Nvidia and AMD Can Recapture $4.5B and $700M from China Sanctioned Chips by Selling Them to Data Centers as Inference Accelerators

As detailed in several of my earlier Substack articles on Nvidia’s H800 and H20 sanctions (Nvidia’s H20 Crackdown Was Inevitable—U.S. Licensing Marks the End of Loophole-Led AI Sales to China), as well as AMD’s MI300 export risks (Nvidia's A100 And H100 AI Chips Are Another Failure Of U.S. China Sanctions), the chips blocked by U.S. trade policy are not obsolete. On the contrary, they are being repurposed for inference workloads—a strategic shift that could allow Nvidia and AMD to recover billions in stranded revenue. This article provides a structured financial and technical analysis of how that reallocation plays out.

Strategic Revenue Recovery from Sanctioned AI Chips

Table 1 outlines the known and estimated financial consequences of the U.S. export ban for Nvidia and AMD. The values are sourced from company filings and updated April 2025 disclosures. While Nvidia reported a $5.5 billion charge tied to lost H20 sales, the company had originally targeted $12 to $15 billion in China-based revenue. AMD likewise disclosed up to $800 million in exposure related to MI308 (a variant of the MI300X). Reallocation for inference could recover a substantial portion of those losses in unrestricted markets.

Chart 1 compares their expected export losses due to U.S. sanctions with the recoverable revenue from repurposing the same chips for inference workloads in domestic and allied markets. Despite the initial hit, both companies stand to recoup a large share of their stranded inventory through new demand.