Nvidia Enters Supercomputer Market With New Chips, Systems

Nvidia Corp., the biggest maker of chips powering graphics cards, entered the supercomputer market with a new product that speeds up artificial intelligence software and applications.

The $129,000 computer, called the DGX-1, was built using the company’s Tesla P100 processor and will be sold to operators of data centers and companies that want to increase their machine-learning capabilities, the company said.

“This is one beast of a machine,” Nvidia Chief Executive Officer Jen-Hsun Huang said Tuesday at a company event. “It’s like having a data center in a box.”

He said the new products represent a commitment to go "all in" on machine learning, an increasingly popular area of AI in which computers, and the software that controls them, can learn and improve without help from human programmers.

Under Huang, Nvidia has escaped the worst of a four-year slump in the market for personal computers by persuading video-game players to upgrade to faster machines and finding new markets for its chips in data centers, research labs and cars.

Huang said the new P100, a board containing various chips, is the result of as much as $3 billion of research and design spending over about three years. The graphics processing unit, or GPU, at the center of it is the biggest processor ever designed or made, containing 15 billion transistors, Huang said.

The P100 is in volume production and will be integrated into data centers and used as the center of new computers made by customers including International Business Machines Corp., Hewlett Packard Enterprise and Dell Inc., he said.

Before it's here, it's on the Bloomberg Terminal.