Nvidia Has 91% of Its Portfolio Invested in 2 Brilliant Artificial Intelligence (AI) Stocks

Nvidia has concentrated 91% of its equity portfolio in two AI plays—86% in CoreWeave and 5% in Arm—signaling where management sees strategic exposure to AI infrastructure. CoreWeave reported Q3 revenue up 134% to $1.4 billion, adjusted EBITDA +121% to $838 million and a revenue backlog up 271% to $55.6 billion; the stock trades at 7.7x sales with a median analyst target implying ~58% upside. Arm is gaining data-center share on its power-efficient CPU architecture, saw adjusted earnings +30% last quarter and doubled data-center royalty revenue, with analysts projecting 23% annual adjusted earnings growth through FY Mar-2027 but pricing at ~68x earnings and ~57% implied upside to the $180 median target. These fundamentals and analyst targets make both companies notable high-conviction AI bets for investors evaluating exposure to AI compute and licensing franchises.

Analysis

Market structure: Nvidia’s concentrated equity exposure (86% CRWV, 5% ARM) signals a near-term winners-take-most market for AI infra where specialized cloud operators (CRWV), energy-efficient CPU architects (ARM), and ecosystem owners (NVDA) capture pricing power. CoreWeave’s $55.6B backlog and 134% revenue growth suggest demand far outstrips available GPU-attached capacity, supporting higher contract rates and faster capex cycles for data-center buildouts over the next 6–18 months. Incumbent x86 vendors (INTC, to a lesser extent AMD) face margin pressure and share loss in AI workloads.

Risk assessment: Key tail risks are regulatory action on ARM licensing or anti-competitive scrutiny of NVDA’s stake, export controls on advanced GPUs, and customer-concentration/realization risk at CoreWeave (large deals with OpenAI/Meta could be renegotiated). Timeline: days–weeks for sentiment shocks (earnings, MLPerf releases), months for backlog conversion and GPU supply shifts, and multi-year for structural CPU share transitions. Watch GPU spot prices, contract conversion rates, and ARM data-center royalty cadence as hidden dependencies.

Trade implications: Favor concentrated exposure to high-growth AI infra: CRWV (growth at ~82% CAGR to 2027) and selective ARM exposure for secular CPU share gains, while underweighting INTC/AMD CPU legacy bets. Use options to express asymmetric upside (9–18 month call spreads) and pair trades that long infra/short legacy CPU names to hedge macro risk. Reallocate ~3–6% net from traditional hardware into AI infra over 1–3 months.

Contrarian angles: Consensus may underprice execution and realization risk—CoreWeave’s 7.7x sales looks cheap only if ~80–90% of backlog converts at forecasted pricing; if GPU availability loosens, pricing and margins could compress. ARM’s 68x earnings assumes continued royalty acceleration—watch hyperscaler insourcing (in-house designs) which would cap upside. A concentrated NVDA stake unwind or regulatory clampdown would be the classic asymmetric downside catalyst that markets underappreciate.

Key Decisions for Investors

• Establish a 2–3% long position in CoreWeave (CRWV) within 4 weeks: size 50% now, add 50% on a <=5% pullback. Target exit 12 months at $125 (consensus median), stop-loss 25% below weighted average entry to limit realization/backlog risk.
• Initiate a 1–1.5% structured ARM (ARM) bullish position via a 12-month call spread: buy 12-month $140 calls and sell $220 calls (or nearest strikes), roll if adjusted earnings continue to beat (>+10% quarterly) and royalties keep doubling. Rationale: capture 57% median upside while capping capital at risk.
• Implement a pair trade: long CRWV 2% vs short INTC 1% (or buy 6–12 month INTC put spread if preferred) to express AI infra outperformance vs legacy x86. Time horizon 3–12 months; rebalance if INTC reduces data-center share loss below 2% q/q.
• Reduce net exposure to legacy CPU/commodity semis (INTC, AMD) by 2–4% and redeploy into AI infra/semicap suppliers (NVDA, CRWV, selective tools) over the next 6–12 weeks. Re-evaluate if GPU spot prices drop >20% or CoreWeave backlog conversion falls <50% of quarterly bookings.
• Monitor three specific catalysts over the next 30–90 days before scaling: (1) GPU export-control updates from US Commerce/DoD, (2) CoreWeave quarterly backlog-to-revenue conversion rate and top-5 customer concentration, and (3) ARM royalty growth and any hyperscaler insourcing announcements—adjust positions if any catalyst deviates >20% from consensus.