China Halts NVIDIA AI GPUs

NVIDIA just hit a triple whammy: China froze direct sales of its prized AI chips, Alphabet started letting select customers run Google’s own TPUs in private data centers, and a subtle laptop memory tweak is pointing to supply constraints in the semiconductor world. With U.S. export controls blocking key NVIDIA models, China’s AI chip market—worth up to $50 billion—has slammed the brakes, leaving billions in inventory stranded at the border. Meanwhile, Chinese firms like Huawei are seizing the moment, using homegrown accelerators for AI inference and scaling up production, potentially shifting the global balance of power in AI hardware.

But here’s the catch: while NVIDIA’s chips still dominate for training AI models, Huawei’s platforms are winning for the actual deployment, or inference, of those models. If the local ecosystem follows this trend, Morgan Stanley says domestic suppliers could capture 86% of China’s AI chip market by 2030. At the same time, Google is moving to challenge NVIDIA’s data center dominance, selling TPUs to a small but influential group of customers with strict privacy or latency needs—changing the conversation about who controls the future of AI infrastructure. All eyes are on whether NVIDIA’s chips clear customs in China—one signal that could reshape the global AI race.

Based on reporting from The Globe and Mail, Morgan Stanley, and insights from Jensen Huang and Sundar Pichai.

Powered by Apisod.com