Solidigm Opens AI Central Lab to Advance GPU and Storage Testing

Solidigm, a storage technology company specializing in enterprise and hyperscale data environments, has opened what it calls the industry’s most advanced facility for testing storage in artificial intelligence workloads. The new AI Central Lab is located at the FarmGPU facility in Rancho Cordova near the company’s headquarters.

This AI Central Lab brings together high-performance GPUs and dense solid-state storage systems to study the real-world demands of AI model training, inference, and emerging workload architectures.

The AI Central Lab is designed to provide insights that extend well beyond component-level testing. Solidigm executives describe it as a resource to evaluate storage at the server, rack, and cluster level under conditions that mimic large-scale data centers. By aligning with common reference architectures, the facility gives customers and partners the ability to test solutions using Solidigm solid-state drives in environments that match production deployments.

Avi Shetty, Senior Director of AI Ecosystems and Partnerships at Solidigm, said the lab addresses a growing industry need. “Our AI Central Lab combines today’s most powerful GPUs with leading storage infrastructure to unlock new levels of testing and joint innovation for our customers and the developer community,” he explained. Mr. Shetty emphasized that the setup makes capabilities available that were previously accessible only to a small group of companies with large budgets. He also stressed the importance of placing storage close to GPU accelerators to maximize utilization and performance.

The hardware profile of the lab reflects that philosophy. Solidigm highlights two record-setting test clusters: one is the highest-performing storage test cluster ever built, anchored by the Solidigm D7-PS1010 SSD. In MLPerf Storage benchmark testing, the performance cluster achieved per-node throughput of 116 GB/s, with scalability across multiple nodes. The second is the densest storage test cluster yet, equipped with 192 Solidigm D5-P5336 SSDs, each providing 122 terabytes of capacity. The configuration delivers 23.6 petabytes of storage within just 16 rack units. Additional infrastructure includes NVIDIA B200 and H200 GPUs, 800 Gbps Ethernet networking, and storage servers from leading vendors.

Testing AI-Specific Workloads

Beyond performance, the lab enables exploration of power consumption and efficiency, a pressing concern as AI infrastructure scales globally. Engineers are able to benchmark system architectures for metrics such as tokens per watt and tokens per dollar – figures increasingly important for evaluating the economics of AI deployment. The lab is also being used to test AI-specific workloads such as retrieval-augmented generation (RAG), vector database tuning, and key-value cache offloading, all of which require careful orchestration of storage and compute resources.

Solidigm is already using the lab to collaborate with ecosystem partners. In one example, work with Metrum AI has shown that moving certain AI data operations from memory to SSDs can reduce DRAM usage during inference by as much as 57 percent in RAG scenarios. The company is also leveraging its partnership with FarmGPU to refine performance and efficiency strategies for AI infrastructure.

For Solidigm, the AI Central Lab is a strategic step in demonstrating how advanced storage design contributes to keeping GPUs productive and AI clusters efficient. By combining dense capacity with benchmark-setting throughput, and by creating a platform for testing real-world workloads, the company is positioning storage as a central enabler of the next generation of AI systems.