NVIDIA Vera Rubin Ramps Into Production
NVIDIA Vera Rubin enters production and, with Vera CPU and RTX Spark, expands data center and PC AI capacity, likely boosting token revenue expectations.
KEY TAKEAWAYS
- Vera Rubin ramped into full production with Taiwan server makers manufacturing systems at scale.
- NVIDIA said the platform delivers roughly 10x agent throughput at scale versus Grace Blackwell.
- Vera CPU was described as 1.8x faster than x86 and RTX Spark targets 1 petaflop PC AI compute.
HIGH POTENTIAL TRADES SENT DIRECTLY TO YOUR INBOX
Add your email to receive our free daily newsletter. No spam, unsubscribe anytime.
NVIDIA Corp. (NVDA) launched the Vera Rubin data-center platform at GTC Taipei on May 31, 2026, announcing the systems are now in production and positioned to increase data-center token revenue.
Vera Rubin Platform Enters Production
NVIDIA said the Vera Rubin data-center platform is ramping into full production, with Taiwan’s top server makers and global supply-chain leaders manufacturing systems at scale. The platform includes rack-scale supercomputer configurations such as the NVL72, designed for what NVIDIA calls “agentic AI factories”—large, clustered installations for autonomous AI agents and factory-style model deployment targeting AI labs, cloud providers, and hyperscalers.
The company described Vera Rubin as the POD-scale foundation for next-generation AI factories, delivering roughly 10 times the agent throughput at scale compared with the previous-generation NVIDIA Grace Blackwell platform. “POD” refers to standardized cluster building blocks for data centers. NVIDIA links this performance gain to higher monetizable AI output, or token revenue, per unit of infrastructure.
Vera CPU and RTX Spark Expand Compute Ecosystem
NVIDIA unveiled Vera as the CPU for agents, stating it is 1.8 times faster than x86 processors on a broad set of workloads. The chip aims to address general-purpose and agentic workloads historically run on x86 servers, expanding NVIDIA’s data-center compute beyond GPUs.
In collaboration with Microsoft, NVIDIA introduced RTX Spark, a new superchip and PC platform that reinvents Windows machines for local personal AI agents. RTX Spark-based systems can deliver about 1 petaflop of on-device AI compute and up to 128GB of unified memory. These systems target thin-and-light laptops with all-day battery life and highly efficient desktops for on-device agent workloads. Dell, Lenovo, and HP are named as initial OEM partners for RTX Spark-powered laptops.
NVIDIA also outlined NVIDIA OpenShell, a software layer designed to run secure agents on Windows and accelerate inference on RTX-class platforms by roughly twofold. This software and hardware suite extends NVIDIA’s RTX and DGX ecosystems for local-agent development.
The company used GTC Taipei, embedded in the Computex 2026 trade show, to present these launches as part of a broader strategic push around agentic AI, AI factories, and personal agents. The trio of products—Vera Rubin, Vera CPU, and RTX Spark—serve as infrastructure to increase monetizable output from data-center deployments and expand NVIDIA’s role across servers and client devices.
“Vera Rubin delivers the POD-scale foundation for next-generation AI factories—with 10x agent throughput at scale compared with the previous-generation NVIDIA Grace Blackwell platform,” NVIDIA said in its press release.
