Superintelligence Europe — No. 016

OpenAI announced Thursday that it is pausing its Stargate UK data centre project, citing the regulatory environment and the cost of energy as the reasons for halting investment. The project — announced in September 2025 in partnership with Nvidia and Nscale during President Trump’s state visit to Britain — was planned to deploy up to 8,000 GPUs at Cobalt Park in Tyneside, with the potential to scale to 31,000 GPUs over time. The facility was positioned as a cornerstone of the UK’s sovereign compute capability, designed to run OpenAI models locally for critical public services, regulated industries such as finance, and national security partnerships. The UK government had billed the project as a central pillar of its ambitions to position Britain as a global AI leader.

In a statement issued to Reuters, BBC, Bloomberg, and CNBC simultaneously, OpenAI said: “We continue to explore Stargate UK and will move forward when the right conditions such as regulation and the cost of energy enable long-term infrastructure investment.” The BBC reported that regulatory concerns include uncertainty over UK copyright rules for AI training — specifically whether the government would proceed with its proposed opt-out model for creative works, which was paused in March following significant backlash from artists and creative sector organisations including Sir Elton John. Energy costs — industrial electricity prices in the UK are among the highest in the world — are the other identified barrier.

The European Commission published its one-year progress report on the AI Continent Action Plan on Thursday, marking the anniversary of a strategy that has become the EU’s most comprehensive answer to the question of how Europe competes in AI. The headline figures are substantial: 19 AI factories are now deployed across EU world-leading supercomputers, with 13 AI Factory antennas providing regional access for researchers and start-ups. More than €1 billion in targeted funding has been deployed. The Data Union Strategy — designed to unlock data access and sharing across the continent — launched alongside the Digital Omnibus. And AI Gigafactories — facilities four times more powerful than AI factories, backed by the InvestAI facility targeting €20 billion in private mobilisation — are in development.

The Commission’s framing is confident: progress is visible across all five pillars of the plan — infrastructure, data, talent, adoption, and trustworthy AI. The Apply AI Strategy and the AI in Science Strategy have both launched. The AI Skills Academy has been funded. AI regulatory sandboxes are targeted for operational status by August 2026. The AI Act Service Desk is live. The European AI Innovation Month from 14 October to 17 November 2026 will serve as the public showcase. But the same day’s Stargate UK pause brings the Commission’s progress report into sharp focus. The EU is building sovereign public compute infrastructure at a pace that would have been unimaginable five years ago. The private investment layer — the Gigafactories, the hyperscaler partnerships, the energy-intensive training compute — is still being negotiated against conditions that the continent has not yet resolved.

US Ambassador to the EU Andrew Puzder gave AFP an interview in Brussels on Thursday that directly targeted the European Commission’s forthcoming tech-sovereignty package — expected next month, covering cloud, AI, and semiconductor sectors. His message was unambiguous: “Europe will not be able to pull itself into the AI economy by bringing other people down.” The framing was a direct critique of the Commission’s approach to European digital sovereignty, which Washington interprets as a mechanism for disadvantaging US technology companies rather than genuinely building European capacity. Puzder drew a sharp distinction between competitive behaviour and exclusionary behaviour: “Is it making Europe more competitive or making other people less competitive? One is a good plan. The other is a bad plan.”

The ambassador also made clear what Washington wants from Europe in the AI race: partnership on the hardware stack, access to US data centres, and an end to multi-billion-euro fines on US technology companies. “To partner with us, they have to have access to the data centres and the hardware stack — and you got to stop punishing the companies that are trying to bring those things to you,” Puzder said. He acknowledged that a US-EU digital dialogue had been announced the previous week and offered conciliatory language on that process. But the substance of the interview was unambiguous: Washington is treating EU AI policy as a trade and geopolitical issue, not only a regulatory one. The timing — on the same day OpenAI paused Stargate UK over regulatory conditions — will be noted in every European capital.

Euronews published an analytical piece on Thursday arguing that Europe’s AI and drone research output is genuinely world-class — and that its ability to translate that research into deployed systems, commercial products, and operational military capability is not. The analysis draws a distinction that is becoming increasingly visible in European AI and defence circles: the continent produces excellent science and falls behind on the engineering, procurement, and deployment pipeline that converts science into competitive products. In AI specifically, EU universities and research institutions rank among the global best. In model development, European frontier labs are competing at the highest level. In adoption by industry, military, and public services, Europe is substantially behind the US and increasingly behind China.

The analysis references the €115 million AGILE defence innovation fund as a structural step forward — a programme explicitly designed to close the lab-to-field gap by prioritising speed, removing consortium requirements, and targeting SMEs and startups. But AGILE is a pilot and its first call opens in early 2027. The deployment gap the piece identifies is not primarily a funding gap or a research gap. It is a procurement, regulatory, and institutional speed gap. Europe’s research institutions produce excellent AI. Its procurement systems, regulatory frameworks, and institutional risk appetite are calibrated for a slower era.