The AI Computing Stock That Big Money Managers Are Quietly Buying

Nvidia has attracted continued buying from prominent investors such as Daniel Loeb (Third Point) and David Tepper (Appaloosa) over the past two quarters despite a roughly 1,100% rally since 2022. The company’s data-center revenue grew 66% year-over-year in the most recent quarter, and its CUDA developer base expanded to 5.9 million in 2024 (from 4.7 million in 2023), underpinning a durable software/hardware moat as AI demand rises. Emerging rival AI chips increase competitive risk, but Nvidia’s ecosystem and a forward P/E below 25 are cited as reasons professional managers are still adding exposure.

Analysis

Market structure: Nvidia (NVDA) remains the primary beneficiary as hyperscalers expand AI data-center capacity and GPU demand (data-center rev +66% YoY). CUDA-driven developer lock-in (5.9M devs in 2024) sustains pricing power and margins, pressuring pure-play ASIC/custom-chip winners (e.g., Broadcom/AVGO) to compete on integration rather than raw ecosystem. Tight specialized GPU supply and multi-quarter lead times signal persistent positive supply/demand imbalance for high-end accelerators, supporting skewed option vols and elevated call demand.

Risk assessment: Key tail risks are sudden policy/export controls (US/China tech restrictions), a faster-than-expected hyperscaler shift to in-house accelerators, or a macro growth shock that compresses AI capex. Immediate (days) risks: headline-driven IV spikes and momentum unwinds; short-term (weeks–months): earnings/guidance misses or supply-chain shocks; long-term (quarters–years): erosion of CUDA moat or commoditization of inference chips. Hidden dependency: NVDA’s value is as much software lock-in as silicon — open-source runtimes or standards could materially reduce switching costs.

Trade implications: Core tactical play — build a 2–4% long NVDA position now, scale to 6–8% on a 10–20% pullback; hedge with a 1–2% long put (3–6 month) if downside below $x (set threshold relative to entry). Options: for defined risk, buy 3–6 month bull-call spreads (ATM to +12–20% strikes) to capture continued data-center momentum while limiting premium. Relative trade: target pair long NVDA / short AVGO (notional 2:1) for 3–9 months to express software-moat vs. custom-chip push, rebalancing if NVDA guidance <30% data-center growth YoY.

Contrarian angles: Consensus underprices the probability hyperscalers vertically integrate, which could cap NVDA’s TAM — consider reducing exposure if two major cloud customers announce internal accelerator deployments within 12 months. The market may also underappreciate regulatory fragmentation risk; set a sell/hedge trigger if substantive antitrust investigations are opened (DOJ/FTC/EC) or if NVDA’s growth rate slips under 30% YoY for two consecutive quarters. Historical parallel: Intel’s loss of ecosystem control shows software+standards shifts can be decisive — don’t assume permanent dominance.

Key Decisions for Investors

• Initiate a 2–4% portfolio long position in NVDA today; plan to scale to 6–8% if NVDA drops 10–20% (buy-the-dip), target holding horizon 6–12 months unless fundamental/guide changes occur.
• Implement a protective hedge: buy 3–6 month NVDA puts equal to 25% of the long position or buy ATM-to-+15% bull-call spreads (ratio 1:1) to cap downside while retaining upside exposure; cut exposure if data-center revenue growth falls below 30% YoY for two consecutive quarters.
• Establish a relative-value pair trade: long NVDA / short AVGO at a 2:1 notional ratio for a 3–9 month horizon to express software-moat vs. custom-chip thesis; trim if NVDA guidance misses or AVGO announces large cloud wins.
• Overweight semiconductor/data-center hardware suppliers (e.g., memory, interconnect) by +2–4% vs. benchmark for 6–12 months to capture AI capex tailwinds; underweight legacy consumer semiconductor exposure by -3–5% due to cyclical risk.
• Set monitoring triggers (actionable): reduce NVDA position by half if (a) DOJ/FTC/EC opens formal antitrust probes within 90 days, or (b) two hyperscalers announce in-house accelerator deployments within 12 months, or (c) NVDA reports <30% YoY data-center revenue growth for two quarters.