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

Nvidia reported $96 billion in fiscal Q2 revenue, up 106% year over year, with trailing-12-month revenue reaching $303 billion. AWS plans to deploy an additional 2 million GPUs through fiscal 2029, while Nvidia's non-hyperscaler revenue rose 138% to $40 billion, supporting diversification beyond major cloud customers. Management guidance implies roughly $700 billion of fiscal 2028 revenue versus analysts' $678 billion estimate, while the shares trade at about 15x next-year consensus earnings; risks include power constraints, higher memory costs and potential regulation.

Analysis

The key investable issue is not whether AI compute demand remains strong, but whether NVDA can convert a platform transition into upside while preserving gross margin. A faster mix shift toward rack-scale systems raises content per deployment and strengthens CUDA/networking lock-in, but it also makes reported margin more exposed to HBM and advanced-packaging availability; revenue upside without a stable gross-margin floor would not support the same earnings-multiple re-rating.

Diversification away from a handful of hyperscalers reduces customer-concentration risk, yet smaller cloud, sovereign, and enterprise buyers are likely more financing- and power-sensitive. Their demand can broaden the cycle over 6-18 months, but it is also the first cohort to defer deliveries if GPU utilization or AI application monetization disappoints. The more direct second-order beneficiaries are memory and power/cooling suppliers—MU, SK Hynix, VRT and ETN—where capacity constraints can create pricing power even if accelerator unit growth moderates.

Consensus appears focused on accelerator demand while underweighting the risk that customer-built silicon shifts the mix toward inference workloads where price/performance matters more than software inertia. This is not an immediate displacement thesis: custom ASIC programs at AMZN, GOOG and MSFT can depress their external-GPU intensity before they visibly affect NVDA revenue. Near-term, the decisive catalyst is the first clean platform-ramp quarter: shipment conversion, networking attach, gross margin, and supply commitments matter more than headline order commentary.

The bullish valuation argument should be treated as an estimate-quality question, not a standalone signal. Falsification would be a material guide miss, a gross-margin decline beyond management's transition framing, or hyperscaler capex plans shifting toward internally designed silicon; confirmation would be raised backlog conversion alongside sustained data-center margins through the new-platform ramp.

Key Decisions for Investors

• Maintain/add NVDA only on post-earnings evidence of platform-ramp execution; use a 3-6 month horizon and size against a defined downside trigger of a material revenue-guide miss or gross margin falling below the company’s stated transition range. Risk/reward is favorable only if estimate revisions follow, rather than merely multiple expansion.
• Pair long NVDA / short an equal beta-weighted basket of AMZN, GOOG and MSFT only if disclosed capex continues to accelerate while internal-silicon commentary remains limited. The pair isolates external AI infrastructure capture; exit if any two hyperscalers disclose meaningful production inference migration to proprietary accelerators.
• Build a 6-12 month satellite long in VRT and ETN rather than adding pure GPU beta after a sharp NVDA move. Data-center power-density constraints are a less discretionary bottleneck, though the thesis is invalidated by project cancellations or a clear decline in hyperscaler construction commitments.
• Monitor MU and HBM supply commentary as a hedge signal: sustained memory-price inflation supports memory suppliers but can cap NVDA incremental gross margin. Do not add a direct MU position without confirmation that HBM mix and contract pricing—not commodity DRAM strength alone—are driving revisions.