Nvidia has started marketing its new Vera AI data center CPU to Chinese cloud vendors, with shipments anticipated as early as August. The company is currently accepting orders, and several Chinese customers have shown interest. A major cloud vendor plans to test over 300 servers equipped with dual Vera CPUs before considering large-scale deployment. The Vera CPU, based on the Arm architecture, is designed for academic AI applications and boasts performance up to 1.8 times faster than competitors. Each CPU is priced significantly above $20,000, with a full rack of 256 CPUs costing around $10 million. Nvidia projects Vera sales revenue to reach $20 billion this fiscal year.