AMD Stock Has Blown Past Nvidia This Year. Is It Still the Better Buy?

AMD has risen nearly 190% in 2026 and briefly surpassed a $1 trillion market capitalization, driven by optimism for its Helios rack-scale AI system and 50% year-over-year revenue growth to $11.5 billion in the quarter ended June 27. The article argues Nvidia is the more attractive investment: Nvidia trades at a 25x forward P/E versus AMD's roughly 40x, while generating $192.9 billion of trailing-12-month net income versus AMD's $6.4 billion. Although AMD expects roughly 41% current-quarter revenue growth, the article highlights elevated expectations and material valuation downside risk relative to Nvidia.

Analysis

AMD’s premium is not merely a valuation issue; it embeds a rapid conversion of announced rack-scale demand into high-margin, repeatable accelerator revenue. The key near-term risk is that Helios is initially deployment-heavy and carries lower gross margin than standalone GPU sales, while customers retain leverage through Nvidia’s software ecosystem and existing cluster footprint. A modest revenue miss or gross-margin guide-down can therefore create a double hit—earnings-estimate cuts plus compression toward a more normal semiconductor multiple—over the next 1-3 quarters.

NVDA’s relative discount should not be read as low expectations. Its multiple is supported by materially higher margins, a larger installed base, and the ability to monetize networking, systems and software around accelerators. The more non-obvious beneficiary of credible AMD share gains is TSMC (TSM): incremental competition generally raises leading-edge wafer demand without requiring TSMC to select a winner. Conversely, hyperscaler capex discipline would hurt both, but AMD is more exposed to delayed qualification cycles because its incremental AI revenue pool is less entrenched.

Consensus may be too binary in treating this as an AMD-versus-NVDA winner-take-all trade. AMD can win meaningful share while still underperforming if its AI business scales at structurally lower gross margins or requires price concessions; NVDA can cede some volume and still compound earnings through platform attach and supply normalization. The next inflection points are customer-specific Helios deployments, MI-series backlog conversion, and whether AMD can sustain AI-related gross-margin expansion rather than simply report top-line growth.

The clean expression is relative, not an outright semiconductor short: AMD’s momentum and retail ownership make timing a multiple reset difficult, while NVDA remains exposed to broad AI-capex de-rating. Watch AMD’s next guide for data-center revenue mix, gross margin and named customer ramps; a revenue beat without margin or backlog validation would be a negative-quality signal. Thesis is falsified if AMD demonstrates two consecutive quarters of AI-driven margin expansion and raises forward AI revenue expectations sufficiently to close the earnings-power gap.

Key Decisions for Investors

• Initiate a 3-6 month long NVDA / short AMD dollar-neutral pair after the next AMD strength-driven rally; target 15-20% relative outperformance, with a 7-10% relative stop if AMD raises AI revenue guidance and gross-margin outlook together.
• For lower beta, buy NVDA and hedge sector exposure with a partial SMH short rather than shorting AMD outright; this isolates NVDA’s platform economics from a broad AI-capex or rates-led semiconductor multiple correction over 1-3 months.
• Maintain TSM as a second-order beneficiary of AI accelerator competition, but only add on evidence that foundry leading-edge utilization is tightening; the required confirmation is management commentary or channel data indicating sustained 3nm/2nm demand rather than inventory restocking.
• Set an AMD event alert for the next earnings release: reduce or avoid relative shorts if management discloses binding hyperscaler commitments, backlog duration, and sustained data-center gross-margin accretion. Absent those metrics, treat an earnings-driven AMD spike as an entry opportunity for the NVDA/AMD pair.