EU Accused of Wasting €20B on AI Computing Dreams

Why It Matters

If successful, the gigafactories could secure Europe’s strategic data autonomy; if not, the €20 bn outlay may become a costly misfire that widens the AI gap with the U.S. and China.

Key Takeaways

• EU earmarks €20 bn ($21.8 bn) for AI gigafactory program.
• Plan targets four to five mega‑centers, each with 100,000 GPUs.
• Critics say Europe lacks enough AI firms to fill the compute demand.
• Heavy reliance on Nvidia GPUs could limit true sovereign compute.
• U.S. rivals pour $500 bn and $50 bn into compute, dwarfing EU spend.

Pulse Analysis

The European Commission’s €20 billion AI gigafactory program reflects a growing urgency to secure domestic high‑performance computing. By clustering 100,000 GPUs in each of four to five mega‑facilities, the EU hopes to create a "sovereign" cloud where sensitive data and proprietary models stay under European jurisdiction. The funding model blends public money with private investment, and the first call for proposals is slated for this spring. While the ambition mirrors U.S. initiatives such as OpenAI’s massive Stargate data center, Europe’s approach is constrained by a fragmented AI ecosystem and a regulatory environment that has traditionally prioritized oversight over rapid scale‑up.

Industry observers question the market demand for such massive compute capacity. Europe currently hosts only a handful of AI developers capable of training large language models, with French startup Mistral being the most prominent. The gigafactories also risk deepening dependence on Nvidia, the dominant supplier of GPU chips, which could undermine the very notion of strategic independence the Commission seeks. Policymakers like German Greens MEP Sergey Lagodinsky and think‑tank analyst Nicoleta Kyosovska warn that without a clear business case, the facilities may become "cathedrals in the desert," underutilized and financially draining.

Globally, the EU’s investment pales in comparison to the United States, where OpenAI has earmarked $500 billion and Anthropic $50 billion for compute infrastructure. China is similarly accelerating its AI hardware rollout. This disparity suggests the EU may be playing catch‑up rather than leading. To justify the €20 bn outlay, the Commission might need to broaden the gigafactories' scope beyond large‑language‑model training, targeting sectors like advanced materials, climate modeling, and industrial automation where European expertise is stronger. Aligning the project with these strategic industries could turn the gigafactories into a catalyst for broader economic competitiveness rather than a costly, underused asset.