NVIDIA's Jen-Hsun Huang Hand Delivers First DGX-1 Deep Learning Server To Elon Musk's OpenAI

NVIDIA has just announced that its first-ever DGX-1 deep-learning server has found a home, and it couldn't be more appropriate. That new home is with OpenAI, the world's largest non-profit artificial intelligence research agency, which is based in San Francisco.

If the OpenAI name sounds familiar but you can't quite place why, it's probably because we've talked about it a couple of times before. With the help of fellow industry legends, Tesla's Elon Musk helped launch OpenAI last year with a major goal: to progress AI with the mindset that it will benefit all humankind.

The integration of the DGX-1 into its first lab was special enough for NVIDIA CEO Jen-Hsun Huang to show up and hand-deliver it -- and, it was for a good reason. "I thought it was incredibly appropriate that the world's first supercomputer dedicated to artificial intelligence would go to the laboratory that was dedicated to open artificial intelligence," said Huang.

OpenAI isn't just a name; it's a statement that its research will be public knowledge. When Elon Musk first helped launch the research company he insinuated that it was research for good, not evil. It's clear that AI can be used for very awful purposes, but those are not the purpose of OpenAI.

While NVIDIA's latest top-end graphics cards, such as the TITAN X, push just over 10 TFLOPs, the DGX-1 pushes a total of 170 TFLOPs. That makes this a "supercomputer in a box", according to the company. With that power, research can be heavily accelerated. Perhaps in time, OpenAI will integrate a second if it's deemed necessary. The potential for data centers to integrate multiple units also seems likely. We could definitely picture one or more DGX-1s at Oak Ridge National Laboratory, which has routinely adopted advanced NVIDIA hardware.

If you're a regular user who would like to score a DGX-1, you might want to put your Tesla Model S 70D preorder on hold: one DGX-1 costs about $130,000 USD. If that seems too pricey: consider the fact that NVIDIA spent a staggering $2 billion over three years crafting the product. Maybe that makes $130,000 a steal?