BlockBeats News, June 12th, according to Reuters, three insiders revealed that NVIDIA has introduced a new Vera central processor for AI data centers to Chinese customers, stating that shipments could start as early as August and customers are currently allowed to place orders. NVIDIA's rapid introduction of the new product aims to regain its market share in China, as its shipments of the H200 chip to China have been blocked, causing NVIDIA's market share in China to drop to nearly zero.
Insiders said that some Chinese customers have shown interest in a dedicated CPU designed for smart AI. A large Chinese cloud service company plans to order over 300 servers, each equipped with two Vera processors. The customer plans to first deploy them in overseas data centers for testing, and then decide whether to place additional orders based on the evaluation results. Due to software ecosystem compatibility and the challenges of migrating AI workloads to domestic AI chips, whether Vera can be widely adopted is still uncertain.
According to the semiconductor research firm SemiAnalysis, the price of a single Vera processor before bulk discounts is estimated to be well above $20,000, while a full rack configuration with 256 chips is priced at around $10 million. Currently, global AI demand is shifting from model training to inference computing, exacerbating the CPU shortage. NVIDIA has predicted that Vera's sales will reach $20 billion by the end of this fiscal year as of January 2026. NVIDIA's introduction has intensified its competition with Intel and AMD, as both companies are ramping up the supply of server CPUs for AI data centers.
Click on the original article link below to join BlockBeats' Feishu AI News Channel, where AI hotspots and news are monitored 24/7.
