Cadence and NVIDIA push AI engineering

Cadence and NVIDIA said on April 15 they are widening their partnership from chip design tools into broader engineering software for robots, industrial systems and AI factories. (cadence.com) The basic idea is to let software agents do more of the engineering busywork while physics simulators check whether a design would work in the real world before anything is built. Cadence said it is combining its design software with NVIDIA CUDA-X, AI physics models and Omniverse digital-twin tools. (cadence.com) (nvidia.com) Cadence said those combined workflows target three areas: semiconductors, physical AI systems and hyperscale AI factories. NVIDIA described the same push in March as part of a broader effort with Cadence, Siemens, Synopsys and others to bring industrial software onto its accelerated-computing stack. (cadence.com) (nvidia.com) In plain terms, a digital twin is a high-detail software copy of a chip, machine or data center that engineers can test like a flight simulator. Cadence and NVIDIA are pitching that model as a way to move from design ideas to deployment faster and with fewer manual handoffs. (cadence.com) The companies attached big performance claims to the stack. Cadence said the expanded collaboration can speed some engineering workflows by up to 100X, and said in a separate March 17 announcement that its newer accelerated design systems can deliver up to 80X greater throughput and up to 20X lower power consumption. (cadence.com 1) (cadence.com 2) That March rollout was still centered on chip design. Cadence said then that its expanded portfolio runs on NVIDIA Grace central processing units and Blackwell graphics processors, and is also offered through the Cadence Millennium M2000 supercomputer. (cadence.com) By April, the message had widened. Cadence said customers including Ascendence, Argonne National Laboratory, Honda Research and Development, Samsung and SK hynix were already using Cadence tools accelerated by NVIDIA. (cadence.com) NVIDIA has been making the same case across industry this spring: software vendors build the engineering tools, while NVIDIA supplies the computing layer underneath. Its March 16 announcement named cloud providers Amazon Web Services, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure as delivery partners for these accelerated design and simulation workloads. (nvidia.com) Cadence is also tying the partnership to newer categories beyond factories and chips. NVIDIA’s industry page says the two companies are using the same accelerated-computing approach for data-center design, operations and drug discovery. (nvidia.com) The immediate takeaway is that Cadence is no longer selling AI mainly as a helper for chip engineers. It is selling a larger engineering stack, with NVIDIA hardware and software underneath, for simulating and optimizing whole systems before they are built. (cadence.com)