This week's developments — from Pentagon AI partnerships to infrastructure attacks — unfold against this backdrop of record capital inflows and mounting market uncertainty.
Watch this week's AI landscape highlights in 60 seconds:
🌍 Landscape
Pentagon AI Battle
OpenAI secured a Pentagon contract with weaker safeguards than Anthropic's rejected terms — relying on existing laws rather than specific prohibitions on autonomous weapons and mass surveillance [NYT]. US Central Command used Claude for intelligence and targeting during Iran strikes despite Trump's ban hours earlier, proving deep military integration [The Guardian]. Pentagon's threatened "supply-chain risk" designation could ban Anthropic from all defense contractors, imperiling 80% of revenue (enterprise sales). Major investors including Amazon, Nvidia, Lightspeed, and Iconiq are now pushing Anthropic and Pentagon to de-escalate. Anthropic's revenue run rate hit $19B, up from $14B weeks ago, as Claude topped App Store downloads [Reuters].
Defense Contractors Preemptively Dump Anthropic Despite Legal Questions
Major defense contractors including Lockheed Martin are expected to purge Anthropic's AI tools from supply chains following Pentagon's order — even though Trump administration's ban may fail in court, government attorneys say. Contractors prioritize access to Pentagon's $1T annual budget over legal uncertainty, signaling Anthropic's isolation accelerating [Reuters].
AI Models Threatened Nuclear Strikes in 95% of War Simulations
King's College London study placed GPT-5.2, Claude Sonnet 4, and Gemini 3 Flash in 21 nuclear crisis simulations. All featured nuclear signaling; 95% saw tactical nuclear use, 76% reached strategic threats. No model ever chose accommodation or surrender. Critical finding: explicit deadlines triggered sharp escalation even in otherwise-restrained models, revealing unpredictable behavior under temporal pressure [King’s College London News].
Nuclear escalation by threshold and model. All models engaged in nuclear signaling, but willingness to actually use nuclear weapons diverged dramatically. Source: King's College London study
February VC Funding Hits Record $189B as Three AI Companies Capture 83% Despite Public Market Selloff
February 2026 shattered venture capital records with $189B raised globally — a 780% increase over February 2025. Three AI companies captured 83% of total funding: OpenAI ($110B), Anthropic ($30B), and Waymo ($16B). This private market euphoria contrasts sharply with public markets, where software stocks shed $1T on AI displacement fears [Crunchbase]. The divide may soon narrow: both OpenAI and Anthropic are exploring IPOs this year, forcing first public scrutiny of sky-high valuations [YahooFinance].
📈 Tailwinds of the Week
Dell AI stock jumped 30% after Q4 server revenue hit $9.5B (342% growth). Company ended with $43B backlog; forecasts $50B AI revenue in fiscal 2027. [Dell earnings report].
AXT semiconductor stock surged 22% on March 2 after analysts raised price targets, citing accelerating demand for indium phosphide wafers critical for AI infrastructure [Trefis].
Workday stock rose 6.7% after Q4 revenue beat ($2.53B vs $2.32B estimate), driven by AI-enhanced products. Company forecast 12-13% growth [Stocks-to-Trade].
📉 Headwinds of the Week
Block's 4,000 layoffs (40% workforce) blamed on AI, but Bloomberg reports critics call it “AI-washing” — cost-cutting disguised as automation-driven efficiency [Bloomberg].
Citrini research report predicting 10.2% unemployment from AI disruption triggered 2% market selloff, extending S&P decline to 38% from October 2026 peaks.
Drone strikes hit AWS UAE data centers during US-Iran conflict, causing structural damage and multi-day outages across 60 services [AWS].
Amazon, Microsoft, Google and Oracle build out regional data center and Al infrastructure; Note: Location points are approximations: Source: Company statements | Akash Sriram (Reuters)
Investment Opportunities
Fig Security emerged from stealth March 3 with $38M (seed + Series A) co-led by Team8 and Ten Eleven Ventures. Founded by ex-Google security engineers, Fig traces data flows through security stacks, alerting teams when changes threaten detection systems [Tech Startups].
Chariot Defense secured $34M seed funding (announced March 3) for AI-powered cyber threat protection in air defense systems, led by Andreessen Horowitz [PRNewswire].
Runway raised $315M, World Labs secured $1B (Nvidia, AMD investors), and Luma AI closed $900M for AI world models — seen as critical path to AGI beyond language models [Pitchbook].
+ News on AI
Anduril Hits $31B Valuation on Pentagon’s Push for AI-Powered Autonomous Arsenal
Palmer Luckey, 33-year-old Anduril founder, has become the Pentagon's favored defense tech innovator under Trump's military modernization push. His vision: AI-powered autonomous weapons (drones, jets, submarines) that reduce costs and save lives. Anduril secured $6B in contracts, $2B revenue, and $31B valuation despite limited production scale. Defense Secretary Hegseth champions Luckey's approach, though officials question whether the startup can match established contractors like Lockheed Martin. Luckey reportedly influenced Trump's Pentagon-to-War Department rename. His 'arsenal of freedom' pitch aligns with Trump's $1T+ military modernization budget [NYT].
London Anti-AI Protest Draws Hundreds but Lacks Clear Investment Impact
Pause AI and Pull the Plug organized London's largest anti-AI march yet, drawing hundreds to protest outside OpenAI, Meta, and DeepMind offices. Demands ranged from extinction risk concerns to AI-generated content complaints. Organizers aim to dry up AI talent pipelines and push regulation, but acknowledge companies are "optimized to not care." Protesters hope public pressure drives government action [MIT Tech Review].
Source: MIT Tech Review
Supreme Court Declines AI Copyright Case, Leaving Human Authorship Requirement Intact
The US Supreme Court rejected Stephen Thaler's appeal seeking copyright protection for AI-generated artwork, upholding Copyright Office rulings that works require human creators. Lower courts confirmed "human authorship is a bedrock requirement." The decision leaves unresolved questions about AI-assisted works as artists using tools like Midjourney also face copyright denials. Legal uncertainty persists during critical AI commercialization period [Reuters].
What to Watch
👀 Can Anthropic survive Pentagon's supply chain blacklist threat while OpenAI phases in as defense AI provider? Legal battle and contractor fallout could reshape competitive dynamics within 6 months.
👀 Citrini's viral AI displacement thesis faces institutional pushback, but Block's 40% workforce cut validates concerns. Track white-collar employment data for thesis confirmation.
👀 Six-month Claude phase-out from classified ops tests Pentagon's ability to swap AI vendors mid-conflict. OpenAI and xAI integration timeline amid active Iran operations signals execution risk.
🎓AI-Learning: Summits, Webinars & Events
USA Artificial Intelligence Summit 2026
📅 June 2026 | 🌐 Washington, D.C.
The USA AI Summit 2026 will convene policymakers, industry leaders, researchers, and innovators for discussions on AI's economic, governance, and global impact. Key topics include federal vs. state AI governance frameworks and preemption, infrastructure requirements , U.S. positioning in the global AI race, government procurement as market-shaping force, R&D investment strategies, open-source models, and workforce development. More info
Valence Thoughts
“Understanding how frontier models do and do not imitate human strategic logic is essential preparation for a world in which AI increasingly shapes strategic outcomes.”
— Professor Kenneth Payne, King’s College London, Department of Defence Studies
That’s it for this week.
Until next time,
The CLNM Capital
[Edited with the assistance of LLMs]
Want to share this newsletter with a friend? Use this link: https://clnm.beehiiv.com/