Hammerspace, a high-performance data platform for AI, has introduced a comprehensive solution to streamline enterprise data access tailored for agentic AI applications. Aligned with NVIDIA's AI Data Platform reference design, this solution adeptly utilizes current infrastructure, obviating the need for costly overhauls or new storage systems, and empowers enterprises to efficiently utilize existing data.
As an esteemed member of the NVIDIA Inception programme, Hammerspace seamlessly unifies unstructured enterprise data across disparate storage architectures. This platform turns raw data into AI-ready intelligence with speed and efficiency, underpinning Retrieval-Augmented Generation (RAG) and other AI-driven processes.
Traditional AI infrastructure often requires data duplication into specialised silos, creating fragmentation. Hammerspace tackles this issue by offering a single global namespace spanning both on-premise and cloud setups. Automated data objectives intelligently tag and tier data, optimising performance and cost while negating the need for manual data handling.
Through support for pNFS, NFS, SMB, and S3 protocols, the platform ensures existing enterprise application compatibility and immediate data access.
Hammerspace's Data Platform leverages NVIDIA AI Enterprise software, paired with cutting-edge NVIDIA hardware for unmatched performance. Incorporating RTX PRO6000 Blackwell GPUs, Spectrum-X Ethernet, alongside integral microservices, it accelerates deployments and refines model accuracy.
At the core, the Tier 0 component unifies NVMe inside GPU nodes, enhancing performance and resource utilisation, while integrated vector database services create searchable embeddings, facilitating seamless business data access.
Delivering a validated, easy-to-deploy solution, Hammerspace's Data Platform facilitates scalable AI adoption in small, project-based configurations through strategic partners. It offers enterprises a practical means to bridge data access challenges, transforming owned data into AI-ready intelligence.
Hammerspace's reference design will make its debut at NVIDIA GTC in Washington D.C., with full availability projected for late 2025 through authorised partners.
