Nvidia announces technology to accelerate chip design at its Artificial Intelligence conference

March 21 (Reuters) – Nvidia Corp. (NVDA.O) CEO Jensen Huang on Tuesday unveiled a new product to accelerate semiconductor design while introducing a suite of new AI technologies based on its chips and software.

Huang said this at GTC, the company’s annual conference for software developers. Analysts are watching the Santa Clara, California-based company to provide more details on how it plans to expand access to processing power like that used to develop rapidly rising technologies like chatbot ChatGPT.

Last month, Huang told investors it would launch its own cloud computing service to offer more convenient access to large systems built using its chips.

Nvidia has come to dominate the business of selling chips used to develop generative artificial intelligence technologies, which can answer questions with human-like text or create new images based on a text message.

These new technologies rely on using thousands of Nvidia chips simultaneously to train AI systems on huge sets of data. For example, Microsoft Corp. (MSFT.O) built a system with more than 10,000 Nvidia chips for OpenAI to use in developing technologies that support the hugely popular ChatGPT.

On Tuesday, Huang announced a new technology called CuLitho to speed up semiconductor design and manufacturing. The software uses Nvidia chips to speed up the step between the software-based design of a chip and the physical manufacture of the lithographic masks used to print that design onto a piece of silicon.

The design of the masks must be modified to account for the fact that many features on the chip will be smaller than the wavelength of light used in the printing process, a step called optical proximity correction. Those calculations can take two weeks to complete, but Nvidia said on Tuesday that its chips and software can handle the task overnight and reduce electricity used from 35 megawatts to 5 megawatts.

Nvidia said it is working with ASML Holding (ASML.AS), Synopsys Inc (SNPS.O), and Taiwan Semiconductor Manufacturing Co (2330.TW) to bring the technology to market. Huang said TSMC will start preparing the technology for production in June.

Huang also said Nvidia is working with AT&T Inc (TN) to make trucking more efficient and with quantum computing researchers to speed up software development.

While Nvidia faces competition in the AI ​​chip market from Advanced Micro Devices (AMD.O) and several startups, the company owns more than 80% of the market for chips used to train AI systems.

The boom in artificial intelligence has helped Nvidia shares rise 77% this year, compared to an 11.5% rally in the Nasdaq Composite. With a market capitalization of $640 billion, Nvidia has grown to be about five times more valuable than longtime rival Intel Corp (INTC.O).

Additional reporting by Stephen Nelis in San Francisco; Edited by Bradley Perrett

Our standards: Thomson Reuters Trust Principles.