Jensen Huang explains why Nvidia will grow an astounding 70% next year

Nvidia CEO Jensen Huang reiterated confidence that revenue could rise 70% year over year in the next fiscal year, implying roughly $680 billion versus analysts' estimate of about $400 billion for the current year. He cited 27% month-to-month sales growth for a Grace Blackwell system incorporating 36 CPUs and 72 GPUs, alongside Nvidia's broad AI ecosystem presence. Huang said Nvidia has reviewed $100 billion of customer contracts tied to companies it invests in, while dismissing concerns over circular financing; competitive custom-chip efforts and eventual AI infrastructure efficiency remain longer-term risks.

Analysis

The key investable issue is not incremental demand visibility but the durability of NVDA's conversion of system-level demand into revenue and gross profit. A 70% growth outcome would force another material upward revision cycle, yet it also implies that customers must sustain exceptionally large capex budgets while absorbing rapidly rising power, networking, memory, and data-center integration costs. The near-term setup remains positive for NVDA and its critical supply chain—especially HBM providers and high-speed networking—but the earnings bar is now tied to shipment execution and customer monetization rather than merely AI spending announcements.

Hyperscaler custom silicon is more likely a margin and mix risk over 6-18 months than an immediate volume threat. AMZN, GOOG, and MSFT can deploy internal accelerators for mature, predictable inference workloads, reducing their marginal dependence on Nvidia; however, frontier training and flexible multi-model capacity still favor Nvidia's software and interconnect stack. The second-order constraint is power availability: if utility interconnects and data-center energization lag equipment deliveries, customers may defer system acceptance, shifting revenue recognition and leaving the most levered AI infrastructure names exposed.

The circular-financing concern should be treated as a credit-quality question, not dismissed as a rhetorical issue. Investments into AI-native customers may be economically rational if third-party contracted revenue is real, but investors need disclosure on customer concentration, financing terms, receivables, and whether demand is backed by end-user cash flow rather than venture capital. Any increase in inventory, receivable days, cancellation provisions, or financing commitments would be a more meaningful inflection signal than continued bullish demand commentary.

Consensus is likely underweight the possibility that Nvidia's strongest revenue guidance becomes a valuation headwind if it pulls forward 2027 demand. A beat-and-raise can still produce a muted stock reaction if gross margin declines faster than expected due to next-generation ramps, customer mix, or systems integration. Conversely, evidence that inference demand expands alongside training—not simply replacing it—would extend the revenue duration and weaken the custom-ASIC bear case.

Key Decisions for Investors

• Maintain an overweight NVDA through the next earnings cycle, but add only on post-results volatility rather than chasing guidance commentary. Base case is further estimates upside over 1-3 months; trim if data-center revenue guidance fails to rise commensurately with the implied 70% growth path or if gross-margin guidance falls materially below consensus expectations.
• Express the infrastructure bottleneck via a selective long NVDA / short CBRS pair over 3-6 months. NVDA retains the broader software, networking, and system-integration moat, while a pure accelerator challenger is more exposed to customer qualification delays and capital intensity; reassess if CBRS reports a named hyperscaler-scale production deployment or materially better gross-margin trajectory.
• Use AMZN, GOOG, and MSFT as a relative hedge against NVDA rather than outright AI shorts: long NVDA versus an equal-weight short basket of the three over the next 1-3 months only if hyperscaler capex guidance remains elevated. Exit the pair if any two of the three disclose accelerating internal-accelerator adoption with flat external GPU spend, which would signal a faster mix shift.
• Set an operational-risk watchlist for NVDA: customer receivable growth above revenue growth, rising inventory days, larger equity/financing commitments to AI customers, or disclosed data-center power delays. Any two signals would justify reducing exposure because they would challenge the quality and timing—not necessarily the existence—of booked demand.