Item 1 of this illustration, AI (artificial intelligence) character and robotic hand miniature, taken on June 23, 2023. REUTERS/Dado Ruvic/Illustration/File Photo
(1/3) AI (artificial intelligence) letters and a miniature robotic hand are depicted in this illustration taken on June 23, 2023. REUTERS/Dado Ruvic/Illustration/File Photo License rights purchased, opens in new tab Aug. 28 (Reuters) – As generative AI takes center stage for businesses and governments, countries are turning to Nvidia's (NVDA.O) chips to build artificial intelligence models in their own languages, spurring already surging demand, a senior Nvidia executive said on Wednesday. Nvidia's third-quarter outlook for increased sales of chips that underpin AI technologies such as OpenAI's ChatGPT failed to meet investors' high expectations. But the company said it was seeing new customers from around the world, including governments looking for their own AI models and the hardware to support them.
Chief Financial Officer Colette Kress said on an analyst call after Nvidia's earnings report that each country's adoption of its own AI applications and models will contribute several billion dollars to Nvidia's revenue in the fiscal year ending January 2025.
That's up from a previous forecast that such sales would contribute billions of single-digit dollars to total revenue. Nvidia now expects total revenue of about $32.5 billion for the third quarter that ends in October.
“Countries around the world want to have their own generative AI that can incorporate their own language, culture and data,” Kress said, describing AI expertise and infrastructure as a “national imperative.”
She gave the example of Japan's National Institute of Advanced Industrial Science and Technology, which is building an AI supercomputer powered by thousands of Nvidia H200 graphics processors.
Governments are also beginning to turn to AI as a way to strengthen national security.
“AI models are trained on data, but for political entities, especially nation states, that data is confidential and the models need to be customized to their unique political, economic, cultural and scientific needs,” said Shane Lau, computing semiconductor analyst at IDC.
“So you'll need your own AI models and custom underlying hardware and software configurations.”
Washington is tightening export controls on cutting-edge chips to China in 2023, blocking Nvidia's sales in the region, as it seeks to thwart AI breakthroughs that could aid the Chinese military.
Companies are working to leverage the government's initiative to build AI platforms in regional languages.
IBM (IBM.N) said in May that Saudi Arabia's Data and Artificial Intelligence Authority would use its AI platform Watsonx to train its Arabic model, “ALLaM.”Bob O'Donnell, principal analyst at TECHnalysis Research, said the growth opportunity for Nvidia's GPUs could be driven by countries wanting to create their own AI models, in addition to heavy investment in the company's hardware by major cloud providers such as Microsoft (MSFT.O).
Sign up here.
Reporting by Arshiya Bajwa in Bengaluru; Editing by Peter Henderson and Jamie Freed
Our standards: Thomson Reuters Trust Principles. Opens in new tab
Purchasing License Rights
Source link
