Nvidia Has $96 Billion in Quarterly Revenue. Here's Why the Stock Is Still a Buy.

Nvidia generated $96 billion in fiscal Q2 revenue, up 106% year over year, and management guidance implies roughly $700 billion of fiscal 2028 revenue versus Wall Street's $678 billion estimate. Non-hyperscaler revenue rose 138% to $40 billion, while AWS plans to deploy an additional 2 million GPUs through fiscal 2029, supporting demand visibility beyond core cloud customers. Vera Rubin began shipping in August and is expected to contribute 20% of data-center revenue in fiscal Q3; Nvidia trades at roughly 15x next-year consensus earnings despite potential upside to estimates. Risks include data-center capex variability, power constraints, rising memory costs pressuring gross margin, and possible regulation.

Analysis

The actionable issue is not the headline valuation claim but whether the Street has modeled the platform transition correctly. If NVDA's implied revenue outlook is credible, consensus is likely missing both accelerator units and higher-value networking/system content; that creates a two-step catalyst path: estimate revisions around the next earnings print, followed by FY28 margin debate as Rubin mix rises. The market will reward a beat-and-raise only if supply availability—not merely customer demand—supports backlog conversion.

The customer-mix shift reduces the obvious hyperscaler concentration risk but introduces a different one: neocloud, sovereign, and startup demand is more financing-sensitive and may carry materially weaker credit quality than AMZN/MSFT/GOOG. A downturn in GPU leasing economics, higher funding costs, or softer AI inference monetization could cause this cohort to defer deliveries, leaving hyperscalers as the marginal demand setter. Memory inflation is also more consequential than consensus framing suggests: a 200-300bp gross-margin miss during a new-platform ramp can constrain EPS upside even while revenue beats.

Second-order beneficiaries are memory suppliers and AI infrastructure vendors rather than the hyperscalers themselves. NVDA shipment growth increases HBM and advanced packaging bottlenecks, while power and data-center build constraints can shift spend from accelerators toward networking, cooling, and electrical equipment. Conversely, internally developed accelerators at AMZN, GOOG, and MSFT become strategically more valuable if GPU lead times or total-cost-of-ownership rise, though they are not near-term substitutes for the broadest training workloads.

Consensus appears too binary—either hyperscaler capex sustains indefinitely or an AI bubble bursts. The more probable 1-3 month outcome is continued spend with volatile delivery timing; NVDA can beat revenue while shares de-rate if gross margin, receivables, or customer prepayments imply lower-quality growth. The thesis is falsified by a material reduction in next-12-month data-center outlook, Rubin contribution below management's ramp target, or hyperscaler capex guidance turning flat sequentially.

Key Decisions for Investors

• Maintain/establish NVDA long only on post-earnings volatility or a 10-15% drawdown; target 15-20% upside over 3-6 months from estimate revisions, with a stop/review trigger if data-center gross margin misses consensus by more than 250bp or forward revenue guidance fails to rise.
• Express relative AI exposure through long NVDA / short equal-dollar MSFT or GOOG for 1-3 months if the objective is to isolate hardware estimate revisions from broad AI sentiment; cover the short if either company discloses accelerating proprietary-accelerator adoption or materially raises capex guidance.
• Do not chase AMZN solely on planned GPU deployment: the spend is a capex and depreciation headwind before it becomes an AWS revenue catalyst. Reassess after AWS discloses utilization, pricing, and incremental AI-service revenue sufficient to offset infrastructure depreciation.
• Set an alert around HBM supply and pricing data; confirmed tightening supports a basket of memory exposure, but avoid a direct recommendation until vendor-level contract pricing and capacity additions are verified. A rapid capacity expansion would remove the bottleneck premium and pressure NVDA system margins.