AMD and Nvidia Just Experienced an Unprecedented Valuation Split

The article argues Nvidia is materially undervalued relative to AMD despite superior AI-data-center exposure, profitability and growth, recommending investors sell AMD and buy Nvidia. Nvidia generated $89.0B of $96.2B in Q2 revenue from data centers, versus AMD's $6.7B of $11.5B; AMD's data-center operating margin was 31%, while its Client & Gaming segment generated a 15% margin. The valuation thesis is that AMD trades at more than double Nvidia's valuation despite Nvidia's substantially stronger business mix and margins.

Analysis

The valuation gap is not, by itself, a catalyst: AMD trades on an option value framework around share gains in AI accelerators, CPU attach, and a broader recovery in enterprise/client demand, whereas NVDA is priced on sustaining an unusually high AI infrastructure profit pool. The key relative question is whether AMD can convert design wins into material gross-margin accretion; incremental AI revenue at lower pricing or with costly HBM supply would support sales but not necessarily the earnings multiple. Conversely, NVDA's lower multiple is deserved only if consensus has correctly modeled a normalization in accelerator pricing, product-transition friction, and hyperscaler capex growth.

For the next 1-3 months, the relative trade will be driven less by reported revenue and more by order visibility, next-generation product availability, HBM allocation, and commentary from MSFT, META, AMZN and GOOGL on 2027 AI capex. AMD has greater downside convexity if AI GPU adoption remains concentrated in a small number of customers or if enterprise GPU deployments are delayed; its non-AI businesses can dilute mix improvement during a weak PC/gaming cycle. NVDA's principal near-term risk is a supply-chain or customer digestion signal that forces forward estimates down, which would likely compress both names but could favor AMD on a relative basis after an NVDA-specific reset.

Consensus may be overstating the importance of a single forward P/E comparison. AMD does not need to replicate NVDA's profitability to outperform; it needs evidence that accelerator revenue becomes recurring enough to re-rate its terminal margin and reduce dependence on cyclical client demand. Still, absent upward revisions to AMD AI GPU volume or gross-margin guidance, the more defensible positioning is NVDA over AMD rather than an outright AMD short, since a credible large-customer deployment announcement can move AMD sharply before financial results validate it.

Key Decisions for Investors

• Initiate a 3-month relative-value position: long NVDA / short AMD, sized beta-neutral, only if the relative multiple remains materially wider than each firm's 12-month EPS revision trend. Target 10-15% relative return; exit if AMD raises full-year gross-margin guidance or identifies incremental AI GPU demand sufficient to lift consensus EPS by more than 10%.
• Prefer NVDA outright on a 6-12 month horizon for portfolios requiring AI exposure, but stage entry around hyperscaler earnings rather than chase momentum. Thesis is falsified by broad 2027 capex cuts, a material inventory correction, or NVDA forward EPS revisions declining for two consecutive months.
• Monitor HBM suppliers SK Hynix and MU (or SOXX as a liquid proxy) for confirmation. Tightening HBM lead times and improving supplier pricing support NVDA's mix and constrain AMD's ability to close the performance/availability gap; easing conditions combined with AMD volume guidance would be a signal to reduce the pair trade.
• Do not treat the article's valuation claim as standalone short evidence for AMD. Upgrade the short leg only after verifying current forward EPS, gross-margin, and accelerator-volume consensus estimates; without those inputs, the setup is an alert rather than a high-conviction directional trade.