How Amazon, Google are transforming the auto industry

Volkswagen engineers working from home were presented with a challenge during the pandemic. The complex calculations and simulations needed to test and validate their developments could not be carried out using the hardware on their personal computers.

In addition, data couldn't securely leave the controlled environment of Wolfsburg's technical development center.
With the help of a solution developed by U.S. software developer Citrix, VW engineers were able to feed and extract huge amounts of data to the automaker's servers via a fast, stable and remote connection over the Internet. According to an internal survey of 179 employees involved in the project, more than 80 percent enthusiastically embraced the process.

Cloud computing already runs everything from online shopping to streaming and gaming, but it is now expected to play a significant role for the auto industry as intelligent, automated data analytics become a central tool for managing all aspects of the business.

Automakers such as VW are employing the technology to crunch massive volumes of data without having to pay for the infrastructure needed to do the work. That is why they are turning to third-party specialists that operate data centers, often built at "hyperscale."

"We are forecasting an explosion in demand for cloud computing resources in the automotive [sector] over the next five years," said Jeff Hood, principal at Deloitte Consulting. The pace of growth will outstrip most other industries as manufacturers endow both their vehicles and their factories with Internet-connectivity, creating entirely new opportunities for the business.

In part that's because automakers have some catching up to do: "They are adopting cloud [applications] later than in some other industries, such as financial services, technology and retail," Hood said.

Manufacturers will be relying primarily on the three major global hyperscalers -- Google Cloud, Microsoft Azure and Amazon Web Services (AWS). Other players include IBM. Meanwhile, Alibaba has grabbed the early lead in the hyperscaler sector in China, where domestic competitors include Huawei, Tencent and Baidu.

Often a cluster of third-party developers, called independent software vendors (ISVs), build upon an underlying infrastructure to offer more tailored solutions for a specific process. Or, like German software giant SAP, they adopt a hybrid strategy, partnering with a company such as Microsoft Azure when it suits while bidding against a partner for the same cloud contracts other times -- an approach known as co-opetition.

"Since the pandemic there has been a big push into cloud-first and cloud-native applications," Hagen Heubach, director of SAP's automotive industry business unit, told Automotive News Europe. "We have had requirements not just from Germany but also the U.S. to deliver a flexible cloud solution very quickly -- in a matter of days -- to react to new demands from business."

Audi and Lucid Motors, for example, both rely on UK-based ZeroLight to design a rich, interactive 3D environment for configuring cars on mobile devices. Powered by Nvidia graphic processors linked together in Ireland and operated by AWS, the immersive experience provide results.  Audi has been able to sell an additional 1,200 euros in options on average since launching the solution.

"Only by using the cloud can we render the car in real time with every single option possible, which in Audi's case involves 10 to the power of 17 different permutations," ZeroLight Chief Product Officer Francois de Bodinat told ANE. "With the Lucid Air, the reservation value was 10 percent higher compared with those customers who used the more traditional picture-based configurator. For a model that can sell for 130,000 pounds, that's not insignificant."

Google Cloud is investing heavily into the business with the aim of being a partner to the automotive sector for all its needs, announcing Monday a deal to become Ford's preferred cloud provider.

"Cloud started out with software-as-a-service, programs like Gmail that are not locally installed, then it became infrastructure-as-a-service, essentially taking what you already store in your own building, putting it somewhere else and renting it back," Dominik Wee, Google's managing director for global manufacturing, industrial and transportation, told ANE. "But cloud services have reached a maturity where the capabilities go far beyond that now. We want to help them solve their biggest business issues, not just rent them a computer."

Networking an automaker's manufacturing operations is a common starting point for hyperscalers. In July, Renault hired Google Cloud to accelerate the digitalization of its industrial footprint and prepare it for the Internet of Things (IoT).

Another area where hyperscalers get involved is making sure a vehicle's architecture is digitally enabled for services such as charging, parking and maintenance. VW teamed up with Microsoft Azure in 2019 to develop an Automotive Cloud, which includes back-end servers that underpin VW brand's We in-car ecosystem.

"The investment level necessary to create a connected vehicle platform is very high and it doesn't provide you any differentiation to the customer, so it's something you want to farm out to somebody else," Microsoft Azure's IoT mobility head, Tara Prakriya, said during a presentation of the deal. "If everyone has to build it bespoke, that's a lot of investment and a lot of maintenance for the long run."

Not to be outdone, BMW partnered in December with AWS to accelerate its pace of innovation "by placing data and analytics at the center of its decision-making." In the process, the premium automaker is migrating data from more than 100 countries to server farms operated by AWS.

The aim of the project is to better forecast demand for BMW models and the equipment options that customers will want worldwide. This is supposed to help the purchasing, production and sales departments while simultaneously boosting customer satisfaction.

Although it doesn't have an estimate for automotive-related cloud spending, IT researcher Canalys estimates the overall global cloud infrastructure market will increase at about 16 percent a year on average, rising to reach roughly $300 billion by 2025.

While services are similar, there are some differences between providers.

"Whereas AWS was first to market, giving it time to build out its capabilities, Google was more of a laggard and had to focus more on specific verticals such as banking," said Matthew Ball, Canalys' chief analyst, about the overall cloud infrastructure market.

While Google Cloud is smaller than its two main rivals, Wee said the business has been growing at a faster rate -- about 50 percent -- in part because of its competitive advantage in semiconductors designed specifically for AI.

Known as tensor processing units, or TPUs, these proprietary chips, which are not for sale, can be trained with visual data to spot blemishes on a car's paint job or defects in a powertrain component using visual clues. (AWS in November began utilizing A100 chips that feature tensor cores for similar applications, but these are graphics processors sold by Nvidia.)

"The reason we have that is because we are probably one of the biggest consumers of machine learning on the planet. Every time you type something into Google, you have just used our algorithms," Wee said.

With the help of AI, Google believes it can improve an automaker's efficiency at a double-digit rate. In the case of Renault this is done by aggregating data from more than 2,500 machines deployed across 22 production sites worldwide.

Just like with Android, Google Cloud believes its commitment to support other open-source software such as Kubernetes and TensorFlow is a key selling point. This enables a customer to more easily switch providers or add new ones, reducing the risk of being locked in with any one particular vendor.

"This was a huge factor for Renault," Wee said. "One of the reasons they went with us was the security that if at any point in time they wanted to move providers, they were at liberty to do so."

AWS, however, is the sector's dominant player, controlling nearly a third of the overall market. The e-commerce giant became the world's leading hyperscaler almost by accident.

Like any retailer, Amazon was subject to the inherent cyclicality of customer demand, which rose dramatically ahead of big holidays such as Christmas.

To cope with these peaks, a massive IT infrastructure was built, but it was underutilized after the holidays. Therefore, the company began renting out its spare processing capacity for a price and in the process AWS was born.

Its most recent income statement shows the service accounted for a little more than a tenth of Amazon's overall $193 billion revenue in the first nine months of 2020. Yet it is extremely lucrative, contributing nearly two-thirds of the group's consolidated operating profit during that the period.

According to the company, its infrastructure can offer companies the ability to deliver three times as many features on average 42 percent faster than a typical customer could get if the company relied on its own legacy IT hardware.

"The biggest challenge for a business is trying to extract all the value from the various silos. Once you have moved that onto a flexible platform like the cloud, you can then start to innovate," Dean Phillips, AWS's automotive tech leader, told ANE.

AWS recently added a new service that allows customers to bid on a spot market for access to Amazon's cloud computer. While they pay below the normal AWS rate, there is one catch: It may abruptly and without warning interrupt the service should the need arise.

"We are able to scale up to a million-plus cores of computing power," Phillips said. This makes it ideal for running autonomous driving simulations, he added: "That entire power could be available to you as a burst, literally now, and when it goes away, the data is stored and you restart later where you left off."

While there is a clear trend toward greater use of cloud computing, there have been notable instances of a reverse migration. Dropbox got its start in the 2000s with help from AWS, but it chose to part ways with the Amazon subsidiary in 2016. This helped Dropbox reduce its operating expenses.

Storing data and running applications off the cloud also means a loss of control. Recently AWS shut down Parler after deciding it would no longer host its business. The damage was so dramatic that the U.S. social media app popular with conservatives has sued the hyperscaler for damages.

HERE Technologies, the digital mapping service part owned by German automakers, re-engineered its platform last year to be "cloud agnostic" -- in part out of concerns it could become too reliant. "We didn't want to be dependent on just one cloud provider," HERE Chief Technology Officer Giovanni Lanfranchi said in an interview.

This is not the only potential downside. The more information that is stored in one location, the more it compounds risks of data theft and other forms of cybercrime. Cloud infrastructures are by nature "a highly attractive target for hackers," said Klaus Ottradovetz, an IoT expert at French blue chip Atos, told ANE.

The SolarWinds attack that compromised a number of U.S. federal agencies grabbed headlines worldwide by highlighting the vulnerability of even the most closely guarded government networks. In Europe, hackers accessed documents relating to the Pfizer-BioNTech coronavirus vaccine stored at the European Medicines Agency in Amsterdam.

In September, these digital weaknesses proved fatal. A patient with a life-threatening condition was refused treatment by a university clinic in Dusseldorf, Germany, after a ransomware attack crippled the hospital's computer systems. Diverted to a hospital a half-hour away, the woman was pronounced dead upon arrival.

"This has been reported as the first cyber-related death in Europe," EU Commission Vice President Margaritis Schinas told reporters last month as he helped unveil broad new measures to combat hacking across the bloc.

In an attempt to ease customer concerns over the safety of their data, Google Cloud can process information while still encrypted with a key that stays with the customer, a new development called confidential computing.

"We have dedicated security and privacy engineers who focus just on protecting all our systems and solutions -- find me a carmaker with this kind of capability," Google Cloud's Wee said.

Notwithstanding the risks, Brussels now wants to extend the EU's single market to data, with the 27 member states announcing plans in October for a European cloud -- in part to level the playing field with the three U.S.-baed hyperscalers. By creating an even footing and harmonizing standards across borders, applications and IT systems, also known as federating, it's easier to encourage ISVs to migrate to your platform and drive innovation.

The IT industry is bullish about the possibilities as 5G telecommunication becomes standard, enabling ultrafast data transfers to the cloud. Amazon CTO Werner Vogels predicted in December that more data will be created in the next three years than over the last 30.

The bulk of this will likely be mobile devices, industrial robots and intelligent cars. According to estimates from the European Commission, these "edge" use cases will comprise 80 percent of the computing needs in 2025.

This is where latency plays a factor. Not all IoT applications are suited for the cloud -- a fully autonomous car navigating busy downtown streets, for example, would not be able to transmit information to and from the cloud for processing fast enough to respond to changing traffic conditions.

"Wireless data transfer can result in latency problems compared with fiber optic cables, where your main constraint is the speed of light," Canalys' Ball said.

After new-car sales plunged last year, manufacturers need to monetize this wealth of data in order to survive, argues consultancy McKinsey in a study published ahead of last month's Consumer Electronics Show in Las Vegas.

"Seven of the 10 most valuable companies in the world built their success on data-based business models," wrote co-author and McKinsey partner Timo Moeller.

According to the McKinsey study, 37 percent of potential car buyers surveyed said they would consider defecting to a rival brand's model if they found it had better connectivity features. That number rose to nearly half for premium brand customers.

Legacy automakers are responding by elevating the importance in their organizations of mining cloud-enabled vehicle data for value. In October, BMW bundled all key software functions in a new business line called Digital Car that reports directly to the management board.

The company hopes to better harness the swarm intelligence of its connected fleet to create added value for its customers. Every day roughly 14 million cars collect 25 million bits of traffic information, according to BMW.

The information is first encrypted, then relayed using a virtual private network (VPN), before it bifurcates into personal data stored safely on its own on-premises servers while anonymizing the rest to feed cloud-based services like BMW Maps.

"The cloud gives you access to pretty much unlimited storage and compute capabilities that you can scale up or down on demand," Canalys' Ball said.

The sky, in other words, is the limit.