Nvidia Stock Just Did Something for the First Time in 10 Years. Here's What History Says Happens Next.

Nvidia trades near $230, implying a forward P/E of about 19x on consensus next-12-month EPS of $12—its lowest valuation reading in roughly a decade and below its long-term average near 32x. The article argues that earnings growth has outpaced the stock's 24% year-to-date gain and estimates that a re-rating to 25x could support a roughly $300 share price, or 30% upside; a return to 32x would imply about $384, or 67% upside. Nvidia also recently announced a record $150 billion share repurchase, although rising memory costs, custom-chip competition, and uncertainty over AI spending durability remain risks.

Analysis

The relevant setup is not simply a low headline multiple: NVDA’s earnings estimates embed unusually high conversion of hyperscaler AI capex into accelerator purchases. A modest deceleration in AMZN, MSFT, GOOG, or META infrastructure budgets would therefore hit the denominator of the valuation equation and could prevent re-rating even if reported results remain strong. Conversely, continued inference demand and tight availability of leading-edge packaging would make consensus estimates conservative, creating scope for both EPS revisions and multiple expansion over the next 1-3 quarters.

AMD’s sharp relative move raises the appeal of a relative-value expression rather than outright beta. AMD needs sustained MI-series share gains and a durable software/customer adoption cycle to justify its relative premium momentum; NVDA retains the better earnings visibility, ecosystem monetization, and capital-return capacity. The second-order issue is memory: higher HBM pricing is less a pure demand signal than a test of whether NVDA can preserve system-level gross margin while suppliers such as MU and SK Hynix capture more of the AI bill of materials.

The contrarian view is that a historical P/E floor is weak evidence in a business whose customer base is now concentrated among a handful of capex-sensitive buyers. The buyback supports downside liquidity but does not resolve the key question of post-deployment returns on AI infrastructure. Over 6-18 months, custom ASIC adoption at GOOG, AMZN and META could cap NVDA’s share of incremental workloads even if total AI spending continues rising; that outcome argues for owning NVDA selectively rather than assuming a mechanical return to prior-cycle multiples.

Key Decisions for Investors

• Initiate a 3-6 month pair: long NVDA / short AMD in equal dollar amounts after confirming the next hyperscaler capex reads remain intact. Target 15-20% relative outperformance; exit if AMD reports material accelerator revenue upside or NVDA guides data-center gross margin below consensus by more than 200 bps.
• For directional exposure, accumulate NVDA only on a 8-12% pullback or after the next earnings report confirms forward revenue estimates are still rising. Use a 6-9 month call spread rather than naked calls to monetize re-rating while limiting exposure to an AI-capex multiple reset; risk is a 10-15% drawdown if hyperscaler capex guidance weakens.
• Maintain a watch alert on HBM supply and pricing, particularly MU earnings and leading memory-vendor commentary. If memory inflation is not offset by NVDA pricing or mix, reduce long exposure before the subsequent NVDA margin guide; this is the most immediate earnings-risk transmission channel.
• Do not underwrite a full historical-multiple reversion without independently verifying the stated forward EPS estimates, share count, and consensus gross-margin trajectory. A downgrade cycle in 1-3 month forward estimates, rather than the current valuation level, is the thesis falsifier.