FT News Briefing
2026-07-06 · Hosted by — · Financial Times
Executive Summary
The FT examines London’s King’s Cross as the epicenter of a European AI boom — home to Google, Meta, OpenAI, Anthropic, and a “DeepMind mafia” of well-funded startups — and the newly urgent debate over British “AI sovereignty.” The catalyst: the US Department of Commerce ordered Anthropic, with just 90 minutes’ notice, to block all foreign nationals from using its frontier models Fable and Mythos, exposing how dependent the UK is on American AI. Despite raising $55 billion globally, DeepMind alumni have deployed only ~$5 billion in the UK, and no British firm is building a homegrown large language model. The government is pursuing a £500 million “Sovereign AI” fund plus grants and research funding, but critics say it falls far short given the UK lacks the power, planning, and chip infrastructure to compete — problems that “could take another decade to resolve.”
Key Stories & Changes
1. The Anthropic Export Order as Sovereignty Wake-Up Call
The US Department of Commerce ordered Anthropic, with 90 minutes’ notice, to stop any foreign nationals from using its latest frontier models, Fable and Mythos.
The move “reopened the debate over whether Britain was right to let ownership of DeepMind slip through its fingers” and pushed AI sovereignty onto the G7 agenda at the meeting in the French Alps.
Investor Nathan Benaich (Air Street Capital): “The most advanced AI is built by a handful of American companies… what the rest of us are permitted to do with it can change on a Friday afternoon.”
Five Eyes intelligence chiefs warned that AI models capable of “fundamentally transforming” cyber security would arrive within months.
2. King’s Cross: London’s AI Cluster and the “DeepMind Mafia”
King’s Cross now hosts foreign outposts of Google, Meta, OpenAI, Anthropic, and Jeff Bezos’ Prometheus (“physical AI”).
Traced largely to Sir Demis Hassabis, who kept DeepMind in London after its £400mn sale to Google in 2014 — now seen as “one of the biggest bargains in dealmaking history” given OpenAI and Anthropic’s near-trillion-dollar valuations.
A “DeepMind mafia” is spawning startups: David Silver’s Ineffable Intelligence and Tim Rocktäschel’s Recursive Superintelligence.
Other arrivals: Cursor (AI coding, recently agreed a $60bn sale to SpaceX), Perplexity, open-model developer Reflection, and Google’s new “Platform 37” office opening this summer.
3. The Homegrown LLM Gap
No DeepMind London alumnus is building a homegrown LLM that UK companies and government could rely on if the US restricts access.
DeepMind alumni have raised $55bn worldwide but only ~$5bn in the UK (Dealroom).
Cosine — backed by BAE Systems, BT, Lloyds Banking Group, and the London Stock Exchange Group — is preparing to train Lumen Sovereign on Bristol’s £225mn Isambard-AI supercomputer, claiming Britain’s first fully sovereign frontier model by year-end, but has raised just $15mn and is seen by insiders as a long shot.
4. UK Fundraising Breakout and Government Strategy
UK AI fundraising broke out this year: Wayve raised $1.5bn at an $8.6bn valuation; Ineffable Intelligence raised $1.1bn (largest-ever first round for a European tech startup, pre-product); ElevenLabs, Synthesia, Isomorphic also raised multibillion-dollar rounds.
Government strategy (per AI minister Kanishka Narayan): £500mn “Sovereign AI” VC fund, £120mn in grants, chip-startup support (Nvidia alternatives), and up to £60mn for “blue sky” labs at Oxford and UCL.
Political turbulence: PM Sir Keir Starmer announced his resignation this week; some senior ministers “deeply alarmed” by the Mythos decision are pressing Number 10 to go further.
5. Infrastructure Bottleneck
SoftBank opted to invest tens of billions in AI facilities in France instead of the UK; OpenAI put its “Stargate UK” Tyneside facility on indefinite hold.
Al Carns MP: “What we don’t have is the power stations to run the data centres, the planning system to build them, or the industrial base to make the chips… We invent. Others build. Others decide.”
A tech executive: UK grid infrastructure, power availability, and cost make large clusters “tough” — problems that “could take another decade to resolve.”
Trends Identified
1. AI Nationalism Goes Mainstream
The Anthropic export order crystallized a decade-old argument (Ian Hogarth’s 2018 “AI nationalism” blog) into active policy debate at the G7 and Westminster. AI is increasingly treated as a strategic national asset, with the US itself embracing the trend — Trump has openly discussed taking government stakes in companies like OpenAI. The episode reframes access to frontier models as a geopolitical dependency, not just a commercial one.
2. Capital Follows Talent, but Value Leaks Abroad
London has re-concentrated AI talent post-pandemic (the “multiplier effect” likened to Canary Wharf and Boston’s Kendall Square), and UK fundraising has broken out to multibillion-dollar rounds. Yet the FT surfaces a persistent leak: research happens in the UK while value creation, infrastructure, and decision-making land in the US — captured in the “We invent. Others build. Others decide” critique.
3. Sovereignty vs. Pragmatism
A genuine strategic disagreement runs through the piece. Campaigners (Baroness Kidron, citing China’s DeepSeek) argue a British LLM is achievable and “bigger is better” is a myth, while advisers and the AI minister argue Britain should “bet on where the future is going” — new applications and smaller domain-specific models — rather than “replay the past” by chasing a frontier LLM. Infrastructure limits (power, grid, chips) tilt the debate toward the pragmatists. —-
Sentiment Analysis
Overall Market Sentiment: Anxious but Ambitious
The dominant mood mixes genuine alarm over US dependency with pride in London’s thriving, well-funded AI cluster — an ecosystem that is booming even as the UK debates how much control it can realistically claim.
Risk Factors Highlighted
US export-control dependency: American labs can restrict UK access to frontier models with almost no notice (the 90-minute Anthropic order).
National security exposure: Five Eyes warns transformative cyber-AI models arrive within months, heightening the cost of dependency.
Infrastructure deficit: Lack of power, grid capacity, planning, and chip-making base — issues that could take another decade to resolve.
Capital scale gap: Anthropic’s single $65bn round nearly matched all of Europe’s VC investment last year; UK capital is dwarfed by US balance sheets.
Talent absorption: Critics blame DeepMind for absorbing UK talent that could have built a British LLM challenger.
Sovereign data giveaway: NHS, Met Office, and BBC datasets are “given away” or managed by US firms like Palantir.
Investment flight: SoftBank chose France and OpenAI froze Stargate UK, signaling the UK is losing large AI-facility investment.
Political instability: PM Starmer’s resignation adds uncertainty to the continuity of AI sovereignty policy.
This episode was covered in today’s The Market Signal — 2026-07-06, a cross-source synthesis of multiple podcast reports.