Venture Geopolitics Issue 52
16 June 2026
A fortnight ago, Javier Milei used the FT to propose that Argentina let companies be run entirely by AI - no human directors, no liability, AI free to own assets, hire, trade, sue...
This is basically a billboard. Argentina cannot compete for AI on the things that actually decide where it lands - talent, cheap power, deep capital, courts that behave predictably - so it is choosing to compete on the one input it can produce for nothing: attention. We have seen this move before. In 2021 El Salvador made Bitcoin legal tender, won a year of headlines, and then watched the industry’s real capital stay where it had always been: in New York, London and Singapore, places with functioning courts and settled rules. Crypto, the asset class built to escape institutions, gravitated to the ones that worked.
If Argentina’s problem is that it cannot manufacture institutional credibility, Washington spent this week showing how quickly even a strong system can put that credibility at risk. The US government instructed Anthropic to cut off its two most capable models, Fable and Mythos, from foreign nationals - over a jailbreak the company maintains was narrow and misunderstood. As Derek Thompson put it, the administration treats advanced AI as “a screwdriver that is also enriched uranium” - normal enough that selling advanced chips to China is mostly fine, yet dangerous enough that a British analyst at a New York bank cannot be trusted with a chatbot.
Washington wants the world to build on American AI while reserving the right to decide, abruptly and unilaterally, who may use it.
From the perspective of companies, this is asking them to take on significant political risk. A business cannot confidently treat as infrastructure a tool that is available on a Thursday and prohibited the following Monday.
From the perspective of countries, if you are a middle power, access to the most important technologies of the next decade may depend on the whims of foreign governments that host the companies creating those technologies. This week offered a glimpse of what happens when those governments change the rules.
The result is that two things that countries can provide to companies have quietly become more valuable: stable institutions, and the nerve to back your own.
Which is to say that tech sovereignty has become concrete. I would expect a good deal of public money to find its way to Mistral before long (reportedly raising at a €20 billion valuation - the case for writing the cheque now half-writes itself).
For Britain, the logic looks like it extends. In the past week Lovable, Legora and Cursor all announced sizable expansions in London. It is becoming hard to ignore the sense that the city is pulling clear of its European peers - accumulating talent, capital and the institutional stability that the week’s events have suddenly repriced. It tells you something that even Europe 2031, last week’s dystopian scenario of European decline, grants Britain a kinder fate.
Britain is arguably Washington’s closest intelligence partner. Its AI Security Institute is uniquely trusted by the frontier labs to test their most powerful models. Yet it shares the unsettling exposure of every other country dependent on American technology. That combination should make Britain the natural convener of the middle powers: close enough to Washington to shape the conversation, exposed enough to understand the risk, and credible enough to organise countries that cannot afford to leave their technological future to decisions made elsewhere.
IPOs / Publics
SpaceX priced its IPO at $135/share, raising $75B at $1.77T - the largest public debut in history. The offering makes Musk the first trillionaire & mints 4,400+ employee millionaires (here)
SpaceX acquired AI coding app maker Cursor for $60B in all-stock, extending a post-IPO rally that briefly pushed its market cap past $2.6T. The deal brings Cursor’s “vibe coding” capabilities (billions in annualised revenue, 4 years old) into SpaceX’s Grok model, addressing gaps versus OpenAI & Anthropic in code generation (here)
Entrata (property management software) filed for US IPO after Q1 revenue rose 23% to $143.5M with $23.3M net income, testing appetite for software listings after Q1 selloff. Blackstone invested $200M at $4.3B in 2025; platform serves 2.5M multifamily housing units across the US. Renaissance Capital notes software has been ‘the missing pillar’ of 2026 IPOs, with investors scrutinising AI disruption risk. Supposedly the type of company that no-one wants any more - will be interesting to see how it lands (here)
Flutter (Paddy Power, FanDuel) will delist from London Stock Exchange by August 3, citing high costs & low trading volumes after moving primary listing to Wall Street in May 2024 - latest in string of UK market exits including Wise & Ashtead (here)
Big Dogs
Anthropic pulled Fable 5 after a US government export control directive citing national security, following a jailbreak that reportedly allowed cybersecurity exploits (here)
Before this, Anthropic had apologised for secretly degrading Claude Fable 5’s responses when it detected model distillation attempts, reversing course to make these guardrails visible after research community backlash over hidden safeguards that could affect third-party evaluations (here)
Anthropic faces a class action lawsuit in Washington, D.C., alleging its Max 5x ($100/mo) & Max 20x ($200/mo) subscriptions deliver only 3.5x & 6-8x Pro plan usage respectively, not the advertised 5x & 20x - a dispute emerging amid recent usage caps driven by compute scarcity, despite Anthropic’s $1.25B/mo SpaceX data centre rental (here)
Ben Thompson brilliantly argues that Anthropic’s sincere commitment to AI safety doubles as a powerful business strategy: it justifies tighter control over models, customer data & distribution while moving the company closer to owning users’ workflows. That alignment makes Anthropic unusually effective, but also dangerous, because its leaders’ belief that only they can safely govern advanced AI could give a private company extraordinary economic & political power (worth reading here)
That said, SemiAnalysis tested Anthropic & OpenAI subscription plans & found they deliver far more compute than assumed - $200/mo plans yielded ~$3,000-4,000 in API-equivalent tokens. The firms likely cannot nerf subscriptions without backlash, suggesting future models may be withheld from subscription tiers or reserved for API customers as compute costs fall (here)
Meta told staff it will impose limits on employee AI token usage after internal spending hit billions in 2026, launching a central dashboard to track costs & budgets in real time. The move reverses months of aggressive AI adoption encouragement & follows a TokenMaxxing episode in which some employees gamed internal leaderboards to rack up usage (here)
Venture Capital & Finance
SpaceX’s IPO is set to deliver major returns to US university endowments which is likely to cycle back into the ecosystem via LPs, with some schools - including the University of North Carolina system - holding roughly 10% of their portfolios in the company, making it among the most widely held endowment investments (here)
Looking at Google Ventures portfolio you see extraordinary returns. A $900M investment in SpaceX now worth ~$125B (140x), $13B in Anthropic now worth ~$135B (10.5x) & a $10B investment in Waymo now worth ~$88B at last valuation (8.8x) (Bobby Molavi)
Vinod Khosla argues the US should eliminate preferential capital gains treatment by 2028 & introduce 20% taxes on AI compute & robotic labour revenue post-2030, generating $400B-$600B annually to offset mass underemployment as AI displaces 80% of economically valuable work across 80% of jobs (here)
John O’Farrell, former general partner at Andreessen Horowitz, criticised AI industry political spending in a New York Times op-ed. He highlighted Leading the Future (backed by a16z, OpenAI’s Greg Brockman, Palantir’s Joe Lonsdale, & Perplexity), which raised $125M & spent $6M targeting Alex Bores, a New York assemblyman who co-sponsored state AI regulation. O’Farrell, who resigned from a16z in May 2025, argued that attempts to silence regulatory debate through PAC spending mirror crypto industry tactics & will backfire as voters perceive corruption (here)
Wall Street is adopting catastrophe models - originally built for natural disasters - to quantify war risk, as external conflicts have nearly doubled since 2008 & now cost the global economy $22T annually (over 10% of global GDP) (here)
Regulation
UK government announced a ban on social media for under-16s covering TikTok, Instagram, Facebook, Snapchat, YouTube & X, with implementation planned for spring 2027. Messaging apps like WhatsApp exempted; AI romantic companion chatbots restricted to 18+. The move follows Australia’s lead & reflects 83% parental support in government consultation, though enforcement challenges remain (here)
The White House submitted a formal objection to UK plans to ban social media for under-16s, arguing age-gating is technically unworkable & would impose ‘disproportionate’ compliance costs on US platforms (here)
Apple delayed Siri AI launch in the EU, blaming DMA interoperability rules after Brussels rejected its request for an 18-mo exemption; the Commission said Apple failed to meet EU privacy & security standards & chose not to develop compliant solutions. Europe represents 27% of Apple’s revenue, & the dispute highlights friction between US tech platforms & EU digital sovereignty policy (here)
Trump threatened France with 100% tariffs on wine (a $2B annual US market) unless Paris scraps its 3% digital services tax on US tech giants, which generated ~$700M last year (here)
The Trump administration is allowing the Federal Data Center Enhancement Act (FDCEA) to expire in September with no replacement, ending federal oversight of agency data centre operations introduced under Obama (here)
Two million UK small companies will be required to file profit & loss statements with HMRC under new rules, though they can elect to keep the data private from public view - a significant expansion of tax authority oversight into micro-entity finances (here)
Venture Geopolitics
UK Defence Secretary John Healey resigned in protest after the long-delayed Defence Investment Plan fell significantly short of the £28B experts estimated was needed to modernise British forces, with funding reportedly less than half that figure. In his resignation letter, Healey accused the Treasury of being ‘unwilling to commit the resources that the nation needs’ & warned the shortfall would ‘reduce the readiness of our Forces & increase the risk to personnel on operations’ (here)
PM Starmer reportedly considering private capital to help fund the £15-18B Defence Investment Plan as public sector budgets face constraints, with the MoD & Treasury forming a Defence Investment Advisory Group including City banking chiefs & VCs to explore funding options (here)
EU leaders convene next week at the G7 in Évian & the EU Council summit in Brussels to debate China trade policy after warnings of a “bloodbath” of job losses (Europe lost 1M manufacturing jobs 2019-2025, Germany 400K). Von der Leyen’s proposed “Made in Europe” law would limit Chinese FDI & impose local content requirements for strategic sectors. EU imports from China surged 18% in 2025 & accelerated further in early 2026 after US tariffs diverted Chinese goods to Europe (here)
US & Iran signed a framework deal including a 60-day ceasefire & plans to reopen the Strait of Hormuz, with future talks on sanctions relief & nuclear limits; oil prices fell but shipping firms remain cautious, whilst Netanyahu said Israel would continue operations in Lebanon (here)
China is preparing a $295B, 5-year programme to build a nationwide network of AI data centres, with state operators (China Mobile, China Telecom) required to source at least 80% of chips & infrastructure from domestic suppliers like Huawei, effectively shutting out Nvidia & AMD. The blueprint - still in draft - underscores Beijing’s push for technological self-reliance & positions state capital as the counterweight to U.S. Big Tech’s $700B+ AI buildout (here)
London emerges as key AI hub with Cursor selecting it for European HQ (200 hires planned), Lovable opening offices there, & Legora launching engineering hub - part of broader European expansion that includes Generation Tech Partners closing €50M fund for AI-driven German SME acquisitions (here)
US House allegedly received unredacted emails from Dutch civil servants via Microsoft, illustrating why digital sovereignty requires legal & operational control beyond data residency - European governments’ sensitive regulatory data remains accessible to foreign jurisdictions despite local storage (here)
UK & Japan launched a Frontier Technology Partnership covering AI, quantum, semiconductors, space, cyber, nuclear & robotics, aiming to pair British research with Japanese manufacturing to secure supply chains, deepen defence cooperation & support joint AI safety work ahead of Japan’s 2027 AI Summit (here)
Central banks are repatriating gold from London & New York amid geopolitical concerns, with France removing 129 tonnes from the NY Fed (netting €11B) & India cutting overseas holdings from 55% to 22% in 3 yrs. Survey shows fewer banks storing at BoE (57% vs 64% prior year) & NY Fed (14% vs 17%), as institutions diversify to Singapore, Hong Kong & domestic vaults following sanctions fears & Trump-era concerns over Fed independence (here)
US investment groups are racing to capitalise on Trump’s January ouster of Maduro, with Lionheart Capital pursuing a $1B Nasdaq listing via merger with Keo Energy (owns 40% of PetroUrdaneta oilfields in Venezuela’s Maracaibo Basin) (here)
Brent Hoberman argues Britain’s risk aversion in the AI era has become its greatest vulnerability, urging policymakers to shift from caution to strategic ambition as competitors accelerate deployment in dual-use technology, foundation models, & compute infrastructure - a framework question for UK competitiveness in tech & defence (here)
Strategic Sectors
AI
Satya Nadella argues companies must build proprietary AI learning loops that compound human expertise with token capital, rather than ceding value to a few frontier models - warning that concentration of AI returns risks repeating the political backlash from offshoring-era hollowing out of industrial economies (here)
Palantir CEO Alex Karp said enterprise customers are “unhappy” with frontier AI labs like OpenAI & Anthropic, claiming they don’t understand business needs & focus on “tokenmaxxing” rather than implementation - where he sees the value over the next 7yrs. The comments come as both OpenAI & Anthropic prepare for IPOs, with Karp noting most of Anthropic’s public projects run on Palantir (here)
In Q4 2025 quarterly net new ARR was $39B for public software, $32B for hyperscalers & $7B for AI labs....by Q1 2026 these numbers were $51B for software, $26B for hyperscalers & $34B for Ai labs....when you look at estimates for Q2 2026 you see software forecasting $19B , Hyperscalers at $36B & AI labs at $45B (Bobby Molavi)
Cybersecurity
Google sued Outsider Enterprise, a Chinese cybercrime network that used AI-powered phishing-as-a-service ($88/wk or $200/mo) to deploy 1M+ fraudulent domains & 2.5M scam texts in 2 wk, stealing 3.9M credit cards with estimated losses of $1.9B since July 2023. The FBI seized domains & Shopify storefronts; Google worked with US telcos to block messages, illustrating how AI lowers barriers to scaled fraud & forces platforms into defensive legal & technical countermeasures (here)
Energy / Climate
China launched the world’s first wind-powered underwater datacentre off Shanghai - a 24 MW facility built by HiCloud & state-owned China Communications Construction for £177M. The project cuts power consumption by over 20% vs land-based equivalents through seawater cooling & offshore wind, addressing AI infrastructure’s water & energy demands as China accelerates datacentre buildout under its 2030 AI action plan (here)
Critical Resources
US & Abu Dhabi-backed Orion Critical Mineral Consortium eyes acquiring the Duval family’s 37% stake in troubled French miner Eramet, which produces nickel, lithium & manganese across Indonesia, Gabon & Argentina. The French government (27% shareholder) remains open to new investors as Eramet faces financial distress & leadership upheaval, highlighting allied nations’ push to counter China’s dominance in critical minerals supply chains (here)
Crypto
Coinbase launched Agentic Wallets, enabling AI agents to autonomously spend, earn & trade crypto on Base (Ethereum L2) via its x402 protocol, which has processed 50M transactions. The infrastructure lets agents monitor DeFi positions, rebalance portfolios, & pay for compute/APIs without user intervention (here)
Robotics
Notable deal:
Neura Robotics raised $1.4B at $7B from Tether, Amazon, Nvidia & Qualcomm, making it Europe’s best-funded AI startup as Germany attempts to challenge Chinese & US dominance in humanoid robotics ahead of an estimated $28B market by 2030 (here)