AMD Is Up 187% This Year and Nvidia Is Up 22%. Prediction: Nvidia Wins the Next 12 Months.

AMD surpassed a $1 trillion market capitalization after a 187% gain in 2026, but trades at roughly 40x 2027 consensus earnings despite estimates already implying that earnings will approximately double. Nvidia reported fiscal Q2 revenue growth of 106% to $96.2 billion and guided for $108.0 billion in fiscal Q3 revenue, while management expects about 70% revenue growth in fiscal 2028, constrained by supply rather than demand. The article favors Nvidia over AMD for the next 12 months, citing Nvidia's lower valuation of about 15x fiscal 2028 earnings and greater room to absorb execution disappointment.

Analysis

The relevant setup is not simply relative valuation: NVDA’s constrained supply creates a pricing-and-mix option that consensus EPS may still understate, while AMD must convert announced design wins into high-margin accelerator revenue rather than lower-value EPYC CPU sales. AMD’s enterprise AI deployments also face a software and systems-integration hurdle; a larger installed base of CUDA-trained developers raises switching costs even when AMD hardware offers attractive price/performance. TSMC (TSM), SK Hynix exposure via HBM, and NVDA networking suppliers should retain stronger visibility than merchant compute beneficiaries if AI capex remains capacity-bound.

Over the next 1-3 months, AMD is more vulnerable to positioning-driven downside than a deterioration in underlying demand: a modest delay in rack-scale qualification, HBM availability, or gross-margin expansion can force estimates lower because earnings expectations embed an unusually narrow execution band. Conversely, NVDA’s principal near-term risk is not demand but a supply-chain release that shifts market attention from scarcity pricing to customer digestion and return-on-investment scrutiny. A broad AI-capex reset would likely compress both multiples, but AMD should exhibit greater downside beta given its dependence on upward revisions.

The contrarian risk to the obvious NVDA-over-AMD view is that hyperscalers may deliberately dual-source accelerators to improve negotiating leverage, making AMD’s incremental revenue more economically valuable than its current share suggests. That thesis requires evidence of sustained accelerator mix gains: AMD data-center gross margin, backlog conversion, and repeat orders—not merely aggregate segment growth. If those indicators improve for two consecutive quarters, the relative short should be reduced even if AMD remains optically expensive.

Key Decisions for Investors

• Initiate a 6-12 month market-neutral pair: long NVDA / short AMD, sized beta-neutral rather than dollar-neutral. Target 20-30% relative outperformance; stop if AMD outperforms NVDA by 15% from entry or if AMD reports two quarters of data-center gross-margin expansion with raised full-year accelerator guidance.
• For directional AI exposure, favor NVDA over AMD into the next earnings cycle, but scale entries rather than chase momentum. The thesis is falsified by a material reduction in forward revenue visibility, evidence of hyperscaler order deferrals, or a gross-margin guide-down attributable to pricing rather than product transition.
• Use TSM as a lower-beta supply-chain expression of sustained accelerator demand over 6-18 months, subject to monitoring advanced-packaging capacity commentary. This is preferable to adding high-multiple GPU exposure if the objective is participation in volume growth with less single-product execution risk.
• Do not treat AMD’s aggregate data-center growth as a clean AI-GPU read-through. Set an alert for disclosed accelerator revenue/mix, MI-series availability, and customer production deployments; absent those data, avoid adding to AMD after sharp momentum moves.