Rumor
Vera architecture reportedly doubles performance per watt over Grace
Monday, January 5, 2026 at 10:31 PM
The Vera architecture is reported to offer double the performance per watt efficiency compared to the Grace architecture.
Context
Nvidia has reportedly achieved a massive efficiency milestone with its upcoming Vera CPU architecture, which is set to deliver twice the performance per watt of the current Grace chips. This advancement is the cornerstone of the Vera Rubin platform, Nvidia’s next-generation data center solution succeeding the Blackwell series. By utilizing TSMC’s 3nm process and a custom 88-core design, the architecture specifically addresses the power consumption bottlenecks that are currently the primary constraint for scaling massive AI "factories."
The Vera CPU will feature a 1.8 TB/s NVLink interconnect, doubling the chip-to-chip bandwidth of previous generations to ensure seamless data movement between processors. Key specifications also include 4.2x higher memory capacity and 2.4x more bandwidth compared to Grace. Mass production is scheduled to begin in the second half of 2026, reinforcing Nvidia’s strategic shift to an aggressive annual release cadence. This transition to the Rubin era signifies a pivot toward chiplet-based designs to maintain a dominant lead in the global AI hardware supply chain.
Sources (3)
Related Companies
Nvidia
NVDA