in

How Nvidia Went From Gaming to AI to Riding With Mercedes

How Nvidia Went From Gaming to AI to Riding With Mercedes
Courtesy Mercedes-Benz

To investors who have followed Nvidia’s transformation over the past five years, this makes perfect sense. It helps explain why the stock has multiplied 18 times in price over that period.

in 1993 to make a new type of computer chip ideal for powering three-dimensional videogames—a few years before games needed such a thing. Fancier games soon followed. This fiscal year, which ends in January, gaming will be worth an estimated $6.1 billion in revenue for Nvidia. But there’s much more to the company than that.

“People thought we were a videogame company,” says Huang. “But we’re an accelerated computing company where videogames were our first killer app.”

To see the second killer app, just look at Wall Street forecasts for this year. Revenue in a category Nvidia calls Data Center is expected to more than double, to $6.5 billion. That will mean gaming is no longer the company’s biggest moneymaker.

Nvidia owes its success in data centers to the rising use of artificial intelligence, because videogame chips just happen to be well suited to that type of computing. There’s a perfectly good explanation for why this is the case—I know this, because sometimes people try to walk me through it, while I stare back at them like a dog watching Cirque du Soleil.

The closest I came to understanding it was this past week, talking with Chris Rolland, an analyst at Susquehanna Financial Group. He described how lighting up a display pixel involves three “vectors”—an x and y coordinate, plus a color. Artificial intelligence also uses three vectors, he said, in an operation called matrix multiply-accumulate. Soon after that my brain felt hot, so I started thinking about bobbing for apples, which helped.

Anyhow, Rolland says a lot of companies are trying to make chips even better suited to AI but that Nvidia has created powerful software for programming its chips, called CUDA, making its role secure for now. Nvidia leads in what’s called parallel processing, or performing many tasks at the same time, whereas Intel (INTC) rules in serial processing, or performing single tasks very quickly.

Right now, there’s massive growth in parallel processing. Jefferies analyst Mark Lipacis calls it the fourth tectonic shift in computing, after the move from mainframes to minicomputers in the 1960s, to personal computers in the ’80s, and to cellphones and data centers beginning in the late ’90s. He attributes the rise of parallel processing to cheap memory, practically unlimited data storage, and, yup, improvements in processing chips and software. In other words, Nvidia’s Huang made the product, and the market for it appeared, just like with videogames.

Artificial intelligence is increasingly useful for turning raw customer data into sales tips, or automatically telling which photos in a collection include Uncle Burt, or spotting cancer on a patient scan that a doctor might miss. It’s also what cars will one day use to drive themselves. If anyone asks how it works, tell them there’s a subset of AI called machine learning, and a subset of that called deep learning, which uses something called artificial neural networks. Then pretend you have to take an emergency call from NASA to help with a super-hard space question.

, in part because it has developed its own hardware and software. Other car makers must either spend vast sums and bone up on AI and chip-making, or else find a partner, if they don’t want to risk falling behind. Rolland says a company called Mobileye is in a good position to be one of those partners. Mobileye, however, was snapped up by Intel in 2017. Nvidia’s advantage, he says, is that its chips are already being used for the job of training autonomous-driving algorithms.

Huang at Nvidia calls the Mercedes deal a transformative moment for the company, because it has gone from hardware for videogames, to hardware and software for AI, to now hardware, software, and services for cars. He says that makes Nvidia a platform company, not just a chip maker. “The first vertical market that we chose is autonomous vehicles because the scale is so great,” he says. “And the life of the car is so long that if you offer new capabilities to each new owner, the economics could be quite wonderful.”

Nvidia’s earnings per share are expected to jump 40% to $8.15 this fiscal year. Shares recently traded at $370, or 45 times earnings. Data centers could be a $10 billion business for the company within three years. Cars are still less than $1 billion.

How long until cars truly drive themselves? I would have guessed more than a decade, but Rolland at Susquehanna says it could be 2025.

Huang says, “We’re going to have humans in the loop for a long time.” I asked him if it’s true that cars will have to be programmed to prioritize whom to hit and to avoid if it comes to that in an accident. A top chip chief told me that over dinner several years ago, and I’ve been trying to figure out ever since how to get on the “avoid” list.

Huang says it isn’t true. “If a car gets in front of you, stop,” he says. “Otherwise just stay in your lane. Simple.”

Write to Jack Hough at [email protected]Follow him on Twitter and subscribe to his Barron’s Streetwise podcast.

Source: www.barrons.com

Report

What do you think?

486 Points
Upvote Downvote

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

GIPHY App Key not set. Please check settings

Here's how this tech firm uses AI to speed up traffic flow

Here’s how this tech firm uses AI to speed up traffic flow

Lyft resumes autonomous vehicle testing on public roads

Lyft resumes autonomous vehicle testing on public roads