On July 30, 2026, the European Commission opened a formal call for bids to build seven AI gigafactories across the EU, launching a €30 billion initiative designed to break Europe’s dependence on American and Chinese computing infrastructure. For enterprise AI teams, this is not a distant policy story. It is a structural change to the global AI compute landscape with direct implications for where you train models, where you run inference, and how you price AI deployment over the next three to five years.
What Happened
The European Commission formalized its InvestAI initiative on July 30 with the launch of a competitive tender for up to seven AI gigafactories to be built across EU member states. The program commits €10 billion in public funding, drawn from EU budgets and national government contributions, and targets an additional €20 billion in private investment, bringing the total to €30 billion.
According to Bloomberg and the Washington Post, each planned gigafactory will house at least 100,000 of the most advanced AI chips available, making each facility approximately four times more powerful than current EU data centers. The initiative is structured to deliver shared compute infrastructure accessible by organizations across member states, not private capacity for a single operator.
Henna Virkkunen, the Commission Executive Vice President overseeing tech sovereignty, framed the announcement in explicit strategic terms: “Access to the raw scale of computing power within AI gigafactories is a strategic necessity for Europe as AI development accelerates.”
Bidding closes November 12, 2026. Award decisions are expected in early 2027, with construction starting the same year.
The Funding Structure
The call is organized into two lots, and the per-project funding scales significantly by phase:
| Lot | Early-Stage Funding | Phase 2 Funding | Target |
|---|---|---|---|
| Lot 1 (smaller) | Up to €100M per project | Up to €400M per project | Distributed across more sites |
| Lot 2 (larger) | Up to €200M per project | Up to €800M per project | Fewer, higher-capacity sites |
The funding structure is designed to de-risk private investment. Public money covers early-stage costs (land, planning, grid connections, initial infrastructure), with the expectation that private capital enters at phase two as the commercial model is established. The Commission’s target ratio is approximately one euro of public funding for every two euros of private capital.
The InvestAI initiative connects to a broader EU ambition: the Commission previously announced a goal to mobilize €200 billion in total AI investment across the bloc. The gigafactory call is the most visible single deployment mechanism for that ambition.
Who Can Access Compute
The gigafactories are explicitly designed as shared infrastructure, not private hyperscaler facilities. According to the European Commission’s official program page, access is intended for:
- AI startups and scale-ups
- Small and medium-sized enterprises (SMEs)
- Large industry enterprises
- Academic and research institutions
- Public authorities across EU member states
The intended use cases span the full AI development lifecycle: training new foundation models, running inference at production scale, and fine-tuning existing models on proprietary enterprise data.
This distinguishes the gigafactory program from the broader EuroHPC AI Factories network already operational across Europe. EuroHPC’s 19 AI Factories and 13 AI Factory Antennas provide subsidized access to smaller compute clusters, primarily serving research workloads and AI prototyping. The gigafactories are designed to handle the scale of production-grade AI training and inference, which EuroHPC facilities cannot currently support.
Why This Matters for Enterprise AI Strategy
Compute Access Is the Bottleneck That Is Rarely Discussed
Most enterprise AI conversations focus on models, prompts, and orchestration. The infrastructure layer is treated as a commodity: pick a cloud provider, pay the token bill. But at the scale of foundation model training or high-volume inference, compute access is not a commodity. It is allocated by relationship, negotiation, and geography, and European enterprises have been at a structural disadvantage.
Currently, European enterprises that need to train or fine-tune large models at scale have three real options: negotiate enterprise agreements with US hyperscalers (AWS, Google Cloud, Azure), use US-domiciled compute at risk of data sovereignty issues, or wait in queue for EuroHPC allocations that are sized for research, not production. The gigafactory program, when operational, adds a fourth option: EU-resident, production-grade, compliant shared compute.
Regulatory Fit Changes the Calculus
The EU AI Act, now in full effect, creates compliance obligations that interact directly with where AI computation runs. High-risk AI systems used by European enterprises face documentation, audit, and traceability requirements that are considerably easier to satisfy on EU-resident infrastructure. Sovereign AI for regulated industries (financial services, healthcare, public sector) is not a nice-to-have. It is increasingly a legal requirement.
Microsoft and Mistral’s sovereign AI partnership from July demonstrated how this pressure is already reshaping compute contracts: enterprises in regulated EU markets are paying a premium to keep workloads on European soil. The gigafactory initiative makes that option available at training scale, not just inference scale.
The Global Compute Power Shift
The EU’s announcement is part of a broader global realignment of AI compute investment. South Korea’s $880B 10-year AI investment plan announced in July 2026 follows a similar logic: national compute sovereignty as a prerequisite for competitive AI capability. The US approach has been to concentrate frontier compute in private hyperscalers and frontier AI labs. The EU is building a public infrastructure layer designed to prevent that same concentration from repeating in Europe.
For enterprise AI buyers, this is a purchasing signal. In 2025, there was one credible infrastructure choice for large-scale AI training: US hyperscalers. In 2026, AMD and Anthropic’s 2GW compute partnership demonstrated that even frontier AI labs are diversifying away from single-vendor dependency. By 2028, the competitive compute landscape will include EU gigafactories alongside US hyperscalers, making multi-region AI infrastructure strategy a standard enterprise planning requirement rather than an edge case.
What Enterprise Teams Should Do Now
Audit your data residency obligations before compute contracts renew. If your enterprise operates in EU member states and handles personal data, GDPR and the EU AI Act interact with where you run training and inference. Enterprises that renew multiyear hyperscaler agreements now, before EU gigafactories are operational, may find those contracts create compliance friction by 2028.
Start the EuroHPC AI Factory application process for prototyping. The existing EuroHPC AI Factories network is already operational and provides subsidized access to smaller GPU clusters for startups and SMEs. The application process takes time, and enterprises that begin the relationship with EuroHPC now will be better positioned to transition workloads to gigafactory infrastructure when it comes online.
Map your AI compute roadmap against the gigafactory timeline. The first gigafactories will be operational in 2027. Enterprises planning major AI training investments in 2026 and 2027 should model two scenarios: one where EU compute becomes available mid-project, and one where it does not. Infrastructure decisions made this year will lock in for 18 to 36 months.
Track which member states receive gigafactory awards. The seven gigafactories will be distributed across EU member states, but the geographic allocation is not yet determined. Enterprises with operations concentrated in specific countries may have better or worse proximity-based access depending on which governments win awards in early 2027.
The EU’s €30 billion compute initiative does not change the frontier AI model landscape directly. What it changes is where European enterprises can run those models at scale, under what regulatory framework, and at what price point. That is a structural shift that will play out over years, but the decisions being made in the next 18 months (which compute providers to contract with, which data architectures to build, which compliance frameworks to adopt) will determine who benefits from it.
If your team is evaluating how the EU compute landscape affects your AI infrastructure roadmap, talk to Enera about planning the transition.
Key Takeaways
- The EU Commission launched a formal call on July 30, 2026 for seven AI gigafactories, backed by €30B in combined public and private funding.
- Each gigafactory will house at least 100,000 AI chips, making them four times more powerful than current EU data centers.
- Access is designed for startups, SMEs, large enterprises, and research institutions across EU member states for training, inference, and fine-tuning.
- Bidding closes November 12, 2026; first gigafactories operational by 2028.
- EU AI Act compliance obligations make EU-resident compute increasingly important for enterprises in regulated industries.
- The program is part of a global shift toward national AI compute sovereignty, with implications for enterprise infrastructure contracts signed in the next 12 to 18 months.
- Track frontier AI governance developments alongside infrastructure shifts: the regulatory and compute layers are converging.