Home Connected Car Self-Driving Intel is too late for the first wave of autonomous vehicles

Intel is too late for the first wave of autonomous vehicles

9 min read
View original post.

Following the completion of Intel’s (INTC) acquisition of Mobileye (NYSE:MBLY), Intel announced that it intends to build a fleet of 100 autonomous vehicles (SAE level 4). However, even this modest sized fleet will not be finished this year, and Intel has only claimed that the “first vehicles will be deployed this year”.

More to the point, Intel has made absolutely no claims regarding the capabilities of its autonomous vehicles. These are still test vehicles that will require human drivers and, by virtue of relatively immature software, will probably require frequent human intervention. As of today, not even Alphabet’s (NASDAQ:GOOG) (NASDAQ:GOOGL) Waymo has an autonomous vehicle system good enough to forego human drivers.

What will Intel build into these test vehicles? We already have a pretty good idea from a recent Intel blog. The blog asserts a lot of self-evident things as shown in the graphic above. Intel’s point that its processors “power hundreds of level 4 and level 5 test vehicles” is something we could have figured. That’s because the first option that experimenters in the field have gone to is the PC in the trunk.

Intel seems to want to perpetuate the PC in the trunk approach. Its preferred solution is a hodgepodge of processors from Atom to Xeon with an FPGA thrown in for good measure. Intel argues that since nobody knows what the final solution will be, autonomous vehicle developers need a little of everything.

It’s not inconceivable that Intel is right, but the PC in the trunk is not a solution that automakers will prefer in any case. This solution has a number of problems. It’s too expensive, too bulky, and too energy consuming for the impending electric vehicle era.

Intel plays catch up

I certainly agree that Intel should try to compete in this space, but I think Intel investors need to be realistic about Intel’s prospects. Intel is competing with other companies that have had their own test vehicles on the road for a year or more. Waymo has been working on self-driving cars since 2009.

Audi’s (OTCPK:AUDVF) 2018 flagship A8 sedan will feature semi-autonomous driving capability (up to about SAE 3) using the Audi zFAS system that has been in development since 2015. Audi’s zFAS processor was a joint project of Audi, Mobileye, and Nvidia (NVDA).

Nvidia first unveiled the Drive PX 2 autonomous vehicle platform at CES in January 2016. At CES this year, Nvidia demonstrated its own autonomous test vehicle, the BB8, which was in development for 18 months, according to Nvidia.

Shortly after Drive PX 2 was unveiled, Cruise Automation began fielding its own autonomous vehicles and was soon acquired by General Motors (GM). Cruise has been building GM’s own fleet of autonomous Bolts, which have put on some very convincing demonstrations around San Francisco.

Tesla (TSLA) currently installs its Enhanced Autopilot hardware on every Model S and X, which has a processor based on Drive PX 2. To date, Nvidia has announced numerous partnerships that will make use of its future Drive PX 3 based on the Xavier system on chip (SOC). These partnerships include Daimler, Audi, Volkswagen, Volvo, and parts suppliers such as Bosch and ZF.

The Xavier based system is far more compact than a PC and only consumes about 30 Watts.

Source: Nvidia

Nvidia is part of the Rethink Technology portfolio and is a recommended buy.

Whatever Intel has to offer this year will not be mature enough for the automakers

The significance of the partnership announcements is that the automakers are currently lining up the hardware that they will use for autonomous vehicles to arrive some time around 2020. Because of the long development times for new auto models, key design decisions are made several years in advance. The automakers have to base their design decisions on what’s available several years before production begins, not on what might be available when production begins.

Automakers are unlikely to commit to anything that Intel might field this year, simply because it isn’t production ready. That’s assuming that Intel even gets something on the road that is fairly close to the performance of the more experienced players.

That’s a big if. Neither Intel nor Mobileye have much experience in the deep learning algorithms that are currently considered the best approach to autonomous vehicles. This perhaps explains Intel’s attempt to de-emphasize the importance of deep learning. Intel lags in the development of deep learning software and the hardware platforms to support it. That also makes Intel a non-starter for the first wave of autonomous vehicles to hit circa 2020.

And I think Intel realizes this, even if some enthusiastic supporters do not. There’s still a market opportunity that exists after the first wave of self-driving cars hits the road. Intel is aiming for that opportunity, thinking and acting long term.

Comments are closed.

Check Also

The near future of urban mobility

View original post.The world of personal mobility, long dominated by the private car, is c…