Lenovo 30KL0002US ThinkStation PGX NVIDIA Grace Blackwell GB10 NVIDIA DGX OS World Wide Multiple

Lenovo 30KL0002US ThinkStation PGX NVIDIA Grace Blackwell Workstation

Overview

The Lenovo 30KL0002US is the ThinkStation PGX, a compact AI workstation built around the NVIDIA Grace Blackwell GB10 superchip and shipped with NVIDIA DGX OS — a purpose-built platform targeting AI inference, model fine-tuning, and edge compute workloads that need data center-class GPU architecture in a deployable form factor. The 30KL0002US ships factory-new and is sourced through established commercial distribution channels.

Connectivity is handled via PCIe and Ethernet interfaces, enabling integration into existing network infrastructure and expansion ecosystems. The unit weighs 9.00 lb, reflecting a compact physical footprint suited for under-desk, rack-adjacent, or edge-node deployments where space is constrained but GPU compute is non-negotiable.

Key Features

• NVIDIA Grace Blackwell GB10 Superchip: The GB10 integrates Grace CPU and Blackwell GPU architectures on a unified platform — this means unified memory across CPU and GPU workloads, which eliminates the PCIe bandwidth bottleneck that limits discrete GPU configurations when moving large tensors.
• NVIDIA DGX OS: Ships pre-loaded with DGX OS, a Lenovo- and NVIDIA-optimized Ubuntu-based environment with pre-configured AI frameworks, drivers, and container runtime — reduces time-to-first-inference compared to bare OS builds that require manual stack assembly.
• PCIe and Ethernet Interfaces: Dual interface types support both local peripheral expansion via PCIe and network-connected deployment via Ethernet, making the unit compatible with standard enterprise LAN infrastructure without additional adapters.
• 9.00 lb Deployable Weight: At 9 lb, this is a field-portable AI compute node — practical for distributed inference deployments, branch office AI, or temporary edge installations where rack space is unavailable.

Integration and Compatibility

The ThinkStation PGX 30KL0002US is designed for environments running NVIDIA-compatible AI frameworks (PyTorch, TensorFlow, Triton Inference Server) via the DGX OS software stack. PCIe connectivity supports expansion cards where supported by the chassis. Ethernet integration fits standard enterprise switching environments. Confirm specific NVLink, storage, and memory specifications with Lenovo's ThinkStation PGX product documentation before finalizing deployment architecture — the distribution feed data available does not enumerate all hardware sub-components.

Frequently Asked Questions

Q: What interfaces does the Lenovo 30KL0002US support?

A: The 30KL0002US supports PCIe and Ethernet interfaces, based on verified distribution feed data.

Q: What operating system does the 30KL0002US ship with?

A: The unit ships with NVIDIA DGX OS, an optimized AI-focused operating system based on Ubuntu.

Q: How much does the 30KL0002US weigh?

A: The unit weighs 9.00 lb per verified distribution specifications.

Q: Where is the 30KL0002US manufactured?

A: Country of origin is China (CN) per the distribution feed.

James Everett

The 30KL0002US is one of the first compact workstations to pair the NVIDIA Grace Blackwell GB10 superchip with a shipping DGX OS image — that combination matters because it compresses the gap between hardware delivery and a production-ready AI inference environment to a matter of hours rather than days of stack configuration.

Technical Highlights:

• Grace Blackwell GB10 Architecture: Unified CPU-GPU memory on a single superchip removes the discrete GPU's fundamental bottleneck — PCIe bandwidth — for large-model inference tasks where tensor transfers dominate latency.
• PCIe + Ethernet Dual Interface: Covers both local expansion and network integration without adapters, keeping deployment straightforward in standard enterprise environments.
• 9.00 lb Form Factor: Compact and portable enough for edge-node deployments where rack infrastructure isn't available — distributed inference at branch locations or temporary field installations.

Deployment Considerations:

• Confirm exact memory capacity, storage configuration, and NVLink topology directly from Lenovo's ThinkStation PGX product documentation before sizing for large-model inference — the distribution feed data does not enumerate these sub-components.
• DGX OS is Ubuntu-based but is not a standard Ubuntu build — validate compatibility with your existing MLOps toolchain and container registry before committing to this platform at scale.

The 30KL0002US fits best in distributed AI inference deployments — branch offices, edge compute nodes, or small team AI development environments — where data center GPU horsepower is needed but full rack infrastructure is not available or justified.