise. The AI capability is now widely accessible via APIs — the key differentiator is the domain knowledge that determines which workflow to target, how to embed AI into it in a way that is genuinely superior to the prior approach, and how to build the customer trust required for adoption in regulated or high-stakes enterprise environments. These are European strengths.
Incumbent software — threat or opportunity?
The public market's assessment of AI's impact on the incumbent software industry has been driven primarily by fear: the fear that agentic AI will replace the human workers whose activities generate seat-count subscription revenue, hollowing out the business models of the software platforms they use. The private market view from the leaders of those platforms is considerably more nuanced.
AI agents can execute workflows autonomously, reducing or eliminating the need for human users of seat-licensed software. As agent adoption grows, seat counts contract, and the SaaS subscription model faces structural headwinds. Incumbent software vendors are passive observers of a transition that benefits AI labs and hyperscalers at their expense.
Incumbent software companies hold what no AI startup has: years of customer trust, established distribution networks, deep implementation expertise, and the proprietary data and workflow knowledge that makes AI genuinely useful in specific enterprise contexts. AI becomes a new revenue driver — not a displacer — for platforms that can ship AI capabilities on top of the data and relationships they already own.
The critical variable is execution speed. Software leaders who move quickly to integrate agentic capabilities into their existing platforms — retooling their product around outcomes rather than sessions, shifting pricing toward value rather than seat count, and deploying the proprietary data they hold into AI models trained for their specific domain — are positioned for a revenue expansion cycle, not a contraction.
The distinction between genuine integration and cosmetic AI branding will become visible in fundamentals within 12 to 24 months. Companies that have genuinely rearchitected their products around AI will show revenue stability and net revenue retention recovery. Those that have applied AI labels to unchanged products will see the competitive gap widen as customers discover AI-native alternatives with superior outcomes.
The CEOs of enterprise software companies are not afraid of AI agents. They are racing to make AI agents work through their platforms — because that is what converts a per-seat business into an outcomes business, which is a fundamentally superior commercial model.
The defence technology surge — Europe's most unexpected tech vertical
Alongside the AI application wave, a structurally distinct and equally significant technology investment theme has emerged in Europe: defence technology. For the better part of three decades, European governments systematically reduced their defence investment as a share of GDP, outsourcing security guarantees to US military infrastructure. That era is over — and the reversal is creating one of the most capital-intensive venture investment opportunities on the continent.
The nature of modern warfare has shifted. The conflicts in Ukraine and the Middle East have made clear that the decisive technological capabilities in contemporary military operations are not tank battalions or artillery pieces — they are drone systems, electronic warfare infrastructure, autonomous platforms operating on land and at sea, and the AI-driven command-and-control software that coordinates them. These capabilities can be developed by software-driven startups, not only by traditional defence primes.
A capability deficit, now acute
European governments spent three decades systematically reducing defence budgets, creating a capability deficit that is now acutely visible in the context of continental security requirements. The gap between what European militaries can field and what the current strategic environment demands is enormous — and closing it requires technology-intensive procurement at speed.
Modern military capability is software-defined
Drone systems, counter-drone platforms, electronic warfare tools, and autonomous maritime systems are areas where agile startups can iterate faster than traditional defence primes — and where the performance gap between software-native and legacy approaches is widest and most consequential.
€5bn+ in European defence VC in 2024
European venture capital investment in defence and deep-tech defence exceeded €5 billion in 2024. Global defence-AI venture deals reached $49 billion in 2025, nearly doubling year-on-year. Governments and defence procurement agencies are actively engaging with startups at a speed and openness that would have been unthinkable five years ago.
Not prototyping — scaling
European defence technology startups are not in the prototyping phase — they are scaling. Companies focused on counter-drone systems, autonomous maritime platforms, and AI-driven defence software are raising growth capital and signing government contracts at a rate that reflects a defence procurement community that has abandoned its historic preference for incumbent suppliers when faster alternatives are available.
The defence technology sector represents a rare combination of rapidly expanding addressable market, government-backed demand, and genuine technological differentiation opportunity. European startups in this space benefit from proximity to the strategic context driving procurement decisions — the war in Ukraine is not an abstraction but a live laboratory for understanding what works and what does not in modern conflict.
The shift in investor attitude toward defence technology has been equally significant. A category that was once considered off-limits by many European venture funds — on ethical, reputational, or dual-use grounds — has moved from the margins to the mainstream. The firms now actively building in this space are attracting capital from generalist tier-one venture funds that previously declined to participate, reflecting both the scale of the opportunity and a broader reassessment of the societal value of defence capability.
The new capital structure — staying private, exiting through M&A
The conversation about European technology exits has long been dominated by the IPO — seen both as the definitive proof of success and as a bellwether for the health of the entire ecosystem. That framing has become increasingly disconnected from how companies are actually structuring their capital journeys in the current environment.
Two connected dynamics are reshaping the exit landscape. First, a growing cohort of European technology companies is actively choosing to remain private for longer — not because they cannot access public markets, but because they judge that the private market offers better conditions for executing on their strategic plans than the near-term public market would. Second, M&A — which has always been the statistical primary exit mechanism for most technology companies — is expected to become more active as AI drives a consolidation cycle across enterprise software.
| Trend | What's happening |
|---|---|
| The strategic private premium | Companies with strong revenue growth, clear competitive positioning, and patient capital backing are increasingly judging that public-market frameworks — quarterly earnings pressure, short-term guidance cycles, macro-driven multiple compression — are incompatible with executing on the multi-year product roadmaps that AI-era leadership requires. Private markets offer the strategic flexibility many of these companies need. |
| AI-driven M&A consolidation | AI is accelerating M&A across enterprise software in two directions simultaneously. Incumbent software platforms are acquiring AI-native capabilities to accelerate their own AI product development. AI-native challengers — having demonstrated product-market fit — are themselves attractive targets for larger platforms seeking to integrate their technology and customer base. This creates a more active near-term exit environment than a narrow IPO focus would suggest. |
| IPO window — selective, not closed | Public markets are not unavailable — they are selective. Companies that can demonstrate durable revenue growth, clear AI integration strategies, and a path to improving unit economics at scale are finding receptive investor audiences. The bar for accessing public markets has risen — and many companies that might previously have IPO'd early are now waiting for the business to mature further before doing so. |
| Government-backed capital | In defence technology and sovereign AI infrastructure, European governments have become significant capital providers — both directly through procurement contracts and indirectly through investment vehicles supporting strategic national capabilities. This government-backed capital is changing the financing structure for some of Europe's most strategically important technology companies, reducing their dependence on traditional venture timelines and exit pressures. |
The practical implication for investors is that the volume of European technology value that becomes accessible through public markets at any given moment underestimates the actual creation of value occurring in the private ecosystem. Strategies that can access the private market — through growth equity, pre-IPO rounds, or M&A activity — are better positioned to capture European technology momentum than those restricted to listed equities alone.
Europe's moment — if the execution follows the opportunity
The conventional narrative about Europe and artificial intelligence has been one of playing catch-up. That framing captures part of the truth: Europe does not have the compute infrastructure, the foundation-model champions, or the capital concentration to compete at the model layer against the US tech platforms that have invested over a trillion dollars in AI capability since 2022.
But the application layer — the software that converts AI capability into genuine commercial value for specific industries, workflows, and customer outcomes — operates by different rules. Domain expertise matters more than compute. Customer relationships matter more than parameter counts. Execution velocity matters more than research prestige. And on these dimensions, a generation of European founders and investors is competing effectively.
The data supports the thesis. More than 400 European tech unicorns. A capital-to-US parity ratio that has improved tenfold in a decade. A new cohort of AI application companies reaching $100 million in ARR faster than any prior generation of enterprise software. A defence technology ecosystem moving from curiosity to mainstream venture priority. And a group of incumbent software companies — many of them European — that hold the customer trust and domain data to make AI work in the enterprise environments where most commercial value is actually generated.
The risks are real. The question of whether European AI application companies can sustain their early momentum as foundation models become more capable, as large language models begin competing vertically in the application layer, and as US platforms deploy their distribution advantages into application categories remains open. The execution advantage of fast-moving European teams can erode if they do not continue to deepen the domain integration and proprietary data moats that currently protect their lead.
But the structural opportunity is equally real — and the evidence that European entrepreneurs and investors are seizing it is accumulating. The race for AI leadership has more than one track. Europe has chosen the right one to compete on.
The US built the engine. Europe is building what the engine powers — the specific, domain-deep, outcome-oriented applications that turn AI from a technological achievement into an economic reality. That is the layer that will generate most of the profit. And it is the layer where Europe is competing to win.
This report has been prepared by Lualdi Advisors for informational and educational purposes only. All data, statistics, and market observations cited in this report draw on publicly available sources including industry reports published by Atomico, Accel, CB Insights, EY, SIPRI, and publicly available press coverage. Lualdi Advisors has not independently verified all third-party data. References to companies, sectors, and market categories are illustrative and analytical in nature; they do not constitute recommendations to invest in or avoid any specific company, security, or sector. This material does not constitute investment, legal, tax, or financial advice and should not be used as the basis for any investment decision. Lualdi Advisors makes no representations regarding the accuracy or completeness of third-party information referenced herein. Past performance and historical trends are not indicative of future results. Forward-looking statements are inherently uncertain; actual outcomes may differ materially.
