Jensen Huang Admitted Nvidia's China Revenue Has Fallen to Zero. Here Is His $20 Billion Comeback Plan.

Nvidia said its new Vera CPUs open a $200 billion addressable market and are expected to generate nearly $20 billion in revenue this year, helping offset the collapse in China sales tied to export restrictions. The Vera Rubin AI computing platform, built for agentic AI, is another growth catalyst and is expected to ship later this year with up to 35x higher inference throughput. Analysts still project full-year revenue growth of 81% to $391 billion, supporting a bullish long-term outlook despite continued uncertainty in China.

Analysis

The strategic read-through is that NVDA is converting geopolitical damage into platform lock-in: lost China revenue is being replaced by higher-quality, software-adjacent compute demand where the company controls more of the stack. That matters because CPU attach rates should improve gross margin mix and reduce cyclicality versus standalone GPU sales; the market is still underappreciating how much of the next leg is coming from system-level selling rather than chip unit growth.

Second-order beneficiaries are the ecosystem suppliers that ride NVDA’s rack-scale standardization: high-speed interconnect, advanced packaging, HBM, and power/cooling infrastructure should see less price sensitivity than the core compute names. By contrast, INTC and AMD are not just competing on x86 share; they are competing against an increasingly “good enough” CPU narrative inside AI clusters, which compresses TAM for general-purpose server CPUs over the next 12-24 months.

The key risk is not product demand, but execution timing and policy reversals. If export rules loosen even modestly, headline growth could re-rate near-term numbers, but if China remains shut and U.S./EU procurement scrutiny expands, the multiple expansion may stall despite strong fundamentals. The market’s mistake is treating China as a lost end-market when it may actually be a forcing function that accelerates NVDA’s move up the stack and deepens dependency among hyperscalers.

Contrarian angle: the stock may be cheaper than it looks if Vera Rubin is the first evidence of a durable systems franchise, but the consensus is still pricing NVDA like a component supplier rather than a platform owner. That creates asymmetric upside over 6-12 months if inference workloads inflect, while the main downside is a delay in shipping cadence or any sign that customers start multi-sourcing around NVDA’s integrated architecture.

Key Decisions for Investors

• Maintain/accumulate NVDA on pullbacks over the next 2-6 weeks; risk/reward is favorable if the market keeps underpricing systems revenue and CPU attach, with upside from a re-rating rather than just estimate revisions.
• Initiate a pair trade: long NVDA / short INTC for 3-6 months. Thesis: AI server CPU share and platform integration should continue to pressure INTC’s enterprise relevance while NVDA monetizes the full rack.
• Consider a smaller long NVDA / short AMD pair for 3-6 months if you want cleaner exposure to AI platform share gains; AMD’s risk is valuation compression if investors conclude the AI CPU opportunity is being captured more by integrated systems than by standalone CPUs.
• Look for a basket long in NVDA-adjacent infrastructure winners over the next 1-3 months: networking, advanced packaging, and HBM suppliers. These names should capture second-order spend even if NVDA absorbs the headline.
• Use call spreads rather than outright longs if entering ahead of product shipping milestones: e.g., 3-6 month NVDA call spreads to express upside from Vera Rubin while capping risk around launch timing or policy headlines.