The S&P 500 Is Near an All-Time High but Trading at Its Lowest Valuation in a Year. Nvidia, Alphabet, and Amazon Help Explain Why.

The S&P 500 trades at a 19.4x forward P/E, its lowest level since April 2025, while Nvidia, Amazon, and Alphabet trade at 24.4x, 20.1x, and 17.2x forward earnings, respectively, as earnings estimates have risen faster than share prices. Nvidia is guiding for fiscal 2028 revenue growth of 70% year over year after Vera Rubin shipments began in August, despite the stock rising 21.9% year to date. Amazon and Alphabet have turned free-cash-flow negative as they accelerate AI infrastructure investment, creating upside to margins if AI data-center spending produces significant revenue but also exposing chip suppliers to any hyperscaler capex retrenchment.

Analysis

The relevant signal is not headline forward P/E, but the durability and cash conversion of the earnings embedded in it. NVDA’s multiple can remain stable only if rack-scale system growth offsets rising memory and networking content costs; a memory upcycle transfers incremental economics toward HBM suppliers (MU, SK Hynix) and away from GPU-system gross margin. The more material near-term read-through is whether hyperscaler capex remains funded internally rather than through incremental leverage or reduced buybacks, as that determines the duration of semiconductor orders.

AMZN and GOOGL face a more asymmetric setup than NVDA over the next 1-3 quarters: their valuation support requires evidence that AI capex is producing either cloud acceleration, higher ad yield, or labor/productivity savings. Absent monetization disclosure, each additional capex-guide increase is likely to compress FCF yield even if revenue estimates rise. MSFT is comparatively insulated because enterprise software monetization provides a clearer path to recoup compute costs; it is the better defensive AI exposure if the market rotates from infrastructure spend to application ROI.

Consensus treats lower forward multiples as proof of cheapness, but estimate revisions are procyclical and may be most vulnerable precisely when capacity additions begin to exceed inference demand. The key falsifier for the constructive semiconductor view is a sequential deceleration in hyperscaler capex guidance or a meaningful increase in depreciation/revenue without corresponding cloud growth. Over 6-18 months, successful AI monetization would favor cloud platforms and power/cooling infrastructure more than merchant compute, since utilization—not chip shipment volume—becomes the binding economic variable.

Key Decisions for Investors

• Maintain a tactical long NVDA through the next earnings cycle only while hyperscaler capex guidance remains positive; use a 8-10% trailing risk limit or put spread hedge. Upside requires sustained system revenue and stable gross margin; a gross-margin guide-down tied to memory costs is the exit signal.
• Initiate a 1-3 month pair: long MSFT / short AMZN in equal dollar exposure. This isolates the monetization-versus-capex debate: MSFT has higher recurring revenue conversion, while AMZN needs AWS growth and retail efficiency to absorb investment. Reassess after both companies disclose capex and cloud-growth outlooks.
• Watch MU and SK Hynix as second-order beneficiaries of continued AI rack deployments; add only after confirmation that HBM contract pricing and bit-demand guidance are rising. The risk is that GPU demand pauses before memory capacity can be absorbed.
• Avoid adding broad mega-cap AI beta solely on aggregate forward-P/E arguments. Set an alert for sequential hyperscaler capex cuts or cloud revenue growth below guidance; either would likely trigger simultaneous estimate reductions across NVDA, AMD and AVGO, overwhelming apparent valuation support.