Because of the increasing complexity of chip designs, developing a new chip can take months or even years if firms don’t forecast and install enough fixed compute capacity on-premises.
Two decades ago, a simulation could be run with only a handful of CPUs on a single server. These days, it requires partitioning a design and running multiple simulations simultaneously across dozens of CPUs. Few IT teams or data centers at chip manufacturers can keep up.
If engineering teams don’t have the customized resources they need to complete each step in the chip design flow, development will slow down, and products will miss their launch dates. As a result, computing and resource limitations have become significant bottlenecks.
Modern chip design needs a high-performance computing (HPC) environment, which requires enormous capital investment in servers, networking equipment, data-center infrastructure, utilities, and IT staff. In addition, it takes time to specify, commission, setup, and provision servers. In the meantime, engineering teams are competing for capacity and are limited in their ability to do their jobs.