Nvidia Walks a Tightrope Between GeForce, AI

TAIPEI—A little more than ii years ago, Nvidia CEO Jensen Huang described his visitor as "all-in" on artificial intelligence and automobile learning. Possibly nowhere exercise his words ring truer than at the Computex trade show this week, where he unveiled a tiny new supercomputer to infuse robots with AI just a few minutes before disappointing legions of PC gamers by announcing that new GeForce graphics cards, the original core of the company'southward business, will non come up to marketplace for "a long time."

Nvidia is billing the new device, nicknamed Jetsen Xavier, equally the first calculator designed specifically for advancing car learning in robots. Packing more than 9 million transistors and capable of processing more than thirty trillion operations per 2nd, it took years to develop, Huang said.

"This little computer is going to be the brains of future of robots," Huang said, before list pretty much everything that has enamored lodge most robots, from cocky-driving cars to Atomic number 26 Homo'due south personal assistant ("Robots that fly, that bulldoze, that swim, that get underground, that option strawberries, aid you in the lab like Jarvis").

Huang's praise of robots was most equal to his reverence for people who play reckoner games, and he expressed regret that cryptocurrency mining has driven upward the cost of GPUs, making them unaffordable for many enthusiasts who like to build their own high-terminate gaming rigs.

"Mining is a bonus to our business just it'south non our business organisation," Huang said. "I would rather that mining demand subsides a chip and allows the GeForce cards to be on the market."

Is AI More than Important Than Gaming?

But actions speak louder than words, and the introduction of Jetsen Xavier is proof that AI is more than of import to Nvidia than consumer GPUs, at least right at present. If Nvidia wants GeForce cards to be on the market—many are currently merely available from hawkers at greatly elevated prices—why not temporarily put the brakes on AI piece of work and advance the development of next-generation GeForce GPUs?

Put another way, Nvidia is seeing record demand for and immense profits from its products, whether they're destined for machine learning, crypto mining, or powering gaming desktops and laptops. And so what's property the company up from releasing more of them?

Part of the reason may be that Huang's affection for Nvidia's traditional client base doesn't consume his business organization strategy. Gaming innovation is alive and well at Nvidia, equally evidenced past the big-format G-Sync gaming displays that showed up at CES this year and are on track to begin sales by the end of the twelvemonth. That ought to satisfy many fans who live on the haemorrhage edge, and who certainly aren't hurting for graphics functioning anyway if they own a GeForce GTX 1080.

Another office of the reason is lack of competition. For more than than a decade, Nvidia was the only game in boondocks when information technology came to serious GPU performance. Recently, AMD has emerged equally a worthy challenger with its Vega fries, but those rival the best of what GeForce has to offer, rather than surpassing it by leaps and bounds.

Just maybe the best explanation for Huang's exact words ("It's a long time from now," he said when pressed on the time frame for introducing new consumer GPUs) is that they're ambiguous. He could be setting anybody's expectations low to dramatize a potential stealthy launch event soon, something Nvidia has done in the past.

Whatever the case with timing, two things are clear: the current Pascal-based GeForce cards are two years sometime now and they are definitely going to be replaced, and impatient gamers have to reconcile their expectations with Nvidia's devotion to AI and the distracting cryptocurrency mining sideshow. Neither factors were nowadays in Nvidia's decision making just a few years agone.

And in a compression, if you lot simply must drop a wad of greenbacks on graphics computing right now, you could always spend $i,299 on a Jetsen Xavier development kit.