Buy and Hold: 5 Artificial Intelligence (AI) Stocks to Own Through 2035

The piece highlights five AI-exposed large-caps—Nvidia, Alphabet, Microsoft, Amazon and Palantir—as multi-year winners in the buildout of AI compute and software, citing Roots Analysis' estimate that the AI market could expand from roughly $270 billion today to more than $5.2 trillion by 2035. Key specifics: Nvidia is said to command as much as ~92% GPU market share in data centers with a ~$500 billion order backlog and a strong CUDA moat; Alphabet has developed TPUs and owns ~7% of SpaceX; Microsoft owns ~27% of OpenAI and operates Azure while maintaining a 23-year dividend increase streak; Amazon holds an $8 billion stake in Anthropic and leverages AWS; Palantir’s AIP (launched mid‑2023) is driving client growth though its valuation is flagged as a risk. Managers should note the mix of hardware and software exposure, private stakes that provide indirect access to non‑public AI leaders, and the valuation caveat on Palantir.

Analysis

Market structure: The clear short-run winners are NVDA (hardware), GOOGL/GOOG and MSFT (cloud + custom silicon), and AMZN (AWS + Anthropic exposure); Palantir (PLTR) is a high-beta AI-software play. Nvidia’s ~92% GPU share and a stated ~$500bn order backlog imply pricing power and a supply-constrained market for the next 12–24 months, lifting semiconductor-equipment and power/metal demand by mid-single digits vs. pre-AI baselines. Incumbent legacy software and CPU-centric datacenter vendors (e.g., INTC) face margin pressure and potential share loss as model training centralizes on accelerator fabrics.

Risk assessment: Primary tail risks are fast-moving regulation/antitrust actions (US/EU investigations within 12–36 months), export controls or fabs delays that extend GPU scarcity beyond 24 months, and an AI-safety shock that temporarily collapses demand. Hidden dependencies include hyperscalers’ willingness to vertically integrate (Google TPU, Amazon custom silicon) which can shave 10–30 percentage points off NVDA’s addressable share over 3–5 years. Key catalysts: quarterly cloud spend beats, Nvidia order-fill acceleration, or public TPU sales announcements that can re-rate multiples within 1–3 quarters.

Trade implications: Tactical allocation: stagger 2–3% portfolio buys in NVDA over 2–8 weeks (dollar-cost to manage gamma), add 1–2% in MSFT and 1–2% in GOOGL via 12–24 month LEAPS (target +20% OTM strikes) to capture optionality to 2027. Start 0.5–1% position in PLTR, only increasing if price drops 30–40%; size a 1% AMZN 12-month call spread (buy-delta ~0.35) to play Anthropic upside. Hedge NVDA exposure with 6–9 month 25% OTM puts (~cost cap ~2–4% portfolio cash) or use collars around earnings windows.

Key Decisions for Investors

• Initiate a phased 2–3% portfolio long position in NVDA over 2–8 weeks; hedge with a 6–9 month put (25% OTM) sized to limit one-sided drawdown to ~20% of the NVDA allocation; trim on any 30% rally from entry.
• Buy 12–24 month LEAPS on MSFT and GOOGL (allocate 1–2% each, target ~20% OTM calls) to capture cloud/custom-silicon optionality while limiting capital; add to positions on confirmed cloud-spend beats or accelerated TPU commercialization within 1–3 quarters.
• Establish a speculative 0.5–1% position in PLTR (common shares) and plan to add another 1–2% only if shares drop 30–40% or after two consecutive quarters of >20% QoQ AIP customer growth; otherwise consider pair-trade long MSFT / short PLTR to express durable SaaS vs. early-adopter risk.
• Buy a 12-month AMZN call spread (cost-funded) sized at ~1% portfolio to play Anthropic upside; if AMZN announces monetized Anthropic integrations or a confirmed $Xbn incremental AWS AI revenue stream (X>2% of AWS revenue) within 6 months, convert spread to outright long.
• Reduce exposure to legacy CPU datacenter names (e.g., INTC-scale exposure) by 1–3% and reallocate to cloud/hardware winners; set a stop-loss threshold: exit reallocations if 10-year US Treasury yield rises >75bps within 90 days (growth re-rating risk).