So, what exactly is Amazon’s playbook on autonomous vehicles? There is little doubt that Amazon will be a major user of autonomous vehicles and will participate in multiple AV use-cases. In this column, I will focus on Amazon’s AV-related activities. This is a follow-on to my previous Amazon column, which provided information on Amazon’s extensive and growing logistics and transportation activities.
One of Amazon’s four guiding principles is “commitment to operational excellence” and AV deployment in multiple use-cases will answer those goals with better safety, lower costs, better customer service and other benefits.
So, what does Amazon’s playbook say, when it comes to go after the AV market?
AVs will also create new business opportunities through mobility-as-a-service (MaaS) especially robotaxi and goods deliveries. There is no doubt that Amazon will use goods AVs as part of its transportation and delivery network—the question is when and how fast. There is more uncertainty in Amazon’s goal with the robotaxi opportunity that are present from the Zoox acquisition.
The next table is my summary and perspectives on what Amazon can and may do in the autonomous vehicle use-case segments. I used data I could find on Amazon’s website and other sources. I also added my own opinions, estimates and guesses. Note that the deployment guesses are entry time and volume deployment will take a decade or so for major impact.
Amazon acquired Zoox in June 2020 for about $1.2 billion. In December 2020, Zoox introduced its special-purpose battery-electric vehicle designed for robotaxi applications that is a very impressive product. I wrote a column on the introduction: Gadzooks! A Worthy Robo-Taxi from Zoox | EE Times. In a recent interview, the Zoox CEO said its robotaxi will not be deployed in 2022. Testing is continuing in San Francisco and Las Vegas and in private testing grounds.
Sidewalk goods AVs
Amazon is developing its own sidewalk goods AV, called Scout, and has its own hardware and software labs in Seattle. Starting in January 2019, six Scouts began delivering packages to customers in a neighborhood in Snohomish County, Washington. It is close to Amazon’s headquarters.
From August 2019, Scout started to deliver packages to customers in Irvine, California. A small number of Scout AVs started delivering packages Monday through Friday during daylight hours. For safety, the sidewalk AVs initially were accompanied by an Amazon Scout Ambassador.
In July 2020, Amazon expanded Scout delivery testing to Atlanta, Georgia and Franklin, Tennessee.
Amazon’s expansion has largely targeted residential communities, with Atlanta as the exception. Dense cities, such as New York and San Francisco, already have crowded urban sidewalks and are skeptical about adding more sidewalk traffic. Amazon emphasize that each Scout can navigate around pets, pedestrians, and other objects and moves at pedestrian speed.
Road goods AVs
The last mile goods delivery segment also has road-legal AVs in a variety of sizes—from vans and small trucks to special purpose goods AV such as Nuro R2. The vans and small trucks have flexibility and can be used for middle-mile goods transport and can have safety drivers for backup during testing stage.
There is minimal information on what Amazon is doing or planning in this segment. The Zoox AV-BEV platform could probably be leveraged for an Amazon special purpose goods AV and may be the best choice.
Cooperation with Gatik is another option since it is being used for last mile and middle-mile goods transport. Gatik is currently working with Walmart and other retailers. Gatik’s VC funding is modest at around $30M, which makes Gatik a potential acquisition candidate for Amazon or someone else.
Nuro is the leader in special purpose goods AVs that only carry goods with no human onboard. This simplifies safety rules and avoids much costly onboard safety equipment. Nuro was the pioneer that got DoT permission to use up to 5,000 R2 goods AVs without human safety equipment (sounds obvious, but regulators move slowly).
Amazon could become a future Nuro customer, but Nuro is probably too expensive for an acquisition since it has received around $1.5 billion in VC funding—most of it from SoftBank.
Predicting when Amazon will enter this AV use-case has a lot of uncertainty and depends on its strategy. My best guess is that this will happen in between 2023 and 2025.
The most likely autonomous truck partner for Amazon may be Aurora due its VC investment in Aurora. Aurora has been cranking up its testing in autonomous trucks in the last year. In January 2021 Aurora announced an autonomous truck partnership with Paccar. Paccar owns two leading truck brands—Peterbilt and Kenworth. Amazon is currently using Kenworth for its own truck fleet. From its press release, Aurora said it will work with Paccar to create an “expansive” plan for future autonomous trucks. Aurora and Paccar plan to work closely on “all aspects of collaboration,” from component sourcing and vehicle technology enhancements to the integration of the Peterbilt and Kenworth vehicles with Aurora’s hardware, software, and operational services.
Amazon has other options and could work with one of autonomous truck software platform companies. I wrote a column on these platforms in early December 2020: Autonomous Truck Software Platform Sweepstakes | EE Times. TuSimple and Plus are among the leaders in truck autonomy and both are working with multiple truck OEMs.
An Amazon acquisition is probably a long shot. Amazon has worked with autonomous truck startup Embark in testing goods delivery on the I-10 interstate highway. The Swedish autonomous truck startup Einride may be interesting to Amazon since it has developed trucks with battery-only powertrain.
Amazon could also use whatever autonomous software platform its OEM truck supplier use. The potential list is quite long and include Paccar, VW (multiple truck brands), Volvo, Daimler and others.
Forecasting when Amazon will enter the autonomous truck use-case is also difficult and depends on its strategy. My best guess is that this could happen as early as 2023 in small scale deployment, with 2024 and 2025 more likely for larger deployment.
Amazon has been a leader in using warehouse AVs since it acquired Kiva in 2012. The Kiva technology revolutionized Amazon’s fulfillment center operation. Since then Amazon developed a new generation of warehouse AVs with a name of Hercules. Amazon has an article on this development at: The story behind Amazon’s next generation robot (aboutamazon.com). This article from March 2019, says Amazon was using over 100,000 warehouse robots at that time. With Amazon’s fulfilment center expansion in the last two years, Amazon is now probably using 150,000 or more warehouse AVs.
It is clear what Amazon is doing in autonomous drones—it is developing its own devices and systems to manage the operational phase. This was covered in the previous column. I am making a guess that Amazon will move from drone delivery testing, which has started, to drone delivery deployment in 2022 in a few locations with continued expansion over many years.
AV and related investments
Aurora is a leading developer of autonomous vehicle software platform. The Aurora Driver is focused on multiple AV use-cases including autonomous trucks, robotaxis and goods delivery. Aurora has received VC funding of $1.1 billion from 13 investors including Amazon.
Deliveroo is a British startup company focused on food delivery using human drivers. Deliveroo has received $1.7 billion in VC funding. Amazon was the lead investor among four investors in a May 2019 VC round of $575 million. Deliveroo and similar companies are expected to use goods AVs in the future.
Rivian is a successful BEV startup that has received VC funding of $8.2 billion with Amazon providing $700 million in February 2019. Amazon also ordered 100,000 Rivian vans on February 2020. Rivian started delivering its vans in late 2020 and Amazon are planning to deploy these BEVs in 16 cities in 2021 with Los Angeles already started in February 2021.
There is a lot of uncertainty in this data, but hopefully it will give you some useful perspectives. I will do a few back-of-the-envelope speculations on what this could mean for last mile AVs for Amazon. This is strictly a what if scenario using multiple estimates for 2025 and 2030.
The methodology is simple, and is only for the U.S. I use an estimate of how many 2020 packages were shipped by Amazon for last mile delivery. Amazon’s transportation network delivers a portion of these packages and the percentage is increasing each year. I then estimate Amazon’s package growth until 2025 and 2030—both total and Amazon’s portion.
The next step is to guess at what portion of the Amazon delivered packages use four different delivery methods—human drivers, sidewalk AVs, road AVs and autonomous drones. To make it simple to read, I have put the spreadsheet summary in the next table.
The core data is how many packages Amazon is shipped in 2020. From various data, I think an estimate of 5.2 billion is reasonable. Amazon delivered about 67% of these packages or 3.5 billion in 2020. All of the deliveries were by human drivers in 2020.
To get an estimate of Amazon package shipment in 2025 and 2030, I slowed the yearly growth from around 13% in 2021 to 9% in 2030. This gave an estimate for Amazon total package deliveries of 9.4 billion in 2025 and 15.1 billion in 2030. For Amazon’s share of last mile deliveries, I used 90% in 2025 and 96% in 2030. This means Amazon package deliveries is estimated at 8.4 billion in 2025 and 11.85 billion in 2030.
The next step was to make a guesstimate for the market share of sidewalk AVs, road AVs and autonomous drones. As you can see from the above table, the shares are miniscule in 2025 and quite small in 2030. The vast majority of Amazon packages will be delivered by human drivers. After all Amazon has ordered 100,000 electric vans with Amazon getting the last ones in 2030. Amazon will use these electric vans for last mile deliveries for a long time.
Even with the small share for the AV-based deliveries, the delivery numbers are quite large. In 2030 the what-if scenario shows sidewalk AV will deliver over one billion packages, road AVs at 945 million packages and drones at 727 million packages. The human delivered packages are 11.8 billion in this scenario. Please make your own estimates if you think this is too low or too high.