I saw this posted on Anandtech forums:

http://blogs.barrons.com/techtraderd...rs-the-future/

Jen-Hsun Huang, CEO of chip maker Nvidia (NVDA), is a veteran of the Consumer Electronics Show, having come to Las Vegas for years and years now, since back when it was the Comdex show.

At the company’s booth on the floor of the Las Vegas Convention Center on Wednesday, he was kind enough to take a few minutes talking with me about Nvidia’s expanding role in automobiles and other aspects of the future.

Huang, who has six cars, including a Ferrari and two Tesla (TSLA) “Model S” (he auctioned off his original Tesla “Roadster,” raising $600,000 for charity), believes autos will be the most significant market for computing in the coming years.

“Think about it: it gives us freedom,” he says, of the fundamental act of getting in and driving.

Someday, he avers, “all cars will be software-defined,” meaning, fundamental aspects of the car can be modified via software updates, much the way aspects of the Tesla have been repeatedly enhanced though over-the-air updates downloaded to the car.

Observes Huang, “The day I bought a gasoline-powered car was the day my relationship ended with that car company.” Old Detroit, he points out, wanted to send you out the door with a widget and never hear from you again. New car makers, such as Tesla, understand that “the sale of the car is just the beginning of the relationship,” says Huang.

Huang unveiled a new chip, “Tegra X1,” to power new model cars this year, during the company’s CES press conference Sunday evening.

In future, X1 parts will contribute to a kind of supercomputer on wheels that he believes will make possible autonomous vehicles.

When? Within two to three years, cars will be on the market that will park themselves just like the demo shown on Sunday. In five to ten years, he believes, cars will achieve more autonomous driving functions.

As for dense urban environments, that’s a little more challenging.

“Boy, New York is tough,” he says of the overall density of car traffic in the Big Apple. Huang expresses amazement and admiration for New York City cabbies, especially the ones who accelerate to 90 mph in a given city block. “They’re just amazing,” he says of New York cabbies.

Will Nvidia, which has had numerous successes in car semis, get a higher share of the bill of materials of cars if its plans for X1 play out as expected? “Yes, I think we’ll get a higher share if we do this right,” says Huang.

I ask if he’s looking for new challenges, new mountains to climb after having figured out the supercomputing needs of cars. “We’ve got to build a lot of cars!” he replies with a smile, indicating there’s still a lot of work to be done.


One word that might describe Nvidia is adroit: Huang has shifted the company out of markets when it appeared the company’s prospects there might not be as high as hoped.

For example, I asked Huang about smartphones, where Nvidia used to be a contender. Last spring, Huang indicated to the Street during the company’s analyst day meeting that Nvidia was no longer a contender in mobile devices such as tablets and smartphones. “We’re still open to things, if customers come to us,” he said, pointing to Nvidia chips running Google’s (GOOGL) “Nexus 9“ smartphone. But the company is not actively pursuing the phone. Huang’s general view, repeated several times, is that it is a slowing market that is not nearly as exciting as autos.

I also asked about servers. Nvidia, I said, had in past talked up prospects as a vendor of CPUs for servers based on ARM Holdings (ARMH) server chips. That was my impression, at least, in the years leading up to the release of its “Denver” project, which turned into the Tegra “K1” processor last year.

As Qualcomm (QCOM) enters the server market, Nvidia seems to have dropped any ambition to be a vendor of general-purpose server CPUs.

I asked if Huang had changed his thinking on the server market.

“I may have changed my thinking,” said Huang. “I think Intel (INTC) has done a very good job in improving the power-performance of Xeon,” Intel’s main server-chip line.

“So, the question is, Why do you need an ARM-based chip?” to run a server. “Why do you need it? The reason used to be that you needed to improve the power efficiency of the server. And there will be some, there will be some business for it [ARM-based chips], maybe Cavium (CAVM) and others. But I think Intel has made a lot of progress.”

On the other hand, Huang is excited about what his company’s “Tesla” line of GPUs can do inside of whatever servers are running in the data center. Tesla provides the ability to sift and sort large data sets for tasks such as Big Data and analytics. “We’ll work with whatever ISA [instruction set architecture] is out there, be it ARM or x86,” said Huang. “We’ll work with whatever people want to use. It might even be POWER [IBM’s (IBM) chip flavor]. We’ll be open.”

Nvidia shares today are up 73 cents, or almost 4%, at $19.86.
Quite an honest interview!!