Nvidia: AI speeds chip design

Designing a chip is usually a long sequence of drafts, checks, and fixes. Nvidia said one graphics processor unit design task that took eight engineers 10 months can now be done overnight with artificial intelligence tools. (tomshardware.com) A graphics processor unit is a chip built to handle many calculations at once, and its design process includes writing hardware logic, laying out circuits, and checking that the result can actually be manufactured. Nvidia research says its design automation work now spans those steps, from register-transfer level code to verification, physical design, and manufacturing checks. (research.nvidia.com) Nvidia’s own researchers have been building software agents for those jobs, including tools that generate hardware code and design-rule checking code with large language models. In a February 2025 technical post, Nvidia said one of those agents reached a 94.2% success rate on the VerilogEval-Human v2 benchmark. (developer.nvidia.com) The company is not saying artificial intelligence can design a whole chip alone. Tom’s Hardware reported Nvidia said it is still “a long way” from fully autonomous chip design without human input. (tomshardware.com) Nvidia is making that case as it pushes Blackwell, its current chip architecture for artificial intelligence and accelerated computing. Nvidia says Blackwell is built for generative artificial intelligence workloads and large-scale data center systems, while its RTX Pro Blackwell line targets workstation and professional graphics use. (nvidia.com (nvidia.com) That ecosystem push showed up in partner announcements on April 13, 2026. Premio said it validated Nvidia RTX Pro Blackwell graphics processor unit support across rugged edge and industrial systems for industrial automation, machine vision, on-premises generative artificial intelligence, and other edge workloads. (premioinc.com) The same day, Boost Run said it achieved Nvidia Exemplar Cloud status on Blackwell architecture. The company said the designation independently validated its cloud infrastructure performance and put it in a small group of providers using Blackwell-based artificial intelligence systems. (prnewswire.com) Nvidia has also been tying Blackwell to industrial software and design tools beyond its own labs. In March 2026, Nvidia said Cadence, Dassault Systèmes, Siemens, and Synopsys were building Nvidia-powered artificial intelligence agents to plan, optimize, and verify chip and system workflows. (investor.nvidia.com) The near-term picture is narrower than “AI designs chips now.” Nvidia is using artificial intelligence to compress specific engineering tasks, while customers and partners are lining up Blackwell systems to run those tools in data centers, workstations, and edge boxes. (tomshardware.com) (nvidia.com)