AI systems / GB10

NVIDIA DGX Spark

NVIDIA documents 1 or 4 TB self-encrypting SSD options, 10 GbE and ConnectX-7 networking. This is a desktop AI system built around NVIDIA’s GB10 Grace Blackwell platform.

Buying check: Check that the listing identifies the GB10 Spark, the included SSD and the power adapter. Partner-built GB10 systems use the same platform but have their own support terms.

Configuration at a glance

PlatformNVIDIA GB10: Arm CPU and Blackwell GPU
Unified system memory128 GB
Storage1 or 4 TB configurations
Operating systemNVIDIA DGX OS

Specifications and configurations can vary. Source: NVIDIA DGX Spark specifications.

eBay asking-price tracker

Tracked configuration: 128GB unified memory. SSD capacity and condition vary across listings; compare the complete configuration before buying.

$8,500 median asking price across 26 matching eBay Buy It Now listings, observed 2026-11-20. These are active asking prices, not completed sales.

Sampled range: $6,749–$10,018. Prices exclude shipping and tax. US listings in USD. Observation dates use UTC.

Samples refresh daily when the eBay API is available. Failed, stale or undersized samples do not set a current price. How price sampling works.

Find listings

These are search results, not verified offers. Confirm the exact system, memory, SSD, condition, warranty and return terms with the seller.

Affiliate disclosure: Affiliate (paid) links: we may earn a commission from eBay Partner Network and Amazon Associates at no extra cost to you. As an Amazon Associate I earn from qualifying purchases. Full disclosure.

What to expect for local AI

GB10 uses coherent memory shared by the CPU, GPU and system, rather than a separate discrete graphics card with its own VRAM. The operating system and runtime also use that memory. Model size, quantization, context and software affect what fits and how fast it runs.

DGX OS is Linux-based and GB10 uses an Arm CPU. Verify Arm support for the particular containers, libraries and tools you plan to run. A vendor’s peak FP4 performance figure is not a measured tokens-per-second result. NVIDIA’s porting guide explains the Arm and shared-memory architecture.

See the GB10 system comparison for alternatives, or compare the platform with a discrete GPU build.