Nvidia And AMD: Trim Your Hare, Buy A Tortoise

AMD is rated Strong Buy while Nvidia is rated Strong Sell, with the article advocating an equal-weighted AMD-NVDA portfolio to improve risk-adjusted returns. Backtesting from April 2025 to March 2026 indicates the mixed portfolio delivered higher risk-adjusted returns and lower volatility than holding either stock alone. The thesis favors AMD's exposure to underserved, cost-sensitive AI demand while highlighting valuation risk at NVDA from faster product cadence and hyperscaler CapEx pressure.

Analysis

The signal here is less about “AMD good, NVDA bad” than about a regime shift in who captures AI spend. When hyperscalers face capital discipline, the winners are not always the highest-performance GPUs; they are the vendors that can monetize budget-constrained deployments, replacement cycles, and mixed-workload inference. That favors a broader set of customers and reduces dependence on a handful of frontier buyers, which is important because concentrated demand tends to compress pricing power once capex growth slows.

The second-order effect is margin dispersion across the AI stack. If NVDA’s product cadence keeps resetting performance expectations while customers stretch refresh cycles, the market may start discounting more of that future roadmap risk upfront, especially in periods when AI spending headlines outrun actual deployment ROI. Meanwhile, AMD can benefit from a “good enough plus cheaper” procurement logic, where buyers optimize for deployment economics rather than benchmark leadership; that tends to improve order durability and can support multiple expansion if share gains look sticky.

The contrarian angle is that consensus may be underestimating how much NVDA can still defend share through ecosystem lock-in, software integration, and switching costs. The bearish case on NVDA becomes much stronger only if hyperscaler spending downticks for more than one budget cycle, not just a few months. If capex re-accelerates or enterprise inference demand broadens faster than expected, the relative spread could compress quickly, making a pure long AMD / short NVDA stance vulnerable to a sharp reversal.

For timing, this is a months-long relative-value trade rather than a days-long catalyst trade. The cleanest setup is to own the lower-volatility beneficiary and hedge the expensive, expectation-heavy leader, while recognizing that both names can rally in a strong AI tape. In that environment, the trade should be managed on relative fundamentals, not absolute price action.

Key Decisions for Investors

• Initiate a 50/50 pair: long AMD / short NVDA over the next 1-2 weeks; target 6-12 months. Expect better risk-adjusted returns if hyperscaler capex growth stays below prior peak rates, with the short leg acting as valuation and expectations hedge.
• Buy AMD on pullbacks rather than breakouts; use 3-6 month horizon. Risk/reward improves if the market continues rewarding cost-efficient inference exposure and broader customer diversification.
• Keep NVDA short only as a hedged relative-value position, not outright beta short. Tighten risk if hyperscaler spend guidance inflects higher or if software/ecosystem commentary suggests pricing power is re-accelerating.
• For options-oriented accounts, consider a call spread in AMD funded by NVDA call overwrites or a modest NVDA put spread. This captures relative upside while limiting damage if AI sentiment turns risk-on broadly.
• If NVDA underperforms AMD by another 10-15% on a rolling 3-month basis, take profits on the short leg and re-underwrite; at that point the trade may have become crowded and less about fundamentals than positioning.