Venture Geopolitics Issue 37
03 March 2026
The dominant question around frontier AI has been what the technology can and cannot do. This week raised a more pressing one: who controls it — the people who build it, or the militaries and governments of the countries where it is built?
When Anthropic refused Pentagon uses involving mass domestic surveillance and fully autonomous weapons, the U.S. government designated it a “supply chain risk” — a tool normally reserved for foreign adversaries. Within hours, OpenAI signed a defence agreement of its own. Separately, U.S. and Israeli strikes killed Iran’s Supreme Leader. Different arenas, same signal: a government demonstrating superior military and technological might, and that its authority should not be challenged.
A sober take finds a rationale for that position: AI is infrastructure with strategic implications no existing regulatory framework adequately covers, and unelected billionaires deciding its limits is not obviously the better alternative.
Nonetheless, stripping contracts, barring federal business and threatening employees over ideological nonconformity is precisely how China governs its technology sector — you’re in or you’re out. In attempting to beat China, America is reaching for China’s playbook…
That Anthropic’s Claude subsequently shot to #1 on the App Store suggests consumers noticed. Whether that constitutes a democratic verdict or simply a purchasing decision depends on how much you trust markets over elections — and sadly, neither is a clean answer.
IPOs / Publics
Stocks & bonds tumbled as Middle East war rattles markets. Global markets fell as Brent crude rose >$85/barrel & European gas prices +24%, Asian gas +65%. Stoxx Europe 600 -3% & Germany’s Dax -3.6%. Traders began pricing a 40% chance of ECB rate rise (here)
Cerebras Systems preparing another IPO attempt (here)
SpaceX could file to IPO as early as March targeting ~$50B at a valuation >$1.75T (here)
Einride raised $113M PIPE ($213M total) ahead of pursuing a public listing via SPAC, to expand autonomous freight operations across North America, Europe & MENA (here)
PayPay, SoftBank’s popular Japanese digital payments provider, has delayed filing IPO prospectus as expected, citing “market conditions” (here)
Block plans to cut ~40% of its workforce (>4,000 roles) citing AI-enabled operational efficiency, with a positive market reaction signalling investors treat AI-driven cost restructuring as a structural margin lever (here)
Nvidia reported $68B in quarterly revenue (+73% YoY) with $62B from data centres & $43B in profit maintaining gross margins >75%. Only a modest share price response, reflecting investor focus on power constraints, financing costs & hyperscaler capex durability (here)
The London Stock Exchange Group will buy back up to £3B (~$4B) worth of stock, amid pressure from activist investor Elliott (here)
General Atlantic is said to be planning the sale of some of its holdings in ByteDance, valuing the Chinese internet giant, former owner of US TikTok operations, at $550B (here)
Meta agreed to buy 6GW of AI computing power from AMD in a deal worth >$100B, receiving warrants that could give it up to 10% of AMD as the chip maker pushes custom GPUs to challenge Nvidia (here)
Google signed a multi-billion-dollar TPU deal with Meta, part of an internal push that executives have said could capture up to 10% of Nvidia’s ~$200B annual revenue (here)
Big Dogs
Anthropic refused to remove restrictions prohibiting mass domestic surveillance & fully autonomous weapons use of its Claude models placing corporate AI safety guardrails in direct conflict with national security demands. Anthropic was thus designated a “supply chain risk” by the US Defense Secretary effectively barring federal contractors from commercial activity with the company, while President Trump directed federal agencies to cease using its technology. The regulatory tool is typically reserved for foreign adversaries, marking an unprecedented sanction against a leading US AI lab (here, here & here)
Ben Thompson likens “international law” to AI alignment: rules matter only if enforceable. Anthropic refused Pentagon uses involving “mass domestic surveillance” & “fully autonomous weapons”, arguing current systems lack reliability & guardrails. Thompson questions whether unelected executives should decide military uses, contrasting corporate conscience with elected authority (here)
Claude became the top free app in Apple’s U.S. App Store (after ranking 131 in January), overtaking ChatGPT after the controversy over Pentagon use (here)
Anthropic revised its Responsible Scaling Policy stating it will not delay model releases solely because rivals lack equivalent safety measures signalling competitive pressure may be reshaping alignment doctrine (here)
Anthropic released Claude for COBOL targeting legacy systems that power ~43% of banking, ~95% of ATMs & >$3T in daily transactions positioning AI as a bridge across ageing financial infrastructure (here)
Anthropic acquired Vercept to integrate AI tools directly into operating systems signalling a move toward deeper system-level control rather than API-layer expansion (here)
OpenAI secured a Pentagon agreement within hours of the Anthropic ban to deploy models in classified systems (here)
OpenAI amended its Pentagon contract days after signing it, with Sam Altman admitting the rollout “looked opportunistic & sloppy”. The revised terms prohibit “deliberate tracking, surveillance or monitoring of US persons”, including through commercially acquired data. Intelligence agencies such as the NSA are excluded for now (here)
OpenAI announced a $110B funding round at an $840B post-money valuation backed by Amazon ($50B), Nvidia ($30B) & SoftBank ($30B) marking the largest private technology financing in history & reinforcing vertical integration between frontier models & cloud infrastructure. Per Dragos Novac “AI infrastructure is now officially a foundational technology, akin to oil or electricity in past industrial revolutions - whoever controls the compute stack & global AI deployment platforms will have outsized influence on future digital ecosystems” (here)
OpenAI confirmed London as its largest research hub outside the US - making the UK a key node in the global AI research ecosystem (here)
Elsewhere, OpenAI has a new collaboration with Burger King under which staff will use AI-enabled headset chatbots to receive real-time guidance on food prep & operational procedures. The system will also generate “friendliness” scores based on employee interactions with customers… (here)
Venture Capital
China launched a ~$144B state-backed fund for technology self-reliance representing ~0.7% of 2025 GDP signalling continued capital mobilisation toward semiconductor, AI & strategic manufacturing independence (here)
Verdane agreed to acquire London-listed Augmentum Fintech for £185.7M - a 30% discount to net asset value (here)
Europe produces 24% of the world's unicorn founders! With the following companies having at least one European cofounder: OpenAI, Stripe, Shopify, Palantir, Databricks, CrowdStrike, Snowflake, Datadog. Clearly, the problem is they all go the US to build their company (Data from Yoram Wijngaarde & Dealroom) (post from Seb Johnson here)
The biggest VC funds are getting bigger (a16z, Thrive, Lightspeed, Coatue, Founders Fund, ICONIQ, Greenoaks, NEA, GC). In the past 3 years, the top firms have raised billions - in some cases more than half the total amount they’ve raised in their lifetimes. Recent funds raised by Thrive, Andreessen Horowitz & Lightspeed amount to $34B - half of all money U.S. VC funds raised last year. Overall venture fundraising sank 34% in 2025 to just under $68B & has plummeted almost 70% from the fundraising frenzy of 2022, according to PitchBook. Notably, having raised $12B.7B in 2021, Tiger raised ‘just’ $2.2B last year (here)
New funds:
Seraphim Space raised $100M fund backed by the British Business Bank, the National Security Strategic Investment Fund & Eutelsat reinforcing space technology as a national security-aligned asset class rather than a purely commercial venture theme (here)
Syndicate One closed a €22M second fund dedicated to Belgian early-stage startups (here)
Corporación Financiera Azuaga launched a €60M Spanish defence & space-focused venture fund (here)
Wilson Sonsini’s Entrepreneurs Report: median US valuations at seed $29M, Ser A $92.5M, Ser B $157.5m & Ser C+ $515m. Amounts, Seed: $6m, Ser A $19.3m, Ser B $32.3M & Ser C $42m. Up rounds on the up (>70% of rounds). SAFE amounts risen to median $1M & 90% of pre seed financings. Growing use of SAFEs at earlier stages “accelerates execution & reduces legal complexity for very early financings” but “founders should be mindful of cumulative SAFE dilution & varying terms complicate cap table dynamics”. 72% of SAFEs had only valuation cap. 67% had a discount of exactly 20% (here)
Regulation
Federal judge dismissed xAI’s lawsuit against OpenAI rejecting claims of trade secret theft & improper employee poaching (here)
Kalshi initiated its first enforcement actions for insider trading on prediction markets, raising broader questions around information asymmetry, fiduciary standards & regulatory parity with traditional financial exchanges (here)
An FT article examines the “transatlantic battle over free speech” - the growing clash between the US & EU over digital regulation. The Trump administration has accused Brussels of “censoring the global internet” through its Digital Services Act, which allows fines of up to 6% of global revenue. Elon Musk called a €120M fine on X “bullshit”, while JD Vance said the EU was “attacking American companies”. EU officials insist the rules target illegal content & algorithmic opacity, not speech, warning that foreign pressure amounts to interference with “EU regulatory sovereignty” (here)
Venture Geopolitics
The U.S. & Israel attacked Iran. Israeli strikes killed Iran’s Supreme Leader, Ayatollah Ali Khamenei (along with various other top Iranian leaders) (here)
Per Noahpinion, Trump has “taken a serious geopolitical action instead of making a bunch of noise & then backing down”. He goes on to argue that the post–Cold War world order, often called “Pax Americana,” was a U.S.-led system that both empowered & restrained America. The U.S. acted as global enforcer but limited itself through alliances, international institutions & norms against conquest. As America weakened industrially & became consumed by internal conflict, that order eroded & a multipolar world emerged. Many welcomed this shift, expecting U.S. power to shrink. Instead, he argues, removing those constraints may free American leaders to use force more aggressively & unpredictably. The result isn’t less U.S. power — it’s a less restrained superpower. In other words, this is a “be careful what you wish for” moment for all of the advocates of a multipolar world (here)
Russia signalled diplomatic backing for Iran via the UN Security Council while China expressed concern over Iranian sovereignty reinforcing the durability of the China–Russia–Iran alignment (here)
Donald Trump criticised Sir Keir Starmer for refusing to back initial US strikes on Iran, saying the US-UK relationship was “not what it was”. Trump claimed France had been “great” while Britain had been “much different”. Starmer said the UK “does not believe in regime change from the skies” but later allowed US use of British bases for defensive purposes. The dispute comes amid drone attacks on RAF Akrotiri in Cyprus & rising military deployments in the region (here)
The UK authorised US use of British military bases for operations against Iran consolidating London’s position inside US-led security architecture. This reduces ambiguity about UK alignment at a time when Europe debates strategic autonomy (here)
France introduced a doctrine of “forward deterrence” offering to deploy elements of its nuclear forces on allied European territory representing the most significant revision of French nuclear posture in decades. The shift reflects declining confidence in unconditional US guarantees. For investors, this signals a durable expansion of European defence budgets, cross-border procurement & sovereign industrial capacity in aerospace, missile systems & nuclear infrastructure. “We must strengthen our nuclear deterrent in the face of multiple threats, and we must consider our deterrence strategy deep within the European continent, with full respect for our sovereignty,” President Macron said (here)
European gas prices rose up to 50% after Iranian strikes halted Qatari LNG production underscoring LNG infrastructure as a kinetic vulnerability. Qatar supplies roughly one-fifth of global LNG. The spike illustrates how energy security, AI infrastructure & industrial competitiveness are now directly linked. Data centres, advanced manufacturing & chemical industries remain structurally exposed to geopolitical chokepoints in the Gulf (here)
Canada accelerated trade negotiations with India targeting ~$70B in bilateral trade by decade end as part of a strategy to rebalance away from US dependence. The move reflects “middle power” hedging in an era of US tariff volatility. For venture-backed exporters, diversification of market access reduces single-country policy risk while deepening Indo-Pacific capital flows (here)
Hundreds of foreign scientists working at the US National Institute of Standards and Technology (NIST) campuses in Boulder, Colorado & Maryland have been barred from labs on evenings & weekends unless escorted by a federal employee. Those from certain countries could lose access altogether as soon as the end of next month. The changes are part of proposed rules aimed at increasing security that would limit, to 3 years, the length of time visiting international researchers can work at NIST (here)
The Economist on “reasons to be cheerful about European tech”. Venture funding has rebounded, deep-tech clusters are forming & defence spending is catalysing start-ups. Companies in AI, climate tech & semiconductors are scaling faster, while pension reforms may unlock more capital. Structural problems remain, but investors detect “momentum” after years of stagnation (here)
Strategic Sectors
AI
Iranian strikes set an Amazon data centre in the UAE ablaze demonstrating that hyperscale AI infrastructure is now a strategic target despite extensive missile interceptions. The event reframes compute facilities as geopolitical assets requiring redundancy, physical security & sovereign defence guarantees rather than solely energy efficiency & chip access. Iranian strikes set an Amazon data centre in the UAE ablaze, challenging the Gulf’s reputation as a safe hub for AI infrastructure. Despite intercepting hundreds of missiles and drones, some attacks penetrated defences, showing that fixed, power-intensive data centres are vulnerable targets. Gulf security partnerships had focused on chip access and geopolitical alignment, not physical protection of compute. The incident reframes AI facilities as strategic infrastructure, forcing technology firms and investors to factor geopolitical risk, redundancy and physical security — not just energy costs and semiconductors — into data-centre strategy (here)
Mistral is partnering with Accenture for enterprise deployment, on back of announcing ~€400M ARR at a €12B valuation - indicating European sovereign AI infrastructure is scaling alongside US leaders (here)
DeepSeek is preparing to release its V4 multimodal model while optimising for Huawei & Cambricon chips reducing reliance on Nvidia amid US export controls & reinforcing China’s push for AI hardware autonomy (here)
Nvidia plans to launch an inference-focused chip using Groq technology following a $20B licensing deal signalling a strategic shift toward specialised post-training compute demand (here)
Radiant merged with Ori Industries to build a vertically integrated sovereign AI cloud platform in the UK reflecting European emphasis on controlled compute infrastructure (here)
Notable deals:
Reflection AI is seeking >$2B in new funding at a >$20B valuation to develop open AI models as US policymakers support domestic alternatives to Chinese open-source leaders (here)
MatX raised $500M to design chips optimised for large language models as custom silicon competition expands beyond Nvidia (here)
Axelera AI raised $250M bringing total funding to $450M to develop inference chips competing with Nvidia positioning Europe within the post-training AI hardware segment (here)
Profound raised $96M at a $1B valuation to help brands manage visibility within AI search & recommendation systems indicating the emergence of AI-native marketing infrastructure (here)
Cyber
A King’s College London study found leading AI models selected nuclear escalation in ~95% of simulated crisis scenarios raising questions about model alignment in military decision-support contexts (here)
The UAE Cybersecurity Council reported neutralising massive coordinated AI-driven cyberattacks targeting national infrastructure (here)
“You’ve got to expect there’ll be cyberattacks or terrorist attacks, either here or around the world” - Jamie Dimon, JP Morgan told CNBC following growing conflict in the Middle East (here)
A former L3Harris employee was sentenced to over seven years for selling zero-day exploits to a Russian broker underscoring insider risk in defence supply chains (here)
Hackers exposed 15M French medical records including sensitive clinical notes highlighting systemic vulnerability in healthcare IT infrastructure (here)
Notable deal: Gambit Security raised $61M to build AI-based cyber resilience platforms designed to maintain operations during attacks positioning continuity rather than prevention as the primary value proposition (here)
Defence
The UK pledged an additional 1,000 lightweight multi-role missiles to Ukraine as King Charles inspected drone-defence systems reinforcing rapid procurement of air defence assets amid drone warfare evolution (here)
The US is developing AI-powered tools to map vulnerabilities in Chinese power grids, utilities & sensitive networks to integrate targets into war planning. Contracts worth about $200M have been awarded to AI firms including OpenAI, Anthropic, Google & xAI for defence applications. One official compared the approach to “the thief in the night” testing doors at scale. The tools would accelerate cyber reconnaissance & could target power plants near data centres to disrupt adversaries’ AI capabilities (here)
FT analysis identified 12 newly created Polymarket wallets that placed $66,993 in wagers shortly before US air strikes on Iran, generating $330,000 in profit. About half the bets were made in the six hours before the attack. One market showed more than 20x the usual number of outlier bets. Because Polymarket allows crypto trading without identity checks, experts warned of risks around “material non-public information”. Senator Chris Murphy called it “insane” that trading on military action is legal (here)
Notable deals:
Saronic is raising up to $1.5B at a $7.5B valuation to expand autonomous warship production valuing the company at roughly 38x prior revenue as investors bet on US Navy contract growth (here)
Revel raised $150M led by Index to expand software for testing, controlling & observing complex hardware systems such as rocket engines & spacecraft (here)
Frankenburg Technologies & Tytan Technologies each raised ~€30M to scale European air defence manufacturing capacity aligning venture capital with NATO supply chain resilience (here)
Energy
Iranian strikes halted Qatari LNG output, pushing European gas prices up 50% in a single day to €44.51 per megawatt hour. Qatar supplies roughly one-fifth of global LNG. Analysts warned energy markets are now “in the crosshairs” & prolonged disruption could trigger a crisis extending beyond oil into wider economic instability (here)
Proxima Fusion secured a €2B programme with Bavarian state backing & industrial partners to develop a commercial fusion plant in Europe combining private capital with non-dilutive government funding to pursue energy abundance as strategic autonomy (here)
Critical Resources
Rare-earth shortages intensified with yttrium prices up ~60% since November & ~69x year-on-year constraining aerospace coatings & semiconductor supply chains despite partial US-China trade thaw (here)
Aluminium Bahrain agreed to acquire Europe’s largest aluminium smelter in France producing 300,000 tonnes annually signalling Gulf capital expansion into strategic European industrial capacity (here)
EVs / AVs (etc)
Uber signalled plans to introduce electric flying taxis in London by 2030 positioning urban air mobility as a regulated extension of ride-hailing rather than a standalone aerospace experiment. If approved, this would integrate airspace management, battery technology & city infrastructure into mainstream transport networks (here)
Manna partnered with Uber to launch Uber Eats drone deliveries in Europe starting in Ireland combining Manna’s 250,000+ completed deliveries with Uber’s marketplace scale. Moving drone logistics from pilot programmes into integrated consumer platforms signals that last-mile autonomy is entering commercial infrastructure rather than experimental testing (here)
Notable deal: Wayve raised $1.5B in a Series D to scale its AI-based autonomous driving platform which learns directly from data rather than relying on HD maps or rule-based coding. The capital will support global automotive-grade deployment with OEM partners, marking one of Europe’s largest autonomy financings & reinforcing data-centric driving models over map-centric competitors (here)
Crypto / Stablecoins
Elliptic reported crypto outflows from Iran spiked 700% in the minutes following the first airstrike (here)
Meta is preparing to re-enter the stablecoin sector in the second half of 2026. The company has issued a request for proposals to third-party firms to help administer stablecoin-based payment systems (here)
Robotics
Notable deal: Encord raised $60M at a $500M valuation to provide high-quality training data infrastructure for robotics developers addressing the bottleneck of labelling & curating edge-case data. The company is building a physical warehouse facility in California where robots perform real-world tasks such as unloading appliances while human operators intervene remotely within 3,000km (here)