Nvidia CEO Jensen Huang has unveiled new cloud computing services with Oracle and others as it rolled out new artificial intelligence technologies based around its chips and software.
Nvidia is working with Microsoft and Alphabet as well to widen accessibility to the massive systems with tens of thousands of chips like those used to develop fast-rising technologies such as the chatbot ChatGPT, Huang said in the virtual keynote address at GTC, the company’s annual conference for software developers.
Nvidia is partnering with AT&T to make dispatching trucks efficient, collaborating with quantum computing researchers to speed software development, and working with industry giant Taiwan Semiconductor Manufacturing (TSMC) to speed up chip development.
Nvidia has paired with a number of large image copyright holders to resolve the legal uncertainties around image-generation technologies. It also announced new chips that would help make running services similar to ChatGPT much cheaper, working with Google’s cloud unit.
Nvidia has come to dominate the field for selling chips used to developing generative AI technologies, which can answer questions with human-like text or generate fresh images based on a text prompt.
Those new technologies rely on the use of thousands of Nvidia chips at once to train the AI systems on huge troves of data. Microsoft, for example, built a system with more than 10,000 Nvidia chips for startup company OpenAI to use in developing technologies that underpin its wildly popular ChatGPT.
Nvidia released a new service called DGX Cloud that it said would give companies and software developers access to supercomputer power by logging on through a browser.
Nvidia said it would work with partners to host the service, starting with a deal with Oracle, which would offer a DGX-based supercomputer that can string together more than 32,000 of Nvidia’s chips at once.
A single instance of the cloud service – which consists of eight of Nvidia’s flagship A100 or H100 chips strung together with its custom networking technology – starts at $36,999 per month, Nvidia said.
Biotech firm Amgen and software firm ServiceNow have started using the service, Nvidia said.
Nvidia released a service called AI Foundations to help companies train their customized artificial intelligence models, enabling them to create products similar to ChatGPT or the Dall-E image creation system but fine-tuned using their own proprietary data.
Huang announced technology to speed up the design and manufacturing of semiconductors, working with ASML Holding, Synopsys and TSMC to bring it to market.
TSMC will start readying the technology for production in June.