Jensen Huang Used 1 Word to Describe AI Demand. It Could Be the Most Important of 2026.

Nvidia reported fiscal Q1 revenue of $81.6 billion, up 85% year over year, with data center revenue rising 92% to $75.2 billion and adjusted EPS jumping 140% to $1.87. The company raised its quarterly dividend 25-fold to $0.25 per share and authorized another $80 billion in buybacks, while guiding for about 95% revenue growth in the current quarter even assuming no China data center compute revenue. Management also cited about $1 trillion of revenue visibility across Blackwell and Vera Rubin from 2025 through 2027, reinforcing the AI buildout thesis despite China-related and customer capex risks.

Analysis

The market is still underestimating how much of Nvidia’s earnings power is being pulled forward by the shift from model training to agentic inference. That matters because inference is stickier, more distributed, and more embedded in enterprise workflows, which should extend demand duration even if hyperscaler capex growth slows. The bigger second-order effect is on the broader AI supply chain: rising utilization and cloud rental prices suggest the bottleneck is moving from chips to deployed capacity, which tends to support networking, power, and cooling vendors even if NVDA itself becomes more mature.

The China assumption is the key hidden risk, but it cuts both ways. Near term, excluding China from guidance creates a cleaner beat path if any incremental demand reopens, yet it also means the stock can de-rate quickly if investors conclude non-China demand is merely being front-loaded ahead of a digestion phase. The supply commitments and inventory build imply management is effectively writing a call option on continued hyperscaler spending; if that spending pauses for even one quarter, margin of safety compresses fast because the business is now operating with much higher fixed operating leverage than in prior cycles.

The contrarian take is that the market is focusing on revenue growth while ignoring the transition from scarcity to availability. If cloud rental economics keep improving, customers may optimize workload placement and reduce urgency for incremental chip purchases, particularly once custom silicon hits scale. That said, the rising older-chip rental rate argues against an imminent demand cliff, and suggests the installed base is still under-monetized rather than saturated.

Key Decisions for Investors

• Stay long NVDA on pullbacks only; use a 2-6 month horizon and size for high quality but lower multiple expansion from here. Risk/reward favors buying weakness rather than chasing strength because execution is excellent but expectations are now embedded.
• Pair trade: long AVGO / short NVDA for 3-6 months if you want AI exposure with less single-name capex risk. If hyperscaler spend broadens into networking and custom ASIC content, AVGO should participate with lower downside if AI hardware orders normalize.
• Long SMCI or a basket of AI infrastructure suppliers vs short software proxies over 1-3 months. If the build-out is still mid-cycle, the market should continue rewarding physical deployment and power/cooling names before the monetization layer catches up.
• Consider selling NVDA upside via covered calls or call spreads into strength over the next 4-8 weeks. Implied volatility is likely to compress after the print, and upside is increasingly dependent on a second-leg reacceleration that may not arrive until the next capex cycle.
• Do not short NVDA outright until there is evidence of hyperscaler digestion; instead use a trigger-based short if cloud capex commentary rolls over. The clean invalidation point is another quarter of accelerating data center demand and improving chip rental economics.