$60 Million AI Cluster with Nearly 10,000 Nvidia A100 GPUs Appears on eBay

Massive AI Infrastructure Hits Secondary Market

Core 4 Solutions, a Minneapolis-based IT hardware reseller serving major tech companies and government agencies, listed an entire decommissioned AI cluster on eBay this week for $59.7 million1

1

2

2

. The AI supercomputer cluster spans 300 server racks and came from a hyperscaler environment. The eBay listing ended on October 28, though it remains unclear whether the system found a buyer. Remarkably, US buyers were offered free FedEx Ground shipping on the massive installation.

Technical Specifications and Scale

The AI cluster contains 9,600 Nvidia A100 GPUs and 4,200 AMD EPYC CPUs distributed across its infrastructure2

2

. Each of the 300 racks houses four compute nodes, with every node equipped with eight Nvidia A100 40GB SXM4 GPUs connected via NVSwitch, eight NVIDIA ConnectX-6 200Gb/s network cards, two 64-core AMD EPYC Rome CPUs, an FPGA, and 1TB of DDR4 memory1

1

. Additionally, each rack includes three 1U management nodes with two more EPYC Rome CPUs apiece. The complete system totals 268,800 CPU cores, 384TB of GPU memory, and an impressive 1.4 petabytes of DDR4 memory overall1

1

.

Understanding the Nvidia A100's Role in AI Hardware

The Nvidia A100 GPU represents enterprise-level AI hardware built on the Ampere architecture. Launched in 2020, it served as Nvidia's flagship data center accelerator for AI workloads and data analytics before the Hopper and Blackwell architectures succeeded it2

2

. Each A100 contains 6,912 CUDA cores specifically designed for AI use cases rather than gaming applications1

1

. The asking price works out to approximately $6,219 per A100, though this figure includes all CPUs, network cards, power supplies, and server racks in the cluster2

2

.

Market Context and Industry Implications

Core 4 Solutions typically sells IT parts to major organizations including Amazon, Lucasfilm, Microsoft, Meta, and the US Department of Justice1

1

. While the company's website features servers priced at $87,000, this decommissioned AI cluster listing stands nearly 700 times more expensive1

1

. The appearance of such infrastructure on the secondary market highlights how quickly AI technology cycles through generations. As newer architectures emerge, even relatively recent systems like this A100-based cluster become available for resale. This development underscores the massive capital investments AI companies make in hardware and the rapid pace at which they upgrade their infrastructure to remain competitive in AI development and deployment.