Should You Buy Advanced Micro Devices (AMD) Stock After Its 23% Drop?

AMD reported fiscal Q3 2025 revenue of $9.2 billion, up 36% year-over-year, with data center revenue of $4.3 billion and client (Ryzen AI) revenue of $2.7 billion (+46% YoY). The company is pushing into AI data-center GPUs with CDNA4-based MI350 series (notable Oracle order for 131,000 MI355X units) and touts a 2026 MI400/Helios rack expected to be ~10x the MI350 performance; AMD also has an OpenAI commitment for up to 6 GW of GPU capacity by 2030. Trailing adjusted EPS was $3.73 (P/E 54.6) while Wall Street consensus projects $6.42 in 2026 (forward P/E ~31.7); the piece contrasts AMD with Nvidia’s larger data-center footprint and faster growth, and notes AMD equity is down ~23% from its recent high. These developments suggest material long-term upside if AMD scales data-center revenue toward management’s $100B-plus thesis, but near-term valuation and competitive dynamics warrant a measured investment horizon.

Analysis

Market structure: AMD is a direct beneficiary (AMD, ORCL, META, DELL, HPQ) as hyperscalers and OEMs buy MI350/RYZEN AI for lower TCO real-time inference and local PC AI; Nvidia retains pricing power in high-performance training but faces incremental share loss in inference cost-sensitive racks. Expect tighter GPU supply through 2025–2026 as demand from OpenAI/Oracle scales—this supports vendor pricing for near-term shipments but creates incentive for hyperscalers to negotiate volume discounts and custom silicon.

Risk assessment: Key tail risks are execution (MI400/Helios delivery slip beyond 2026), customer-concentration (OpenAI/Oracle caveats), and regulatory/antitrust scrutiny around hyperscaler supply deals; any one could compress multiples by 20–40%. Time horizons: days—IV and sentiment swings; weeks/months—quarterly data center revenues and Oracle/OpenAI deployment updates; years—MI400 and 2030 6GW realization drive structural upside or downside.

Trade implications: Favor long-biased exposure to AMD with defined sizing and option overlays rather than unhedged speculative longs; use 12–24 month LEAPS to capture MI400 optionality while selling short-dated calls to fund cost. Consider pair trades (long AMD / short NVDA) to isolate execution risk vs. ecosystem moat; overweight hardware OEMs (DELL, HPQ) for Ryzen AI pickup over 6–12 months.

Contrarian angles: Consensus underestimates software/stack risk—TCO claims only monetize if drivers, libraries and partner deployments scale; conversely the market may be under-reacting to a successful MI400 (10x claim) which would re-rate AMD materially. Unintended consequences include price competition and margin erosion across inference GPUs, and hyperscaler migration to proprietary accelerators that bypass commodity suppliers.

Key Decisions for Investors

• Establish a 2–3% portfolio long position in AMD (AMD) with a 12–24 month horizon; add incremental size if quarterly data‑center revenue > $4.5B or if OpenAI confirms multi‑MW deployments in 2026; set a tactical stop‑loss at -25% from entry and target 40–60% upside on successful MI400 cadence.
• Buy 18‑month AMD LEAPS (Dec 2026/Jan 2027 expirations) sized 0.5–1.0% notional to capture MI400 optionality; hedge ~30–40% of premium by selling 1–3 month OTM calls when implied volatility >30% to reduce basis.
• Implement a relative‑value pair: go 1.5% long AMD and 1.0% short NVDA to express potential narrowing of the valuation gap; immediately unwind the short if NVDA data‑center revenue growth >50% YoY or unwind the long if AMD publicly delays MI400 past Q4 2026.
• Overweight hardware OEMs (DELL, HPQ) by 1–2% for 6–12 months to play Ryzen AI adoption; concurrently reduce 2% exposure to high‑multiple cloud/software names where capex-driven margin pressure is likely over next 4 quarters.