Several European AI clouds, server makers, and software companies have started building on NVIDIA Vera, the data center processor NVIDIA is positioning for AI agent workloads. NVIDIA published the list on October 20, 2026, ahead of GTC Berlin, with Nebius, Nscale, and Verda among the cloud providers, and Bull, Dell, HPE, Lenovo, and Supermicro among the systems makers.
The key facts about the NVIDIA Vera CPU in Europe in 30 seconds
- AI clouds Nebius, Nscale, and Verda are building infrastructure with Vera, and Aleria also appears in the ecosystem.
- Bull, Dell, HPE, Lenovo, and Supermicro are preparing Vera-based servers for HPC and AI.
- ClickHouse, SAP, Elastic, and Starburst report good results in early data-analytics testing.
- Daytona, E2B, Modal, and Vercel use Vera for the sandboxed environments where AI agents run code.
- NVIDIA has given no general-availability date or figures from its partners’ tests.
NVIDIA’s pitch starts from how AI agents work. Each task chains together steps that depend on the CPU: looking up information, calling tools, running code, and checking results. According to the company, Vera is designed to hold up that pace when thousands of tasks are running at once, and it can be used in CPU-only servers or alongside NVIDIA GPUs.
What Vera offers on paper
NVIDIA’s blog post credits the performance to per-core throughput, LPDDR5X memory bandwidth, and its NVIDIA Scalable Coherency Fabric, but it doesn’t include figures. The specs the company presented in July, as reported by Digital Citizen, describe 88 proprietary Arm-compatible Olympus cores, two tasks per core through Spatial Multithreading, up to 1.5TB of LPDDR5X memory on SOCAMM modules, and up to 1.2TB/s of bandwidth.
NVIDIA also claims Olympus delivers 50% more instructions per cycle than Grace, its previous CPU, and up to double the performance of current x86 server processors depending on workload and configuration. Those are the company’s own figures, pending independent testing. Revista Cloud has already reported how NVIDIA Vera beat Intel Xeon 6 in a DuckDB benchmark, in tests run by NVIDIA.
European clouds and server makers
In the European bloc, NVIDIA names Nebius, Nscale, and Verda as AI cloud providers building with Vera to cover the compute needs of enterprises and developers, with the CPUs handling data processing and software execution alongside the GPUs. Aleria, a sovereign AI infrastructure and platform provider, also appears in the ecosystem, though the blog post doesn’t specify what it’s doing with Vera. None of them has disclosed deployment volumes or dates.
The manufacturers named are Bull, Dell Technologies, HPE, Lenovo, and Supermicro, which are preparing Vera-based systems for HPC (high-performance computing) and AI data centers. Bull, the French supercomputing and server brand, is the only European manufacturer on the list. Other manufacturers have already shown hardware: Revista Cloud reported that ASUS packed two Vera CPUs into a 2U server for agentic AI.
Data, software, and sandboxes for agents
NVIDIA says data-processing companies ClickHouse, Cloudera, Duck Labs, Elastic, Polars, Redpanda, SAP, and Starburst have gotten good results in early tests with analytics, ETL (extract, transform, load), and semantic search workloads. The blog post doesn’t publish the results.
Among software companies, Atlassian uses Vera to speed up the data processing behind the semantic search indexes for its AI products. Inherent has tested the processor and plans to adopt it in settings where AI agents work alongside engineers on scientific research, and Nous Research uses it to authenticate agent identity and validate their actions before they’re executed.
The last group is cloud sandbox providers — the isolated environments where agents run code, use tools, and test results: Baseten-Blaxel, Daytona, E2B, Modal, Prime Intellect, and Vercel. NVIDIA stresses that these environments spin up millions of times a day, so startup time matters a lot at scale. None of these companies published improvement figures in the announcement.
Frequently Asked Questions
What is NVIDIA Vera?
It’s NVIDIA’s data center CPU, with 88 Arm-compatible Olympus cores and up to 1.5TB of LPDDR5X memory, built for AI agent workloads and designed to work alone or alongside GPUs.
Which European companies use NVIDIA Vera?
NVIDIA names AI clouds Nebius, Nscale, and Verda, sovereign AI provider Aleria, and Bull among server makers.
Which manufacturers will sell Vera-based servers?
Bull, Dell Technologies, HPE, Lenovo, and Supermicro are preparing Vera systems for HPC and AI data centers.
When will NVIDIA Vera be available?
NVIDIA hasn’t given a general-availability date in this announcement.
Sources:
