1 Artificial Intelligence (AI) Stock I'd Never Sell

AMD forecasts 35% revenue CAGR over the next three years and 60% CAGR for its data-center segment, putting it close to Nvidia's ~65% data-center growth and signaling data center will become the majority of revenue from roughly half today. CEO Lisa Su's turnaround, the MI450 AI accelerator roadmap, and a deal with Meta to power next-generation AI infrastructure are cited as key catalysts. Valuation is currently a trailing P/E of ~77 but expected to fall to a forward P/E of ~31 versus the S&P 500 ~29, suggesting upside if profit growth materializes. The author takes a 'never sell' stance, framing AMD as a rapidly strengthening long-term competitor to Nvidia in AI chips.

Analysis

AMD’s recent trajectory forces a multi-front battle that is as much about supply chain control and software entrenchment as it is about raw silicon performance. The next 6–18 months will be dominated by foundry and HBM allocation — whoever secures priority access to advanced nodes and memory stacks can convert technical wins into sustained share gains without materially changing roadmaps.

Software and ecosystem friction remain the single biggest persistent moat for incumbents; a customer won on silicon can still revert if the software stack, tooling, or benchmarking costs spike. Expect migration decisions to be staged — initial pilots over 3–6 months, broader infra commitments over 12–24 months — which creates multiple discrete catalysts and short windows for re-pricing.

Second-order winners are in packaging, memory, and cloud services procurement: advanced multi-chip modules and HBM capacity will capture economic surplus if demand curves steepen, while hyperscalers’ bespoke infra programs can compress vendor ASPs if they internalize design wins. The investment case should therefore be treated as a sequenced optionality bet — buy execution exposure (AMD) while hedging ecosystem and concentration risks (software lock-in, foundry bottlenecks, hyperscaler negotiation leverage).

Key Decisions for Investors

• Long AMD equity (1–2% NAV) with a 12-month horizon; pair with a 6–9 month AMD ATM put (protect ~20% downside) to limit tail risk. Rationale: capture upside from share gains while capping downside from execution or supply shocks; take profits into staged tranches at +30% and +60%.
• Pair trade: Long AMD / Short INTC equal notional for 12 months (target relative return >2:1). Thesis: asymmetric share migration in servers and accelerators; stop the pair if AMD underperforms INTC by >15% over a rolling 30-day window to cut idiosyncratic execution risk.
• Options spread to express convexity: Buy AMD 12‑month call debit spread (buy longer-dated OTM call, sell higher strike longer-dated call) sized to 0.5–1% NAV. Finance by selling 3–4 month AMD call premium to improve carry; this concentrates payoff into successful multi-quarter execution while limiting premium loss if momentum stalls.
• Supply-chain longs (TSM, MU, ASE) sized 0.5–1% NAV each with 6–18 month horizons — these act as levered plays on aggregate accelerator demand. Trim into any >25% rally; these names are exposed to cyclical node and HBM tightness and will re-rate quickly if bottlenecks persist.