“A Gigafactory does not ultimately produce megawatts. It converts megawatts into intelligence.”
The race is moving from "who builds the data center?" to "who owns the intelligence stack inside it?" Three developments since the previous check materially deepen the European Gigafactory thesis. The most important is not a new site — it is the emergence of a European semiconductor-capital stack around the AI Factory network.
Dutch AI-chip company Axelera AI said today it has signed multiple contracts to supply chips to AI factories while launching its second-generation Europa accelerator, to be integrated into selected Dell and Supermicro systems. Crucially for Europe, Axelera is already working with Dell and E4 as a supplier to the EU-backed IT4LIA AI Factory in Italy and MeluXina in Luxembourg. At the same time, another Eindhoven company, EUCLYD, announced a €200+ million Series A to develop an alternative AI-compute architecture spanning processors, memory and data-center systems — co-led by Samsung, Somerset Capital Partners, the EQT-managed Scaleup Europe Fund and Innovation Industries, with former ASML CEO Peter Wennink as chairman.
These are not merely semiconductor stories. They alter the economics of the European Gigafactory race.
The Second-Order Effects
four consequences that redraw the map
-
Our framework has concentrated on CAPITAL → POWER → GRID → PERMIT → CONSTRUCTION → OPERATING MW. Today another layer becomes unavoidable: OPERATING MW → SILICON → SYSTEMS → SOFTWARE → MODELS → TOKENS. Europe could succeed at constructing enormous AI campuses and still transfer much of the economic rent abroad. That creates two sovereignties: infrastructure sovereignty (the data center sits in Europe) and economic sovereignty (Europe captures a meaningful share of hardware, software, IP and cash flows). The second is much harder — and it is exactly what Axelera and EUCLYD attack. signal: The strategic prize is economic sovereignty, not just a European postcode for the campus.
-
If European AI Factories become early customers for emerging European semiconductor companies, public infrastructure serves as both compute capacity and anchor customer. The causality chain runs: public procurement → reference installation → technology validation → lower customer-risk perception → private orders → scale → lower unit cost → stronger chip ecosystem. Europe need not reproduce Nvidia overnight; it can use its €30+ billion programme — with up to €10 billion of public funding designed to mobilise at least €20 billion of private capital — to create demand pull for European alternatives. IT4LIA is therefore not merely compute in Bologna; it can become a semiconductor commercialization platform. signal: Public AI Factory procurement is the demand engine for a European chip ecosystem.
-
Europe remains far behind the US in frontier-model training. But inference economics are different: they reward energy efficiency, latency, specialization, deployment close to industry and predictable operating costs. That plays into European strengths — industrial customers, telecoms, automotive, robotics, manufacturing, regulated enterprises and sovereign workloads. Axelera's Europa targets enterprise workloads with Titania planned for data centers and supercomputers; EUCLYD attacks inference through processor-memory and system-level optimization. The asymmetry: America may dominate frontier training while Europe builds competitive advantage in industrial inference. signal: Turn European factories, banks, utilities and machines into distributed inference markets.
-
Christine Lagarde supplied the strongest macro-financial framing yet: Europe hosts only about 5% of global AI computing capacity against roughly 75% in the US, and the capacity shortfall could widen from around 3 GW in 2025 to about 20 GW by 2036 — with eliminating a 19-GW gap costing as much as €600 billion including servers and chips. Euro-area companies are expected to devote roughly 10% of investment to AI in 2026, and AI-related borrowing was about a quarter of the growth in corporate credit in Q1 2026. Bank lending alone cannot finance this; it will require equity, bonds, private credit and deeper capital markets. signal: The AI supercycle is becoming the project around which Europe may have to build its capital markets.
Spain: The Regulatory Debate Is Becoming More Technical
and therefore more solvable
From Bankable MW To Bankable Intelligence Capacity
the new metric
Country Implications
where the strategic weight is shifting today
Italy
Moves up strategically. IT4LIA's importance rises because it is not only an AI Factory but potentially an anchor market for European semiconductor suppliers such as Axelera — linking Bologna's compute ecosystem to European semiconductor industrial policy.
Spain
High-potential, high-policy-risk. Today's APECDATA proposals show the conflict may still convert into a workable architecture. Spain also holds Semidynamics, an inference-chip competitor alongside Axelera and EUCLYD — making it vital to connect domestic semiconductor capability with the Gigafactory project rather than treat them separately.
France & Germany
No comparably material new site, financing or grid announcement today. The implication is clear: their Gigafactory consortia should begin treating European silicon procurement as part of industrial sovereignty, not merely construction.
Nordics
Strongest power proposition, but accelerator-level energy efficiency is now another competitive variable. The cheapest megawatt matters — so does how many economically valuable inference tokens can be produced from it.
the fourth race begins
- The first race was for sites, the second for power, the third for capital — the fourth is for value capture inside the AI Factory.
- Axelera's AI Factory contracts and EUCLYD's €200+ million financing show European capital forming around the semiconductor layer as Europe prepares tens of billions in sovereign AI infrastructure.
- The AI Gigafactory race is now industrial + energy + semiconductor + banking + capital-markets policy simultaneously.
- Infrastructure sovereignty is achievable; economic sovereignty is the harder, more consequential prize.