Exclusive / Defense Secretary Pete Hegseth jabs Anthropic over safety policies

The Pentagon added xAI’s Grok to its Genai.mil platform as part of an accelerated AI adoption strategy, a move that has heightened tensions with Anthropic over military use and guardrails. Anthropic, which bars its models from weapons development, resists ceding control over permissible battlefield use while the Defense Department seeks models “free from usage policy constraints,” creating procurement and policy friction that could influence defense AI vendor selection and contracting dynamics.

Analysis

Market structure: The Pentagon’s move to add xAI’s Grok and Google’s Gemini to GenAI.mil re-routes incremental DoD AI spend toward frontier-model providers that accept fewer usage constraints, benefiting cloud providers and GPU suppliers while penalizing AI vendors who impose strict safety gates. Expect 6–18 month revenue tailwinds for Google Cloud (higher contract stickiness) and for GPU vendors as procurement and pilot programs scale; smaller safety-first startups will face slower public-sector monetization and potential valuation compression.

Risk assessment: Key tail-risks include a high-profile battlefield AI failure (weeks–months) triggering procurement freezes or restrictive regulation, and a legal/regulatory backlash (6–24 months) that could force model constraints and reduce addressable market. Hidden dependencies: DoD adoption is contingent on secure on-prem/cloud integration, FIPS/NIST compliance, and congressional NDAA language; failure in those integrations could delay revenues by 6–12 months. Catalysts: RFP awards, contractor partnerships, or a GAO report in the next 30–90 days will materially accelerate share moves; an incident or bipartisan bill could reverse gains quickly.

Trade implications: Favor public tech and defense primes that capture DoD AI spending — Google (GOOGL) and NVIDIA (NVDA) for cloud+compute, and RTX/LMT for systems integration — with a 6–18 month horizon. Use size-limited options (3-month call spreads) to express asymmetric upside in NVDA and covered calls or call spreads on GOOGL to monetize time decay if near-term volatility is elevated. Reduce allocations to private AI/safety-first names that explicitly block military uses; expect private-round repricing risk over 12 months.

Contrarian angles: The consensus that “ungoverned” models win may be short-lived; if regulators or DoD impose auditability and explainability standards, safer-by-design vendors (Anthropic) could regain advantage and command a premium in procurement over 12–36 months. Market currently underprices the probability of a regulatory pause: a 10–20% drawdown in NVDA/GOOGL is plausible within 3 months after any major incident. Consider hedges sized to 1–2% of portfolio to protect multi-quarter exposure.

Key Decisions for Investors

• Establish a 2–3% long position in GOOGL within 2 weeks; add up to +1% on confirmation of a DoD contract or partnership (watch for contract announcements in next 30–90 days). Target 12–18% upside over 6–12 months; hard stop-loss at -8%.
• Allocate 1% notional to NVDA via a 3-month call spread (buy 1 20% OTM call, sell 1 40% OTM) to capture increased GPU demand; unwind if NVDA implied vol rises >40% from current or if quarterly GPU revenue guides down sequentially.
• Initiate a 1–2% equal-weight long exposure to defense primes RTX and LMT (0.5–1% each) with a 12–24 month horizon to capture systems integration and autonomy budgets; trim on two consecutive quarters of negative guidance or if DoD program awards go to competitors.
• Establish a 1% hedge: short 1% notional of MSFT (or buy put protection) as conservative pair downside to GOOGL exposure — reasoning: MSFT/OpenAI alignment policies may limit DoD monetization versus Google; cover hedge if MSFT announces major DoD-centric changes within 60 days.