Nvidia is a 'best in breed' stock on sale. Here's why I'm buying

Nvidia has authorized an additional $150 billion in share repurchases, lifting its remaining buyback capacity to $235 billion through fiscal 2028, equivalent to roughly 4% of its market value. The bullish analysis argues that NVDA trades at 16.7x forward earnings—its lowest valuation in at least a decade and well below its roughly 35x five- and 10-year average—despite 106% year-over-year revenue growth in the latest quarter and management's expectation for about 70% growth in fiscal 2028. Nvidia's roughly 97% server-GPU share, AI demand exceeding supply through 2028, and expanding Blackwell, Vera Rubin and software platforms support the positive outlook.

Analysis

The relevant signal is not the repurchase headline itself—an authorization spread across multiple years is unlikely to alter near-term float materially—but the market’s apparent shift from treating NVDA as a scarcity-growth asset to valuing it closer to a cyclical semiconductor. That creates upside if upcoming results demonstrate that system-level demand converts into sustained revenue rather than merely extended customer capex commitments. The key multiple catalyst over the next 1-3 months is evidence that product-transition execution preserves gross margin and that hyperscaler orders are not being pulled forward.

Second-order beneficiaries of sustained AI infrastructure intensity are MU and the HBM supply chain, where constrained qualified capacity can produce stronger incremental pricing than in GPUs. Conversely, a lower-cost NVDA equity valuation makes it harder for AMD and custom-ASIC alternatives to win share on total-cost-of-ownership arguments: customers are less likely to redesign workloads solely to diversify supply if NVDA can defend both platform economics and software lock-in. The more important competitive risk is not a near-term GPU substitute, but hyperscaler capex discipline after clusters are installed; that would hit NVDA’s earnings duration and memory suppliers simultaneously.

Consensus may be too focused on a headline forward P/E that is highly sensitive to the denominator and assumes a clean earnings ramp. A cheap multiple is not independently bullish if it reflects accelerating depreciation, power constraints, or a normalization in accelerator utilization. We would require independent confirmation through hyperscaler capex guidance, HBM contract pricing, and NVDA backlog-to-revenue conversion before underwriting a structural rerating. Falsifiers are a material gross-margin guide-down, a major cloud customer reducing AI capex, or evidence that lead times normalize faster than expected; any of these would likely compress both estimates and the multiple within days.

Key Decisions for Investors

• Initiate a 3-6 month long NVDA / short SOXX pair only after the next earnings report confirms gross-margin stability and forward demand visibility; target 10-15% relative outperformance, with exit if guidance implies material sequential deceleration or gross margin falls below management’s prior normalized range.
• Add MU as a higher-beta AI infrastructure adjunct on weakness ahead of the next memory-pricing datapoint, sized smaller than NVDA. The payoff is HBM and DRAM tightness translating into operating leverage over 6-12 months; stop the thesis on declining HBM pricing or customer inventory rebuilding rather than end-demand growth.
• Do not treat the buyback authorization as a near-term catalyst. Monitor actual quarterly repurchase deployment versus free cash flow and share issuance; weak execution would remove a key downside-support narrative without changing operating fundamentals.
• For event risk, use defined-risk NVDA call spreads dated 1-3 months beyond the next earnings release rather than outright calls if implied volatility is elevated. The trade requires verified estimates support; absent that, maintain NVDA on watch rather than chase a valuation-based rebound.