China Fears Affect Nividia and AMD in Very Different Ways

NVIDIA reported $81.6B revenue (+85.23% YoY) including Data Center at $75.25B, but said it is excluding all China data center compute revenue from outlook and noted H20 shipments to China fell from $4.6B to zero. AMD posted $11.54B revenue (+50.11% YoY) with Data Center up 107% to $6.72B, effectively “barely flinching” versus the export shock, with China only mentioned around a prior $800M MI308 inventory charge. Net: geopolitical export controls are a near-term headwind for NVIDIA’s guidance (Q2 must extend $91B without China), while AMD’s outcome hinges on Helios ramp, raising relative uncertainty despite strong quarterly numbers.

Analysis

The market takeaway is not that China is “bad” for the AI cycle; it’s that NVDA can absorb a geopolitical haircut and still clear a premium margin hurdle. That improves quality perception, but it also increases the stock’s dependence on a narrow set of U.S. hyperscaler buyers, so the next 1-3 month risk is less China policy and more any softening in AMZN/META/cloud capex commentary. In other words, the immediate reaction should be calm, but the earnings base becomes more concentrated if the Western demand engine wobbles.

AMD’s setup is the opposite: China is mostly an opportunity cost, not a current cash-flow problem, which means the stock is trading more on execution optionality than on a tangible geopolitical P&L hit. That makes it a high-beta call on share gains, but the valuation leaves little room for a delayed ramp, software friction, or underwhelming attach rates in the next two quarters. If the market starts doubting Helios timing, the multiple can compress faster than the fundamentals would suggest.

The contrarian miss is that China removal may be incrementally positive for NVDA’s mix and regulatory overhang, while AMD’s “limited China exposure” does not become monetizable unless it wins real Western sockets. Over 6-18 months, the real separator is whether NVDA keeps resetting the bar on supply and networking attach, versus AMD converting roadmap promise into installed base. Falsifiers: any capex pause from AMZN/META, or evidence that AMD shipment cadence slips beyond the current ramp window.

Key Decisions for Investors

• Go long NVDA / short AMD in equal dollars for 1-3 months: relative-quality pair with NVDA trading at materially lower valuation and higher margin support; thesis breaks if AMD raises conviction on Helios revenue timing or NVDA reports hyperscaler capex deceleration.
• Add NVDA only on pullbacks, not strength: use a 3-5% dip after the next cloud-capex headline as entry, targeting a 10-15% move if Blackwell demand stays intact; cut if AMZN or META signal a capex step-down.
• Do not chase AMD pre-proof: wait for confirmed shipment cadence or first meaningful rack-scale revenue before initiating a long; if you must own it, prefer a limited-risk call spread rather than outright equity due to the ~182x earnings multiple.
• Set an alert on AMZN/META capex language and SOXX relative strength: any slowdown there is the earliest indicator that both names lose multiple support, with AMD likely the first to re-rate lower.