This Artificial Intelligence (AI) Stock Just Landed a Deal Worth Over $100 Billion. Is It a Buy?

AMD locked in an agreement with OpenAI that could generate more than $100 billion in revenue and has additional deals with Meta (6 GW of Instinct GPUs), Microsoft, and Oracle; AMD holds a 4% data-center GPU market share versus Nvidia's 92%. AMD reported 2025 revenue of $34.6B (+34% YoY) and diluted EPS +26%, with a net margin of 12.3% (vs Nvidia 55.6%), indicating solid growth and a lower-cost competitive position that could gradually chip away at Nvidia’s dominance.

Analysis

AMD’s move into hyperscale GPU supply is less about immediate share gain and more about shifting the marginal economics of AI compute procurement. If AMD can prove a 10–20% lower $/token or $/inference cost across large transformer deployments, that advantage scales non-linearly because hyperscalers buy in multi-year, multi-exaflop tranches; even small per-unit savings cascade into meaningful revenue and negotiating leverage with foundries and memory suppliers over 12–36 months. This is a capex arbitrage play for cloud buyers — they will mix vendors to optimize cost curves, not to crown a single winner.

Second-order winners include HBM and PCIe switch suppliers that can service a broader multi-vendor install base; increased AMD volume should exert downward pressure on spot HBM pricing by ~10–25% over 12–24 months, improving gross margins for GPU makers that are memory-cost constrained. Conversely, the used-NVIDIA market may see softened prices as hyperscalers rotate fresh inventory and offload older generations, temporarily compressing resale spreads and creating short-lived arbitrage for third-party resellers.

Key catalysts to watch are (1) large-scale public benchmark disclosures from major cloud customers showing TCO comparisons, (2) cadence of validated software stack parity (driver/runtime stability, compiler + libraries) and (3) foundry capacity signals (N3/N5 allocations). Primary risks: a single-architecture breakthrough from an incumbent that re-establishes a material performance wedge, or supply-side bottlenecks that prevent AMD from meeting synchronous multi-hyperscaler demand — both can reverse the thesis inside 6–12 months.

From a portfolio perspective this is a liquidity- and conviction-sensitive setup: the upside is concentrated if AMD converts a few hyperscalers to repeat large orders, but downside is asymmetric if customers re-centralize spend with a dominant incumbent. Treat positions as event-driven with explicit hedges and 3–18 month time horizons.

Key Decisions for Investors

• Pair trade (12–18 months): Long AMD equity (size 1–2% NAV) vs short NVDA (size 0.6–1% NAV) to express asymmetric capture of hyperscaler wins. Rationale: capture share redistribution while limiting macro/AI-factor beta. Hedge: buy 30–45 delta puts on AMD for the first 6 months to cap downside; target R/R ~2:1 if AMD wins visible multi-cloud design-ins within 12 months.
• Options trade (9–12 months): Buy AMD 12-month calls (delta ~0.35) financed by selling NVDA 3–6 month covered calls (near-term premium) — expresses conviction in secular AMD upside while monetizing NVDA gamma. Risk: if NVDA gaps higher short calls can be rolled; ideal payoff if AMD shows validated cost/TCO wins in public benchmarks.
• Event-driven buy (3–6 months): Accumulate small AM D position ahead of any major customer-infrastructure reveal (earnings/cust announcements) and set a tactical stop at -18% off entry. Take profits on a 35–60% move or upon confirmed multi-cloud deployment metrics.
• Long cloud beneficiaries (12 months): Buy or add to MSFT/META positions (size 1–2% NAV) selectively because diversified sourcing reduces marginal AI unit costs and increases gross margin tailwinds for cloud services. Expect 12–24% upside if multi-vendor sourcing materially lowers customer acquisition cost for AI services; monitor contract disclosures and capex guidance.
• Contrarian hedge (6–12 months): Buy deep OTM NVDA puts (low-cost tail hedge) sized to cover systemic shock where NVDA reasserts dominance through a product/architecture lead — protects portfolio pairs if incumbent regains exclusivity. Target cost <0.5% NAV with payoff >5x if NVDA rerates downward >30%.