[News] NVIDIA Reportedly Opens Vera CPU Sales to China as Early as August as H200 Shipments Stall - TrendForce

While NVIDIA faces challenges in delivering its H200 AI accelerators to China, the company may be opening a new channel via the Vera CPU, according to reports from Reuters, Tom’s Hardware, and Nikkei.... Read original →

NVIDIA’s pivot to Vera CPUs is a calculated workaround against U.S. export controls, shifting AI compute focus from training accelerators to inference-optimized processors. This move directly boosts demand for TSMC’s 3nm capacity and SK hynix’s HBM3E memory, reshaping China’s AI server supply chain. Compliance risks remain: if Vera’s software stack enables FP8 or sparsity acceleration, it could trigger renewed BIS scrutiny, raising localization costs. Intel and AMD will likely counter by fast-tracking AI-enhanced x86 CPUs with custom instruction sets, partnering with Chinese OS vendors to lock in ecosystems. Over the next 18 months, as large models shift toward edge inference, heterogeneous CPU-accelerator architectures will dominate—allowing NVIDIA not just near-term revenue but strategic control over the AI infrastructure stack, evolving from a GPU vendor into a full-stack compute architect.