Supercomputers on wheels. That’s what the world’s leading semiconductor manufacturers are essentially working on. As they continue to invest heavily into the auto sector, one could say they are rocking its very core. The gameplay is changing; and so too is the face of a once traditional car industry.
What is the “next big thing” in the computer world? Cyber glasses, artificial intelligence and digital assistants – or the good old car? Judging by the recent actions and acquisitions of the big tech companies, you would be right in thinking ‘all of the above’. Indeed, semiconductor manufacturers are increasingly shifting their focus onto the transport industry. This shouldn’t come as a shock given that Intel – the world’s largest chip manufacturer – recently estimated the market for systems, data management and all autonomous driving-related services to be worth 70 billion U.S. dollars in 2030. It’s a market there for the taking: and that is why electronics specialists with automotive expertise are becoming more and more attractive targets for collaborations.
TWO HEADS ARE BETTER THAN ONE: THE ACQUISITION ERA
In the past year or so, the landscape of automotive electronics suppliers has changed substantially – thanks to a string of mergers, acquisitions and joint ventures. For starters, the IT giant Qualcomm took over NXP – the largest chipmaker in the car business – in a deal worth 47 billion U.S. dollars. Samsung is currently in the midst of taking over Harman – one of the leading specialists for infotainment and networked driving – for eight billion U.S. dollars. Continental, who had previously acquired the Elektrobit Automotive Group in 2015, founded a joint venture with the mighty China Unicom to develop “new mobility services”. Intel last year announced cooperation with both BMW and Israeli visual technology specialists Mobileye. The latter has now been integrated into Intel’s new business area devoted to autonomous driving – in what was an important strategic move. With the leading experts on image comprehension now on board, Intel offers more than just software expertise.
Indeed, time is of the essence. In this league of the 800-pound gorillas, interest is quickly growing in the smart mobility of the future, with Alphabet, Apple and Microsoft having already staked their claims. And let’s not forget Nvidia – who has managed to emerge as the darling of the car industry despite no major takeover to date. At the beginning of this year, while speaking at the CES, its CEO Jen-Hsun Huang announced a ‘new era in the transport industry’. We are on the cusp of an era that will see the next generation of semiconductors shrink a data center with 30 TOPS (trillion operations per second) to the size of a fingernail, or the ‘Xavier’ Supercomputer for Artificial Intelligence run on just 30 watts of power while meeting all the safety requirements of a passenger car. It is no wonder then, that all major automotive suppliers are following Audi’s lead and queuing up to forge a partnership so that they too can get a piece of the action.
“It is normal in the auto industry for everyone to work on their own platform. But the effort required to program such platforms and cloud environments is incredibly large as well as expensive,” says Douglas Davis, Head of the Autonomous Driving division at Intel. In his 32 years at the company, Davis has made huge contributions to advancing the standardization process for computer chips. He has also witnessed firsthand that in this elephant race, only the largest and fastest will remain. Given that the development of a new generation of semiconductor costs in excess of a billion US dollars, one can understand why. In the end, therefore, the number of autonomous driving platform vendors could be whittled down to just a few. That few will likely provide the software brain, while car manufacturers produce only the hardware around it.
We caught a glimpse of the fruits of these sorts of partnerships on May 3 this year, when Intel presented the first results of its cooperation with BMW and Mobileye. The car in question was a rather inconspicuous BMW 7 Series, which is set to clock up thousands of kilometers as part of a 40-strong fleet of test vehicles. As part of the cooperation, Delphi contributes the sensor and software expertise.
However, the supplier will continue to work with other customers such as Audi, Ford or Mercedes, according to R&D head Glen De Vos: “The highly automated systems for 2019 have long been a part of our serial production and the technology partnership with Intel and Mobileye has been well-established for many years. We will start our first technology generation via field tests in Shanghai with retrofitted vehicles; the fully automated BMW iNext will then be an optimization stage in 2021.”
Supercomputers are not only required to handle the flow of data from the 26 sensors covering the car’s entire environment within the vicinity of up to 250 Meters. They are also involved in subsequent phases upstream, such as environmental understanding, roadway planning and the driving strategy. On top of that, the mountains of data on all levels must also be evaluated by self-learning neural networks.
Already today, luxury limousines use up to 100 million lines of programming code. In the future, the data sets will be so large and the requirements so complex that they will not be able to be processed using fixed systems. In fully automated systems, more than two terabytes of data are generated per hour. This is about the equivalent of 2000 computer hard disks. Such information can only be evaluated with the aid of large computers – both inside and outside the car itself. Jason Waxmann, who is responsible for data centers at Intel, expects that by 2030, 10 percent of all data centers worldwide will be used for autonomous driving alone: “For each autonomous vehicle, there are currently about one million euros in costs for the external data backend. For series production, this effort simply has to shrink to a fraction of the price,” warns Waxmann. Safe to say that chipmakers have enough to be getting on with.
- The automotive landscape is changing rapidly as traditional suppliers and OEMs look to team up with the biggest players in the semiconductor industry in an attempt to win the race to autonomous driving.
- The imminent “next generation” of semiconductors will be characterized by increased power with decreased size and energy requirements: exactly what is needed to perform autonomous driving operations.
- Given the costs of development, chipmakers will likely standardize across all platforms leading to a concentration of power. This, in turn, could mean car manufacturers only supplying the “hardware around the chips”.
- Processors will have to handle much more than just sensor data: with two terabytes of data to process per hour, autonomous driving will have to call on external data centers to handle the data backend.
What do you think? What lies in store for the automotive industry? Will carmakers essentially stay as carmakers while semiconductor manufacturers power autonomous driving?