Superintelligence Europe — No. 011

Nebius Group, the Amsterdam-listed AI cloud company, announced Tuesday that it is constructing a 310-megawatt AI factory in Lappeenranta, in Finland’s South Karelia region near the Russian border. The facility — Nebius’ tenth globally — is being built by Finnish developer Polarnode and is valued at more than $10 billion. First capacity is expected to be available to customers in 2027, making it one of the largest dedicated AI factories in Europe when fully deployed. Shares of Nebius rose 4.5 percent in premarket trading on the announcement.

The Lappeenranta site will surpass Nebius’ existing 240MW project near Lille, France, as the company’s largest European facility. Its current operational Finnish site — a 75MW data centre in Mäntsälä — became the first European deployment of the Nvidia GB300 NVL72 platform earlier this year. Nebius is targeting more than 3 gigawatts of contracted capacity by end of 2026, including recently secured contracts worth more than $40 billion with Microsoft and Meta. The Lappeenranta factory will account for roughly 10 percent of that total. The facility will span approximately 100 acres, use a closed-loop liquid cooling system eliminating the need for water intake from local supplies, and integrate a heat recovery system to supply excess server heat to the local district heating network — a model already delivering approximately 4,000 tonnes of avoided CO₂ per year at Mäntsälä.

Google launched its most ambitious UK AI skills initiative to date on Tuesday. “AI Works for Britain” is backed by nearly £2 million in Google.org grant funding and targets what the company’s own research calls a “stuck majority”: while 65 percent of the UK population has tried AI tools, only one in ten considers themselves an advanced user, and 76 percent say they feel stuck in their progression. Three-quarters of 25-to-34-year-olds said having an AI assistant in their pocket gives them the professional confidence to apply for roles that previously felt out of reach. The gap between AI access and meaningful AI use is the problem Google is now directly funding to close.

The programme has three delivery mechanisms. First: “Squeeze the Juice” pop-up bars launching in Leeds, Liverpool, Birmingham, and UK AI Growth Zones — walk-in hubs where members of the public receive hands-on demonstrations of practical AI use: turning handwritten notes into structured professional quotes, automating admin tasks, and using agentic tools for research. Second: a Gemini University Tour taking AI training directly to campuses nationwide, targeting students before they enter the workforce. Third: AI Skilling for Students, providing free Grow with Google digital skills content and AI-specific certification pathways. The initiative is framed not as corporate philanthropy but as an economic mobility intervention — Google explicitly connecting AI skill development to career progression and income outcomes for workers who feel the system has not worked for them.

Singapore-headquartered Bitdeer Technologies Group announced on Monday — with full Tuesday media coverage confirming the story across Scandinavia — that its subsidiary Tydal Data Center AS has engaged Norwegian contractor Data Center Installations AS to convert its Tydal bitcoin mining facility into a 180-megawatt AI data centre. The project targets completion by December 2026. It is designed primarily for co-location of Nvidia’s Vera Rubin AI technology, following Nvidia’s reference design. Once operational, the Tydal facility is expected to be Norway’s largest operational AI data centre and one of the largest in Europe by installed capacity.

The Tydal site sits in Trøndelag county, powered by renewable hydroelectric energy — making it structurally well-suited for AI workloads that face scrutiny over power consumption. The project’s heat recovery system is also designed to route excess server heat to adjacent food production operations on the same industrial site. DCI, a Norwegian contractor and subsidiary of Sweden’s Sparc Group, will manage design, construction, commissioning, and maintenance.

Governance analysts confirmed on Tuesday that the AI Act’s Digital Omnibus trilogue is targeting a final agreed text by April 28, 2026 — the date of the second scheduled trilogue session. The Parliament voted its position on March 26 (569–45). The Council adopted its position on March 13. Interinstitutional negotiations formally began the same day as the Parliament’s vote, with a political “handshake” trilogue followed immediately by technical working group sessions. The Cypriot Council Presidency — which has made the Omnibus a priority — is pressing for an agreed text before August 2, 2026, the date when the original AI Act’s high-risk obligations would otherwise come into force without the deadline extensions the Omnibus provides.

The broad alignment between Parliament and Council on the major issues makes the April 28 timeline credible. Both institutions have agreed on fixed application dates: December 2, 2027 for Annex III high-risk systems (employment, biometrics, law enforcement); August 2, 2028 for product-embedded AI. Both have agreed on the nudifier ban. Both have agreed on reinstatement of the EU database registration requirement. The primary open divergence is the watermarking transitional period — Parliament wants three months, the Commission proposed six — which is resolvable. The EU AI Act service desk is due to launch in July 2025 to support compliance regardless of when the Omnibus is finalised.

Credible accounts and publications that shaped Tuesday’s European AI conversation. Filtered for genuine signal.