Part of The Mobility (R)Evolution series
By Jim Witham, CEO at GaN Systems, and Uwe Higgen, Managing Partner at BMW i Ventures
This is the fourth and final article within the series. Click here to read part 1, “The Transformative Forces of Electrification and Autonomous Driving”.
As autonomous vehicles (AVs) become generators and users of unprecedented quantities of data, not only must their onboard technology and relationship with data change, the mobility industry’s relationship with data centers, the cloud, and the high-speed networks that connect them – must also drastically evolve.
Data Volume, Data Center Growth, and Hardware Energy Efficiencies
Using the benchmark of 1.5-2 petabytes of data per year generated by each AV, the impact on data center needs can be estimated – and they will be significant. Looking at US vehicles only, if 10% of current vehicles became AVs, that would result in the creation of 40-50 zetabytes of data each year, requiring 400-500 million servers distributed across 4-5,000 massive data centers.
Running those data centers requires significant amounts of energy. If that is not pursued with an emphasis on efficiency and clean renewable sources, data centers will end up replacing cars as the mobility industry’s prime contributor to the global CO2 emissions problem. The management of a data center’s energy use is not only an ethical concern around sustainability, but also around business economics. Forty percent of a Tier 1 data center’s annual operating expense is around energy, so energy efficiency must be an area of operational focus for the new AV data centers. Operators will need to look to data center energy pioneers like Google and Apple for evolving best practices in using renewables, as well as investigate new hardware solutions around energy efficiency.
GaN power semiconductors can play an integral role in reducing the carbon footprint and operating costs of future AV datacenters by making the hardware in their power systems more compact and efficient, with the current data centers more dense – thereby delaying build-outs of new facilities.
Two Types of Data Centers in the World of Autonomous Vehicles
With the mass-adoption of AVs, there will be not one, but two kinds of data center-cloud-vehicle connections.
1. Lower volume edge data centers (‘micro-data centers’) will run the analytics on smaller data sets offloaded from vehicles such as mapping and road conditions. Similar to the Apple iPhone-iCloud relationship, this cloud will be ‘owned’ by the auto OEM who will also be responsible for the data protection and cyber-security of the individual vehicle owner’s data. That data itself will be owned by the driver and should only be shared with their consent.
2. Massive data centers that are globally distributed will perform the machine learning algorithms to reveal deep analytics on datasets forwarded from the edge data centers. Low latency and high performance computing will be key as they ingest and process the data from many simultaneous sources to feed the algorithms necessary for AV functionality. While there may be dozens of AV OEMs, there will only be a few ‘owners’ of these kinds of AV data centers. Consortiums and partnership will form to create these new platforms that must aggregate enough vehicles on the road to feed their algorithms
The Growing Importance of Energy Efficiency and Data Intelligence in Vehicle Manufacturing
In manufacturing and scaling both EVs and future AV production, OEMs are already abandoning some of the old methods of production. The significant expertise and intellectual capital that has been built over decades in building internal combustion engines has become less and less of a competitive edge.
Robotics in manufacturing, AI-assisted design, 3D printing of components, and new materials such as carbon fiber are the elements enabling the conception and building of EVs and AVs with the shorter time to market that consumers are demanding. In an industry where 7 years is a normal concept-to-market cycle, and that is developing dramatically more complex vehicles that still have to perform in any geography, speed coupled with high product reliability is a daunting challenge.
Energy and data intelligence are important factors in this manufacturing evolution. To be most effective, autonomous robots will need to run longer without recharging by having highly energy efficient motors coupled with high-density batteries. And as in the case of AVs, if they are to be truly intelligent and autonomous, when they do need to recharge, this should be autonomous. Semiconductors that take advantage of the unique size and power efficiency capabilities of GaN will be used in factory robots to deliver the performance that modern vehicle manufacturing environment demands.
New Technologies, Business Decisions, and Social Values Must All Inform the Near Future Evolution of Mobility
New technologies around electrification and autonomous driving are enabling the auto industry to rapidly evolve into the promise of something better, even ‘utopian,’ in the form of an integrated sustainable mobility ecosystem.
While the industry upside is great, the global downside is equally vast – if these technologies and the accompanying new businesses decision criteria, are not managed well. EVs and AVs must be seen as a part of a dynamically and evolving system that requires not only technologies, business acumen, and relationships that have never existed before, but also new and equal attention and action on delivering the previously unaddressed human right of individual mobility and the broader social values of global sustainability.
This article concludes the Mobility (R)Evolution article series.
For access to the three prior articles in this series: