BofA Names Top AI Compute Stocks By Investing.com

Bank of America set price targets of $300 for NVIDIA (28x 2027E PE ex-cash), $450 for Broadcom (26x 2027E PE) and $280 for AMD (27x 2027E non‑GAAP EPS), citing AI compute demand as the primary valuation driver. Catalysts include NVIDIA leadership and Mistral’s $830M chip purchase, Broadcom’s $970M five‑year DISA contract and Tomahawk 6 shipments, and AMD’s Celestica rack‑scale collaboration, while key risks are gaming cyclicality, China shipment restrictions, regulatory/competition scrutiny and Broadcom’s ~ $60B net debt.

Analysis

AI compute demand is now a high-conviction but high-volatility cash flow stream: revenues concentrate in a small number of hyperscalers and a handful of large projects, which creates multi-quarter lumpiness and long lead-times for inventory consumption. That amplifies earnings skew — upside is rapid if a hyperscaler orders rack-scale deployments, but downside can be swift if a large program is delayed or China-related constraints bite, so realized vol will remain elevated over the next 3–12 months.

Second-order winners include systems integrators and contract manufacturers that reduce time-to-deploy for rack-scale solutions; firms that own assembly/integration (Celestica-like) will see outsized margin leverage versus pure-play fabs or GPU vendors when turnkey demand ramps. Conversely, sustained NVDA-style share dominance increases incentive for hyperscalers to internalize accelerators or co-design ASICs, a 2–5 year risk that would structurally compress vendor TAM and margin multiples.

Key catalysts to watch: hyperscaler capex cadence announcements, quarterly data-center bookings, defense procurement awards, and any incremental export-control developments out of China or the US. Tail risks that would reverse current bullishness are faster-than-expected in-house ASIC adoption by the largest cloud players, acute power/real-estate constraints slowing DC buildouts, or regulatory action that materially impairs pricing power over a 12–36 month horizon.

Trading should favor asymmetric payoffs and pair hedges — buy optionality around deployment inflection points, and use short calls or pair shorts to monetize near-term sentiment while preserving upside exposure to a multi-year AI spend cycle.

Key Decisions for Investors

• Long NVDA call spread (9–12 month LEAP vertical): buy an at-the-money-to-+25% call and sell a higher strike to finance premium; size 3–5% of equity risk. Timeframe 6–12 months; target ~2.5x premium if a material hyperscaler ordering wave occurs; max loss = premium paid. Rationale: captures asymmetric upside from concentrated AI capex while capping theta decay.
• AVGO long with protective collar (buy stock + 6–12 month 15% OTM put financed by selling a 30% OTM call): size 2–4% of portfolio. Timeframe 12–18 months; target capture of mid-teens upside if networking + software attach continues, downside limited to put strike. Rationale: preserves exposure to cash-flow/return profile while hedging design-out and cyclical semiconductor risks.
• AMD tactical long call spread (6–12 months): small-sized bullish call spread to capture rack-scale accelerator ramp tied to Celestica integration; size 1–2%. Timeframe 3–12 months; reward if early MI400 wins materialize, loss limited to premium. Rationale: asymmetric exposure to execution-driven upside without full equity downside.
• CLS (Celestica) selective long equity trade: buy on any NVDA/AMD-led pullbacks, size 0.5–1%. Timeframe 12 months; target 30–50% upside on outsized systems-integration demand; stop-loss 20% below entry. Rationale: second-order beneficiary from turnkey rack-scale deployments with lower binary risk than chip vendors.