Nvidia has just announced RTX Spark – an AI server technology that is reaching the consumer realm with a Grace CPU (20 cores), a Blackwell GPU (6,144 CUDA cores) and 128GB of LPDDR5X. Now here’s what’s next for servers – and maybe one day for consumer devices too.
The new Vera CPU Vera is half the CPU of the Rubin platform – the other being the Rubin GPU. Vera promises 1.8x average speedups on “leading x86 CPUs” (Nvidia didn’t explicitly name them).
Vera is huge – it has 88 Olympus cores (based on the ARM instruction set) with spatial multithreading for 176 threads per socket. The processor can be paired with up to 1.5TB of LPDDR5X RAM, which can provide up to 1.2TB/s of bandwidth, which is important for AI inference.
Vera can be used as a standalone CPU for agentic AI workloads, reinforcement learning, data processing, and analytics. Nvidia has also designed it vera cpu rackWhich has 256 CPUs for a total of 22,528 cores and 45,056 threads (oh boy).
Alternatively, Vera can be a host CPU used in conjunction with the Rubin GPU. For example, NVIDIA Vera Rubin NVL72 It has 36 Vera CPUs and 72 Ruby GPUs. The CPU and GPU can talk to each other at 1.8TB/s using the Nvidia NVLink-C2C interconnect.
Nvidia has already secured major customers – Anthropic (Cloud), OpenAI (ChatGPT) and SpaceXAI (Grok) will use Vera CPUs and so will hyperscalers like ByteDance, CoreWave and Oracle Cloud Infrastructure.
Additionally, Dell, HP, Lenovo, and Supermicro will manufacture standalone Vera CPU systems. Also: Asus, Compal, Foxconn, Gigabyte, Pegatron, Quanta Cloud Technology, Wistron and Wiwin. Even the New York Stock Exchange is interested in this – NYSE Procedures 1.1 trillion messages per day, so it’s working with Redpanda and HP to build new infrastructure.
