It is common knowledge that desktop computing power is now increasing mainly by the change to multi-core chips. This is a challenge for the software community in general, but is a particular problem for audio processing. Our needs are increasingly towards real-time and low latency. We propose a number of possible paths that need investigation, including multi-core and special accelerators, which may offer useful new musical tools. We define this as High-Performance Audio Computing, or HiPAC, in analogy to current HPC activity, and indicate some on-going work.