By Stephen Nellis
(Reuters) - Graphics chipmaker Nvidia Corp
With new software it calls RAPIDS, Nvidia is hoping to move beyond the fast-growing but relatively small market for artificial intelligence chips. It's aiming for the much larger market for chips that do basic business data crunching such as forecasting inventories. Nvidia announced the software at a conference in Germany.
Nvidia got its start in the 1990s with graphics processors to make video games look more realistic. But in recent years its chips have become used as so-called "accelerators" in data centers to speed up artificial intelligence work, like training computers to recognize images or human speech.
Nvidia's processors are not direct replacements for Intel's general purpose computing chips. But when businesses use a lot of Nvidia's chips, they often buy Intel's chips at a slower rate.
Nvidia's strong position in the artificial intelligence field came from its software efforts rather than its chips alone. About a decade ago, Nvidia created a platform that would let developers program its chips for a wide range of tasks rather than just video game graphics. Artificial intelligence researchers embraced it.
But for all the buzz artificial intelligence generates, other types of computing - like perfecting analytical models for more accurate inventory forecasts - are more common in businesses. Nvidia's new software aims to speed up that kind of work.
Nvidia has teamed up with a range of companies, including Hewlett Packard Enterprise
"It's really an ecosystem we're trying to build out," Jeff Tseng, head of product for artificial intelligence infrastructure at Nvidia, told Reuters in an interview.
But Nvidia will have a tough fight against Intel, whose chips are already the industry standard for business analytics work. The key will be for Nvidia to find early customers for whom the speed benefits outweigh the cost of new chips, said Stephen O'Grady, an industry analyst with RedMonk
"In many cases where we're talking about basic data science, it's going to be computationally faster," he said. "This is their big play to get into the general purpose data science side of things."
(Reporting by Stephen Nellis; Editing by Cynthia Osterman)