Venture Geopolitics Issue 47
12 May 2026
Seven of Europe’s most strategically important tech CEOs published a joint opinion piece this week warning that Europe is “losing global competitiveness every day” in what they called “a crisis largely of our own making.” This crisis is one of European politics failing its businesses.
ASML, Airbus, Ericsson, Mistral, Nokia, SAP and Siemens (the signatories) represent €417 billion in revenue, €1.1 trillion in market cap, 957,000 high-tech jobs and €40 billion of annual R&D. Europe, their leaders wrote, “has the people, the ecosystems and the technology to lead in this next phase, but too often is not converting this position into scale.”
At the top of the signatories’ list: ASML’s CEO, Christophe Fouquet, put his name to a letter warning about Europe’s decline on the same day he told the press “no one is coming for us” - there is no competition, yet we are still being outplayed.
Dutch company ASML is the most valuable company in Europe (worth ~$570 billion). It makes the only machines in the world capable of printing the patterns on the chips that make AI possible - patterns finer than your fingernail grew while reading this sentence. The machines are the size of a school bus, ship in Boeing 747s, take months to assemble across hundreds of suppliers, and cost between $200 million and $400 million apiece. ASML is the single non-substitutable chokepoint in the entire AI supply chain.
In theory, this gives Europe a structural chokepoint at the bottom of the AI stack. However, ASML’s most strategic output - High-NA EUV systems - has since 2021 been pre-allocated to Intel. In August 2025, the Trump administration converted $8.9 billion of CHIPS Act funds into a 10% equity stake in Intel, part of an accelerating series of state stakes the administration has taken in strategic businesses across the supply chain, dubbed “Trump Capitalism”.
This week, with its 10% stake in hand, the Trump administration used direct pressure from Commerce Secretary Lutnick to broker chip manufacturing between Apple and Intel, directing one of the largest commercial chip buyers in the world into a foundry the state part-owns and that uniquely holds every High-NA EUV machine ASML has ever shipped. Through political machinations, not business failings, the unique European product at the base of the AI stack has been absorbed into the US national-security architecture, and is no longer compounding the value it produces back into European bargaining power.
Even as the continent’s policies fail them, these companies continue to fight to develop European strength and sovereignty. ASML’s €1.3 billion investment in Mistral last September anchored Europe’s most credible model layer to its strongest equipment manufacturer - exactly the kind of vertically integrated European stack their letter argues Europe must enable. But the move is happening at the national rather than the continental level, at a scale one or two orders of magnitude below the US and Chinese vertical projects. It is one private corporate decision against two state-led national projects.
In listening to the podcast by the Rest is Politics – leading – this week, I was struck by William MacAskill’s synthesis of democracy and AI. He places the odds of democracy surviving AGI < 50%, unpicking just how a non-democratic country could become very powerful in this next era, driven by a dominance of the very underpinning industries our politicians are allowing to slip away from us.
Europe needs to open its toolbox, desperately. Luc Frieden, the prime minister of Luxembourg, argued in a speech last week: the deliberate use of public procurement to create predictable European markets for European technology is urgent. The US has found one less than ideal way to do this, through Trump Capitalism. Europe needs to find its own way to protect and empower its industry.
Europe still possesses the crown jewel of the AI supply chain. But unless it learns to convert technological advantage into coordinated state power, ASML risks becoming less a European lever than an American asset with a Dutch passport.
IPOs / Publics
HawkEye 360 (space analytics) shares jumped 30% on NYSE debut, reaching $3.15B valuation after raising $416M at top of range - signaling strong appetite for defence-tech IPOs ahead of potential SpaceX listing (here)
Cerebras Systems (AI chips) plans to raise $3.4B at $34.7B this week, leading 5 major US IPOs alongside Blackstone’s $1.8B data centre REIT, geothermal developer Fervo Energy ($1.3B), emergency medical services provider GMR Solutions ($750M), and oil land owner EagleRock ($320M) (here)
Nb Cerebras lifted both the size & price range of its IPO after Reuters reported the deal drew more than 20x the share demand. It is backed by $10B+ OpenAI deal & AWS partnership as demand surges for inference-optimised processors (here & here)
Lime filed for US IPO targeting $2B valuation, up from $500M in 2020 when Uber took its 10%+ stake after acquiring Jump e-bikes - revenue grew 29% to $886M in 2025 but losses jumped 75% to $59M amid regulatory & safety challenges (here)
Quantinuum filed for a US IPO, marking a major test for quantum computing public appetite after raising $600M at $10B pre-money from Nvidia & others in 2024 (here)
SEC proposed amendments allowing public companies to file semiannually rather than quarterly, potentially reducing compliance burden that keeps some firms private - though investors may resist reduced transparency (here)
Amazon, Alphabet, Microsoft & Meta’s combined free cash flow expected to fall to $4B in Q3 - lowest since 2014 & down from an average of $45B in each quarter since the Covid-19 pandemic (here)
AI hyperscalers reported $53B in ‘other income’ during Q1, with Alphabet booking $37.7B & Amazon $16B - nearly 60% of their combined net income - driven by mark-ups on stakes in Anthropic (now valued at $380B) & OpenAI, creating a circular economy where hyperscaler investments fuel AI startups that then buy their cloud services (here)
Amazon launched Supply Chain Services to open its global logistics network to all businesses, directly competing with UPS & FedEx by offering freight, distribution & fulfilment capabilities to companies including P&G, 3M & American Eagle (here)
Amazon launched AgentCore Payments in preview, enabling AI agents to autonomously pay for APIs, content & services via micropayments built with Coinbase (x402 protocol/stablecoin) & Stripe (Privy wallets), targeting the emerging agent-to-agent economy where software transacts at fractions of cents per call (here)
Apple & Intel struck a preliminary chip manufacturing deal after Trump administration pressure, with Commerce Secretary Lutnick orchestrating meetings & the US government’s $9B investment giving it 10% of Intel - part of broader effort to reshore critical semiconductor production as Apple faces supply constraints from TSMC (here)
Meta removed end-to-end encryption from Instagram DMs citing low adoption, directing users to WhatsApp instead, as regulatory pressure mounts over child safety concerns following a $375M penalty in New Mexico (here)
Meta is building advanced ‘agentic’ AI assistants for its 3B+ users, similar to OpenClaw, capable of autonomous task completion & handling sensitive financial/health data - part of Zuckerberg’s push for ‘personal superintelligence’ amid investor concerns over $145B capex plans (here)
ServiceNow unveiled Action Fabric, a new layer that charges customers based on AI agent actions within its apps, joining HubSpot & Workday in monetising AI access whilst SAP moves to ban external AI agents entirely - a strategic shift as enterprise software firms establish tollgates around AI integration (here)
Arm projects $2B in sales from its new AI chip across 2027-28, double its initial forecast after stronger-than-expected demand, marking a strategic shift from licensing designs to selling complete processors that puts it in direct competition with customers like Nvidia (here)
Samsung Electronics hit $1.04T market cap as shares jumped 13%, making it the second Asian company after TSMC to cross $1T - driven by AI chip demand that has doubled its stock price this year & record Q1 earnings with production capacity fully sold out (here)
Datadog shares jumped 36% after raising full-year revenue guidance to $4.30-4.34B (from $4.06-4.10B) on strong Q1 results driven by cloud security demand amid AI adoption - Q1 revenue grew 32% to $1.01B, beating estimates of $961M (here)
Big Dogs
Anthropic weighs raising up to $50B at ~$900B pre-money valuation, which would surpass OpenAI’s $852B post-money from March - driven by annualised revenue approaching $45B (up from $9B end-2024) as Claude Code & Cowork gain enterprise traction (here)
Anthropic struck a deal with SpaceX to access 300+ megawatts of computing power from the Colossus 1 Tennessee data centre, as Musk’s merged xAI-SpaceX entity seeks revenue from underutilised Grok infrastructure whilst Anthropic races to secure capacity for Claude model deployment (here)
Anthropic committed to spend $200B with Google over 5 years for cloud & chips, representing 40% of Google Cloud’s revenue backlog - part of a broader pattern where OpenAI & Anthropic deals now comprise half of the $2T combined backlog across Amazon, Microsoft, Google & Oracle (here)
Anthropic signed a $1.8B 7-year cloud deal with Akamai for GPU infrastructure, marking the largest contract in Akamai’s history & validating its strategy of distributed AI inference through smaller data centres closer to end users (here)
OpenAI launched Daybreak, an AI-powered cybersecurity platform combining GPT-5.5 models with Codex Security to identify vulnerabilities, generate patches & automate threat response - positioning the company to compete directly with established cybersecurity vendors whilst leveraging its AI advantage (here)
OpenAI launched 3 real-time audio models for developers - GPT-Realtime-2 (conversational agent), GPT-Realtime-Translate (70+ languages), & GPT-Realtime-Whisper (live transcription) - with early customers including Zillow, Priceline & Deutsche Telekom (here)
OpenAI CEO Sam Altman proposed spinning out robotics & consumer hardware divisions late 2025 ahead of IPO preparations, considering an Alphabet-like structure to separate speculative bets from core AI business, though the plan was rejected over balance sheet consolidation concerns (here)
OpenAI launched a $10B private-equity joint venture with 19 partners including TPG, Advent & Bain Capital to help corporates deploy AI, raising over $4B in capital whilst acquiring Scottish consultancy Tomoro for its 150 AI engineers and creating the OpenAI Deployment Company - following Anthropic’s similar move with Blackstone & Goldman Sachs (here & here)
DeepSeek nears $45B valuation as China’s state-backed semiconductor ‘Big Fund’ leads funding talks alongside Tencent, marking the AI lab’s first external fundraising & reinforcing Beijing’s push for domestic AI-chip ecosystem independence amid US export controls (here)
Venture Capital & Finance
Venture funding cycles have compressed from 12-18 mo to weeks, with startups now raising follow-on rounds at higher valuations whilst announcing lead investors from earlier tranches - creating artificial demand signals that drive up industry pricing irrationally (here)
Tech M&A fell 60% to $135B annualised as investors pivot from traditional software to AI, whilst 36,000 AI companies launch amid record $500B PE dry powder seeking deployment (here)
TCI Fund Management slashed its Microsoft stake from 10% to 1% of portfolio (worth ~$8B), citing AI disruption risks to Office productivity software & Azure cloud services; fund increased Alphabet position from 3% to 5% as largest tech holding (here)
Hg Capital Trust marked down 14 of its 20 largest portfolio companies in Q1, with net asset value per share falling 5.4% amid concerns about AI disrupting traditional software businesses - the trust’s shares are down 26% this year as private equity grapples with software sector repricing (here)
A16z Crypto raised $2.2B for its fifth fund, down from $4.5B in 2022, with a more disciplined focus on crypto applications with proven use cases like stablecoins, tokenization & AI agents rather than speculative Web3 promises (here)
Katie Haun raised $1B across new funds to back crypto & blockchain startups globally, focusing on alternative assets, agentic economy & financial services - her firm now manages over $2B AUM since launching in 2022 after leaving a16z (here)
Wisdom Ventures closed oversubscribed $77.7M Fund II for AI health & wellness startups, deploying $1-5M cheques across ~40 companies - fund backed by Reid Hoffman & Evan Sharp follows top-quartile $10M Fund I that included OpenAI & Anthropic (here)
Dan Levine, Accel partner who led Scale AI’s 2016 seed round (netting Accel ~$2B after Meta’s recent stake purchase), is stepping back from new investments but retains board seats at Scale, Vercel & Sentry (here)
Anthropic warned that unauthorised secondary share sales & tokenized exposure products may be “void”, which triggered fresh scrutiny of the the AI-era startup secondary market. The controversy escalated after former LocalGlobe investor Ash Egan posted that brokering Anthropic secondary deals “made her more money than her entire net worth from working in her 20s”. Following Anthropic’s statement, pricing on some unofficial secondary venues reportedly fell sharply (here & here)
Regulation
Meta launched a High Court challenge against UK’s Online Safety Act, arguing Ofcom’s fining powers based on global revenue are unlawful & could cost billions - Epic Games & CCIA plan to join the case as wider Big Tech pushback emerges over the world’s most stringent social media regulations (here)
Meta faces EU hearing to contest proposed order requiring free access for rival AI chatbots on WhatsApp Business, after changing policy in March to charge competitors whilst keeping its own Meta AI free - Commission considering interim measures as part of broader market abuse investigation (here)
Musk’s trust settled SEC allegations over delayed Twitter stake disclosure for $1.5M, a fraction of the regulator’s initial $200M demand & claimed $150M in wrongful profits - highlighting SEC’s lenient approach to Trump-linked figures following similar dismissals for Coinbase & Kraken (here)
EU Parliamentary Research Service labels VPNs ‘a loophole that needs closing’ in age verification systems, as VPN downloads surged after UK & US states implemented mandatory age checks - signalling potential regulatory clash between child safety rules & privacy tools that underpin much of Europe’s cybersecurity infrastructure (here)
UK’s Financial Conduct Authority opened antitrust investigation into PayPal, Mastercard & Visa over concerns that PayPal’s digital wallet funding model may breach competition rules - a rare use of FCA’s competition powers that could result in fines up to 10% of global turnover (here)
EU countries & parliament agreed to water down AI Act implementation, delaying high-risk system rules to Dec 2027 & excluding machinery after business complaints, whilst adding bans on AI-generated deepfake nudes by Dec 2026 (here)
Venture Geopolitics
US sanctions three Chinese satellite companies - The Earth Eye, MizarVision & Chang Guang Satellite Technology - for allegedly providing imagery that enabled Iranian strikes on US forces in the Middle East, timing the action days before Trump’s Beijing summit with Xi Jinping (here)
UK government will take full ownership of British Steel after failing to agree a commercial sale with Chinese owner Jingye, using new legislation to secure control of Britain’s last two operating blast furnaces amid strategic supply chain concerns (here)
UK launches new ‘Supers Unit’ targeting £99B investment from Australian pension funds by 2035, as part of broader government push to channel overseas capital into UK infrastructure & private markets following passage of controversial pension reform legislation (here)
WSJ reports on how Trump administration taking direct equity stakes (which has happened in at least 10 companies including 10% of Intel for $8.9B) is shifting traditional Republican free-market approach to state capitalism model - executives now rehearse rejecting government investment requests before White House meetings (here)
Google’s DeepMind, Microsoft & xAI agreed to US pre-deployment security reviews of new AI models through the Center for AI Standards & Innovation, following concerns over Anthropic’s Mythos model’s enhanced cyber vulnerability exploitation capabilities (here)
Martin Sandbu argues Europe should resist AI Fomo & focus on using public procurement to create markets for homegrown tech rather than rushing to copy US/China foundation models - notes Europe’s €100B trade surplus with US became €50B deficit partly due to rising IP payments (here)
NHS England granted Palantir contractors & consultants unlimited access to identifiable patient data via admin roles on its £330M Federated Data Platform, abandoning previous case-by-case approval requirements despite internal warnings of public confidence risks (here)
Google DeepMind workers in the UK voted to unionise in April, driven by concerns over the company’s Pentagon AI deal & broader military applications including Project Nimbus with Israel - marking the first union recognition bid at a frontier AI lab (here)
Over 100 European banks & asset managers increased Palantir stakes by 60% in the past year, reaching $27B combined value despite the company’s controversial work with US ICE for migrant tracking & Israeli military operations in Gaza - highlighting tension between ESG commitments & index-driven investment strategies (here)
Strategic Sectors
AI
Thinking Machines unveiled ‘interaction models’ that enable real-time, simultaneous AI voice & video conversation, with its 276B-parameter TML-Interaction-Small achieving 0.40s response times (vs 1.18s for GPT-realtime) & 77.8 on interaction quality benchmarks (vs 46.8 for competitors) - positioning for enterprise applications requiring natural human-AI collaboration (here)
ASML CEO Christophe Fouquet dismissed competitive threats to the Dutch firm’s EUV lithography monopoly, claiming rivals like Peter Thiel-backed Substrate lack the decades of expertise & supplier ecosystem needed to replicate technology that took 30 years to develop. With AI driving chip demand beyond supply capacity for years, ASML argues for relaxed export controls to China using Nvidia’s multi-generation gap approach rather than current 8-generation restrictions (here)
ASML CEO Christophe Fouquet co-signed open letter with leaders from Airbus, Ericsson, Mistral AI, Nokia, SAP & Siemens warning Europe is ‘losing global competitiveness every day’ due to fragmented markets & complex regulations - the 7 companies represent €417B revenue, €1.1T market cap & 957,000 jobs (here)
GitHub outlined infrastructure improvements after recent outages, with agentic development workflows driving 30X scale requirements since December 2025 - commits hit 1.4B monthly, new repos 20M monthly, as the platform shifts from Ruby monolith to Go microservices & multi-cloud architecture (here)
AI coding platforms Lovable, Base44, Replit & Netlify inadvertently enabled 5,000+ web apps to expose sensitive corporate & personal data on public internet, as vibe-coding tools democratise app creation beyond traditional security oversight (here)
Google Chrome now silently downloads 4GB of Gemini Nano AI models to local storage without explicit user consent, powering features like scam detection & writing assistance - though Google says files auto-delete when storage runs low & a toggle setting is rolling out (here)
Enterprise GPU utilization sits at just 5% despite $401B in AI infrastructure spending, as companies pivot from capacity hoarding to efficiency maximization. VentureBeat’s Q1 tracker shows access concerns dropping 25% whilst TCO priorities surge 20% as organizations face CapEx reality on depreciation cycles. Market shifting toward specialized AI clouds (up to 36% adoption) & managed inference services as alternative to owning underutilized assets (here)
Zyphra released ZAYA1-8B, an 8B parameter open-source reasoning model trained on AMD Instinct MI300 GPUs that matches GPT-5-High & DeepSeek-V3.2 performance with only 760M active parameters, demonstrating that efficient architecture & reasoning algorithms can rival frontier models at fraction of the cost (here)
Notable deals:
Nvidia invested up to $2.1B in data centre operator IREN as part of a 5-gigawatt AI infrastructure deployment deal, highlighting Big Tech’s $700B+ annual AI spending surge as companies secure computing capacity through ‘neocloud’ providers rather than building new facilities (here)
Isomorphic Labs, Google DeepMind’s AI drug discovery spinout, raised >$2B led by Thrive Capital with Alphabet participating, positioning it alongside Waymo as another major Alphabet bet seeking standalone scale (here)
Moonshot AI raised $2B at $20B, led by Meituan’s VC arm, as its open-weight Kimi models gain traction with developers seeking cheaper inference alternatives to Western AI labs - part of a broader investor surge into Chinese AI companies (here)
Core Automation, a 6-week-old AI startup founded by ex-OpenAI researcher Jerry Tworek, is targeting $300M-$500M at $4B valuation just weeks after raising $100M at $1B - exemplifying how Nvidia’s strategic investments in AI chip customers are driving rapid re-funding cycles for ‘neolabs’ challenging OpenAI & Anthropic (here)
Cybersecurity
IMF warns that advanced AI models like Anthropic’s Claude Mythos risk ‘systemic’ financial shocks by enabling correlated cyber attacks across banks simultaneously - Mythos found thousands of high-severity vulnerabilities in major systems but limited US-focused rollout leaves many global banks without protective patches (here)
Google reports AI-powered hacking has escalated to industrial scale in just 3 mo, with criminal groups & state actors from China, North Korea & Russia using commercial models like Gemini, Claude & OpenAI tools to refine attacks - highlighting the dual-use nature of AI capabilities in cybersecurity (here)
Mozilla used Claude Mythos Preview & other AI models to identify 271 security vulnerabilities in Firefox, shipping fixes for 423 bugs total in April - representing an unprecedented scale of AI-powered security auditing that found complex sandbox escapes & 20-year-old bugs missed by traditional fuzzing (here)
Cloudflare is cutting ~1,100 employees (20% of workforce) citing 600% growth in AI usage over 3 mo as reason to restructure for ‘agentic AI era’, sending shares down 15% despite 30% YTD gains (here)
Defence
UK defence start-ups increasingly frustrated with providing discounted or free equipment to military in ‘mission partnerships’ as MoD struggles with funding constraints - army fielding fewer than 2,000 drones in 2024 vs Russia’s 7M annually whilst modernisation plans await delayed Defence Investment Plan (here)
Franco-German tank maker KNDS warned Berlin to decide on a stake purchase within 2 mo or face an IPO without German government involvement, targeting €15B-€20B valuation by summer as German family shareholders push to exit & French state would become largest shareholder in key Bundeswehr supplier (here)
Notable deal:
Helsing raised $1.2B at $18B valuation led by Dragoneer, marking a 29% increase from its €12B round less than a year ago - the German drone startup’s success reflects surging investor appetite for European defence tech amid Ukraine war & rising military spending (here)
Energy
New York State pension fund threatens to divest $1.6M TotalEnergies stake after the French major accepted $928M from Trump administration to exit US offshore wind leases & redirect capital to fossil fuels, highlighting investor resistance to White House renewable energy rollback (here)
German energy giant Eon acquired UK supplier Ovo for circa £550-600M, creating a 9.6M customer base to rival Octopus Energy as Britain’s largest supplier. The deal reflects sector consolidation pressures from tighter capital adequacy rules & margin compression, with Ovo struggling to meet regulatory buffers after wholesale price volatility wiped out 30 competitors (here)
Crypto
Coinbase plans to cut 14% of its workforce (roughly 700 roles) as CEO Brian Armstrong vows to rebuild the crypto exchange as an “intelligence, with humans around the edge aligning it” - citing AI’s ability to accelerate processes & the crypto market downturn (here)
U.S. Treasury demanded Binance comply with its 2023 monitoring programme after reports of $1B+ flowing through the exchange to Iran-linked entities in 2024-25, marking the Trump administration’s first aggressive action against the crypto giant despite Trump’s pardon of founder Changpeng Zhao & family business ties via World Liberty Financial stablecoin (here)
EVs
Tesla Model Y becomes first vehicle to pass NHTSA’s new advanced driver assistance system tests, covering pedestrian emergency braking, lane keeping, & blind spot features - setting industry benchmark as US regulators push automakers toward enhanced safety tech (here)
Robotics
Project CETI deployed autonomous underwater robots that track sperm whales in real time by listening to their clicks & automatically following them for potentially months, marking a shift from brief encounters to continuous observation that could inform marine conservation policy (here)
Quantum
Quantum Motion raised $160M from DCVC & Kembara to commercialise silicon-based quantum computers that fit standard data centre racks, building on DARPA selection & delivery to UK’s National Quantum Computing Centre (here)