GPU Cluster Expansion

Summary

The tender covers the expansion of the Fraunhofer institute’s existing artificial intelligence infrastructure in Karlsruhe with one NVIDIA DGX B300 GPU system with 2,304 GB of memory, including the cables needed for connection. The supply also includes eight five-year NVIDIA AI Enterprise licences and five years of on-site 8x5x4 maintenance. The work includes installing, configuring and integrating the system with the existing NVIDIA DGX and NetApp storage infrastructure, installing software, testing and documentation. Project management and workshops are required, along with a one-day on-site initial training course for up to eight participants. No additional storage systems are being procured. Bidders must also price, as an option, 100 NVIDIA Learning Credits and up to 75 hours each of NVIDIA- and NetApp-certified system engineer services over five years; the contracting authority is not obliged to order these hours. All components must be delivered within no more than six months, and the system must be fully operational no later than two months after the hardware is delivered. Bids are evaluated 70% on price and 30% on qualitative award criteria; delivery information is worth up to 10 weighted points and the service concept up to 20. One responsible provider must offer the complete integrated solution. Participation requires the specified qualification and compliance documents, including turnover figures for the last three financial years, at least three DGX computing references and three storage references, and valid company and staff certifications. A service concept of no more than 15 A4 pages and delivery information are also required.