How the PLA Continued Procuring Nvidia AI Chips After U.S. Export Controls

WireScreen analyzed more than 3,800 Chinese military and government procurement records to uncover how the People’s Liberation Army continued acquiring advanced Nvidia AI accelerators — including A100, H100, A800, and H800 GPUs — years after U.S. export controls were imposed.

‍

The report traces a 374-company procurement ecosystem spanning PLA units, defense universities, state-owned enterprises, shell companies, and commercial intermediaries. In many cases, procurement notices named Nvidia products and were published directly through official Chinese military procurement channels.

‍

The findings show that export controls introduced friction and shortages — but did not stop Chinese military access to advanced U.S.-origin compute.

‍

Key Findings

• Export-controlled Nvidia GPUs continued appearing in PLA procurement records after successive rounds of U.S. restrictions.
• Procurement activity spans operational military units, cyber warfare entities, and Entity-Listed defense institutions.
• Military buyers increasingly relied on intermediaries, shell firms, and compute-as-a-service providers to obtain advanced AI infrastructure.
• Procurement records frequently disguised American accelerators inside “domestic” Chinese server systems.

‍

About the Author

John Costello
Director of Strategic Affairs, WireScreen

‍

Former senior official at the White House Office of the National Cyber Director (ONCD) and the Cyberspace Solarium Commission, with extensive experience in U.S. cyber policy, national security, and intelligence.

‍

Request access to download the full report.

‍