NVIDIA Ships Vera, Its First Agent-AI CPU, With First Server Delivered to AWS
NVIDIA has begun shipping Vera, its first CPU designed for agentic AI, with AWS receiving the first server. According to NVIDIA's official account and blog, Ian Buck personally delivered AWS's first Vera CPU server along with a Vera Rubin GPU to AWS headquarters in Seattle, handing them to Willem Visser and Supreeth.
Confirmed
- Vera features 88 custom Olympus cores and 1.2TB/s of memory bandwidth
- It targets the CPU bottleneck in agent workloads: as models shift from answering to acting, Python execution, tool calls, concurrency and real-time tasks demand far more from the CPU
- NVIDIA claims Vera delivers better tokens per dollar, faster user responsiveness, and a compute foundation for scaling AI factories
- The first delivery went to AWS, including a Vera CPU server and a Vera Rubin GPU
Why it matters
- This is NVIDIA's first CPU explicitly built for agent workloads, extending the AI compute race from GPUs to agent-optimized CPUs
- Multiple authors (e.g. @rohanpaulai, @timzaman, @nordicinst) note that compute demand in the agent era is shifting from model inference to environment interaction and code execution — exactly the gap Vera addresses
- The AWS delivery signals that top cloud providers are already bringing custom CPU servers into their agent infrastructure
2026-08-27 ~ 2026-08-28 · 6 related posts
Primary sources
- [source] NVIDIA Delivers First Agent-Built Vera CPU to AWS — nvidia · 2026-08-27
- [source] NVIDIA Blog: Why Vera CPU Is Built for Agents — nvidia · 2026-08-27
- Nvidia Ships Vera, Its First CPU Built for Agents — nordicinst · 2026-08-27
- [source] NVIDIA hand-delivers first Vera CPU server and Rubin GPU to AWS — timzaman · 2026-08-27
- NVIDIA Vera CPU ships at scale to accelerate agentic workloads — rohanpaul_ai · 2026-08-28
1 near-duplicate retellings: rohanpaul_ai