Nvidia's Forecast Assumes No Data Center Chip Sales to China. The Sept. 24 U.S.-China Summit Could Change That. - The Motley Fool

Nvidia's $108 billion Q3 revenue guidance—implying 89% year-over-year growth—carries a striking footnote: zero data center compute revenue from China. This is not conservatism; it is the arithmetic of... Read original →

Zero China data center revenue in a $108B quarter is not prudence—it is the arithmetic of a structural break. The pivotal signal is not the September 24 summit but the fact that USTR has already carved AI compute out of the negotiation perimeter, institutionally capping the probability of meaningful policy relief. The 25% tariff layered on case-by-case inspection renders even licensed H200 shipments marginally uneconomic at the gross-profit level. The deeper shift is irreversible: Huawei's Ascend 910C and SMIC's 7nm capacity have transitioned from roadmap items to active procurement options for ByteDance and Tencent, eroding CUDA's lock-in effect in the Chinese market over time. Even a full H20-era restoration adds roughly 7% to guidance—material but not transformative. At 14x forward earnings and a $5.4T valuation, the equity is priced on ex-China demand. China is an option, not a foundation. The 18-month question is not whether export controls ease; it is whether China's AI compute market has already closed its independent technology-commerce loop. If it has, Nvidia's re-entry window narrows permanently, regardless of policy.