Superintelligence Europe — No. 004

ECB Executive Board member Philip R. Lane delivered one of the most significant economic assessments of AI’s European implications yet, projecting that accelerated AI adoption could generate total factor productivity gains of 0.3 to 0.4 percentage points per year — compounding to more than four percentage points of additional productivity growth over the coming decade. The speech positioned AI as a general-purpose technology on par with previous transformational shifts, with compounding effects across industries as adoption spreads beyond early adopters.

But Lane was precise about what stands between Europe and those gains. Employee AI usage has risen to roughly 40 percent in 2025, but Europe holds just 3 percent of global AI patents against the United States’ 9 percent. Investment and deployment lag. SME access to AI tools remains uneven. Energy costs are elevated. And Europe’s bank-heavy financing ecosystem struggles to fund the intangible assets — data, models, workflows — that AI-intensive businesses depend on. Lane explicitly called for completing the EU Savings and Capital Markets Union as a structural prerequisite for closing the gap.

Nebius Group, the Amsterdam-headquartered AI cloud company that emerged from Yandex’s international restructuring, announced the closing of approximately $4.34 billion in convertible debt financing — one of the largest single capital raises in European AI infrastructure to date. The proceeds are earmarked for a $16 to $20 billion 2026 capital expenditure programme covering GPU cluster build-out, data centre expansion, and cloud platform development.

The raise follows major contract announcements with both Meta and Microsoft, positioning Nebius as a European-domiciled hyperscaler-adjacent platform filling demand that US-headquartered clouds cannot fully serve for European sovereignty reasons. The company’s Amsterdam base and post-Yandex Western governance structure have been central to its pitch to European enterprise and government customers.

Wired published a detailed infrastructure investigation on Monday documenting how the global AI race is forcing European utilities into an improvised response they were not designed to deliver. Data centre connection queues have exploded: the UK alone has more than 30 gigawatts of pending grid connection requests, against a backdrop where traditional grid build-out takes seven to fourteen years from planning to energisation. Projects are cancelling or delaying not because of a lack of capital or demand, but because they cannot get a grid connection.

European utilities are innovating in response — line upgrades, demand flexibility services, dynamic connection agreements — but the pace of AI-driven demand growth is outrunning grid modernisation in almost every major market. The report also flags a geographic concentration problem: data centres are clustering in locations with existing fibre density and cooling access, creating hotspots of grid strain rather than distributing load across the continent.

Peter Vandermeersch, a senior journalist at Mediahuis — one of Europe’s largest newspaper groups — was suspended from a fellowship role after an investigation found dozens of hallucinated quotes inserted into published articles. Vandermeersch acknowledged that he had been relying on LLM-generated summaries and had not independently verified the attributed statements before publication.

The case is the highest-profile instance yet of AI-fabricated content reaching publication at a major European media group, and it landed on Monday in every European newsroom actively debating where to draw the line on LLM-assisted journalism. The details matter: this was not a junior reporter cutting corners, but a senior figure at a major institution, using AI as a research summarisation tool in a way that collapsed the verification step entirely.

Credible accounts and publications that drove Monday’s European AI conversation. Filtered for genuine signal.