NVIDIA Introduces Vera CPU to Chinese Customers: Single Chip Over $20,000, Initial Customers Plan Overseas Testing and Deployment

As reported by Reuters on June 12, three industry insiders have revealed that NVIDIA has launched a new central processor, codenamed Vera, for AI data centers to Chinese customers. Shipments are expected to begin as early as August, with orders currently open. NVIDIA is rolling out this new product rapidly to recapture its share in China, after shipments of its H200 chip to the region were blocked—a move that pushed NVIDIA’s Chinese market share to nearly zero. Insiders noted that some Chinese customers have expressed interest in this smart AI-specific CPU. A major Chinese cloud services firm plans to order more than 300 servers, each equipped with two Vera processors. The client intends to deploy the units first in overseas data centers for testing, and will place additional orders based on evaluation results. However, uncertainties remain over Vera’s widespread adoption, including challenges related to software ecosystem compatibility and the difficulty of migrating AI workloads to alternative AI chips. According to semiconductor research firm SemiAnalysis, the list price of a single Vera processor (before bulk discounts) is well over $20,000, with a full rack configuration containing 256 chips costing roughly $10 million. Current global AI demand is shifting from model training to inference computing, which is worsening CPU shortages. NVIDIA has projected that Vera’s sales will hit $20 billion by the end of its fiscal year ending January 2026. NVIDIA’s launch has also intensified competition with Intel and AMD, both of which are ramping up supplies of AI data center server CPUs. For more details, click the original article link to join BlockBeats’ Feishu AI News Channel, which monitors AI hotspots and news 24/7.