On Tuesday, Nvidia Corp (NVDA.O) Chief Executive Jensen Huang announced plans to make the powerful and pricey supercomputers used to create AI technologies such as ChatGPT accessible for leasing to practically any firm.
While that access will not be cheap – $37,000 per month for eight of Nvidia’s flagship A100 or H100 chips strung together—making it available to a broader range of business customers could accelerate the AI boom that has driven Nvidia shares up 77% this year, making it roughly five times more valuable than longtime rival Intel Corp. (INTC.O).
The Santa Clara, California-based company already dominates the artificial intelligence chip market. It has assisted partners such as Microsoft Corp (MSFT.O) in building massive systems for ChatGPT creator OpenAI’s services to answer questions with human-like text and generate images from prompts.
DGX supercomputers
At Nvidia’s annual software development conference on Tuesday, Huang said that the firm was working with partners such as Oracle Corp (ORCL.N) to provide access to Nvidia’s DGX supercomputers with up to 32,000 Nvidia processors to anybody with a web browser.
“The iPhone moment of AI has begun,” Huang said during the virtual keynote talk, alluding to how Apple Inc (AAPL.O) pioneered the smartphone industry.
Huang also said that Nvidia was collaborating with Microsoft and Google Inc (GOOGL.O) to provide its supercomputers, which are used to develop new AI technologies, as a service. Nvidia revealed new processors and software on Tuesday that would make products like chatbots considerably cheaper to run daily after they have been produced using supercomputers.
These devices “are light years ahead of the competition,” according to Hans Mosesmann, a semiconductors analyst at Rosenblatt Securities. “Nvidia’s supremacy in AI software is not just enormous; it is also accelerating.”
Nvidia is also cooperating with AT&T Inc to improve truck dispatching efficiency, with quantum computing experts to accelerate software development, and with industry behemoth Taiwan Semiconductor Manufacturing Co to accelerate chip development. Huang continued.
Nvidia’s new renting service, DGX Cloud, may allow many more developers to access tens of thousands of its processors simultaneously. According to Nvidia, biotech business Amgen Inc (AMGN.O) and software provider ServiceNow Inc (NOW.N) have begun utilizing the service.
Nvidea to make AI model development more accessible
Nvidia also unveiled AI Foundations, a solution that assists businesses in training their own artificial intelligence models. Some big stock picture database owners want to utilize the service to avoid legal issues over the copyright of photos used to produce AI content.
Huang also disclosed technologies to accelerate semiconductor design and manufacture. The program uses Nvidia chips to accelerate a step between the software-based design of a chip and the manufacture of the lithography masks required to print that design on a piece of silicon.
Such calculations may take two weeks for a typical processing processor to complete. Still, Nvidia announced Tuesday that its chips and software could finish the operation overnight and cut the amount of power consumed from 35 megawatts to 5 megawatts.
To get it to market, Nvidia said it was collaborating with ASML Holding (ASML.AS), Synopsys Inc (SNPS.O), and TSMC. Huang noted that TSMC would begin preparing the technology for manufacturing in June.
