Opening Keynote: HPC: A Heterogeneous Future

In this opening keynote, Martin Schulz of the Technical University of Munich and the Leibniz Supercomputing Center argues that heterogeneity is the inevitable future of high-performance computing. He traces how AI, data analytics, and traditional simulation are converging into fused, workflow-centric workloads that have already reshaped system design through GPU acceleration, changed I/O paths, and new operational modes, and contends that GPUs alone will not overcome the slowing of Moore's law, the power wall, and memory bandwidth limits. Instead, centers will manage portfolios of specialized accelerators. Schulz surveys quantum computing in depth, explaining superposition, entanglement, and gate-based operation, and describes the Munich Quantum Valley ecosystem, EuroHPC's growing fleet of quantum systems, and integration models ranging from network-attached NISQ devices to fault-tolerant QPUs requiring substantial co-located classical compute. He presents the Munich Quantum Software Stack, including the QDMI backend interface that lets multiple vendors plug into one compilation and resource-management pipeline, and draws parallels to neuromorphic platforms such as SpiNNaker 2 from TU Dresden, as well as photonic and DNA-based approaches. Key challenges he identifies are matching kernels to devices through co-design, programmability across divergent programming models, hierarchical scheduling that coordinates HPC, offload, and on-device schedulers to avoid fragmentation, and sustainable, long-term software maintenance beyond individual research projects. Audience questions address the energy balance of cryogenic quantum infrastructure, the timing of standardization efforts, and cross-center interconnection of quantum and neuromorphic resources.