# Supermicro ships PCIe Gen 6 all-flash storage servers for AI

Published 2026-11-20 · AI.info Team · AI.info

> Supermicro has begun shipping Petascale all-flash storage servers with AMD EPYC 9006 processors. The company says the systems offer up to 100% greater sequential read bandwidth than PCIe Gen 5.

[Supermicro](https://www.supermicro.com/en/pressreleases/supermicro-ships-first-all-flash-storage-servers-delivering-pcie-gen-6-performance) has begun shipping all-flash storage servers with PCIe Gen 6 support for workloads including AI training and inference. In a release, the company said the systems deliver up to 100% greater sequential read bandwidth than PCIe Gen 5.

The systems are part of an expansion of Supermicro's AMD EPYC-based storage portfolio. The shipping Petascale system uses sixth-generation AMD EPYC 9006 processors, previously codenamed Venice; more products are to follow as AMD launches new SP8 processor versions.

## Capacity and networking

The first system in the new family has a 2U chassis supporting up to 32 E3.S NVMe solid-state drives. With 122 TB drives, it can provide up to 3.93 PB of capacity.

The system's input/output architecture supports up to four high-performance x16 networking cards. Supermicro says the design is intended to provide sustained bandwidth for applications including AI data lakes, large-scale analytics and high-throughput storage networking, while allowing customers to configure systems for their workloads and networking requirements.

The company is showing the Petascale all-flash server at Supercomputing 2026 in Chicago from Nov. 15 through Nov. 20. The products are designed to be customizable for high-performance computing, AI training and inference workloads.

## Software partners

Supermicro says it is validating the platform, which includes its CloudDC, GrandTwin and Hyper families, with software-defined storage partners for large-scale AI deployments. The partners named in the release are Hammerspace, [IBM](https://ai.info/companies/ibm), MinIO, Nutanix, Scality, VAST Data, VDURA and WEKA.

Hammerspace says its Data Platform can incorporate the new storage into a single global namespace alongside existing storage capacity. IBM Storage [Scale](https://ai.info/companies/scale-ai) on the AMD-based Petascale server is designed for an inference pipeline spanning data ingestion, model loading, retrieval and KV cache, according to IBM.

MinIO says its AIStor and MemKV products can be combined with Petascale for data and memory management across ingestion, training and inference. Nutanix described a combination of its software-defined storage, AMD EPYC processors and the Petascale platform for workloads including real-time inference and retrieval.

## Other platform combinations

VDURA says its software-defined data platform runs on the latest Petascale systems using AMD's Venice processor. Its approach brings Supermicro all-flash and hard-disk-drive platforms together in a single VDURA namespace.

WEKA says its NeuralMesh software is distributed and can scale across the new Petascale server. VAST Data says it is combining its AI Operating System with the AMD EPYC-based hardware as part of its partnership with Supermicro.

Scality identified its Autonomous Data Infrastructure and AI Inference Factory products as storage solutions for AI. It said the latest Petascale platform provides a foundation for those products as part of its continuing collaboration with Supermicro.

## Sources

1. Original source · Press release [Supermicro](https://www.supermicro.com/en/pressreleases/supermicro-ships-first-all-flash-storage-servers-delivering-pcie-gen-6-performance) supermicro.com

Other pages read

2. [Supermicro (PR Newswire)](https://www.prnewswire.com/news-releases/supermicro-levert-eerste-all-flash-opslagservers-met-pcie-gen-6-prestaties-en-brede-ondersteuning-voor-softwaregedefinieerde-opslag-voor-ai-training-en--inferentie-302946815.html) prnewswire.com

## Related stories

- [Cerebras puts split AI inference into production with AMD GPUs](https://ai.info/news/cerebras-disaggregated-inference-amd-production)
- [FlashInfer on ROCm routes AMD GPU kernels automatically](https://ai.info/news/flashinfer-rocm-automatic-kernel-routing-amd-gpus)
- [Unsloth Desktop adds local inference on AMD XDNA 2 NPUs in Ryzen AI PCs](https://ai.info/news/amd-unsloth-xdna-2-npu-agentic-ai-apps)
- [vLLM details opt-in FlyDSL MonoKernels for faster decode on AMD GPUs](https://ai.info/news/vllm-flydsl-monokernels-agent-inference)
- [AMD publishes Nitro-P for pixel-space image generation](https://ai.info/news/amd-nitro-p-pixel-space-image-generation)

## Explore

[More articles](https://ai.info/news)

---

Canonical: https://ai.info/news/supermicro-pcie-gen-6-all-flash-storage-servers
