Since its initial publication, ARK has scrutinized the assumptions and refined the estimates. Even with more conservative figures, ARK finds that Amazon could offer drone delivery for $1 and still turn a profit. Here are the assumptions, broken down into three parts:
Part 1: Estimated Volume of Packages Shipped by Prime Air
Amazon AMZN is on track to reach $60 billion in revenues for 2015 in North America.1 Amazon’s average order size is roughly $60 for two items,2 implying one billion orders and two billion items delivered per year. A subset of these packages would fit the requirements for drone delivery.3
- Packages must weigh fewer than five pounds; and
- The delivery distance must be within ten miles of Amazon’s facilities.
Jeff Bezos has said that 86% of Amazon’s deliveries weigh fewer than five pounds.4ARK assumes that 25% of these deliveries will be within ten miles of an Amazon facility, and fit Amazon’s target market for Prime Air. If each item is delivered separately, roughly 400 million packages would be eligible for drone shipment.
Part 2: Assumption for Amazon’s Capital Costs
ARK estimates that Amazon will need to spend about $50 million on infrastructure for the Prime Air program. This covers the cost to retrofit existing fulfillment centers with additional space, equipment, and software.5
The cost for small drones is relatively low. The drone that Amazon shows in its demo videos is a small octocopter, which costs $1,000 – $3,000 on average.6 At an average of 30 deliveries per drone per day, 30,000- 40,000 drones would be needed at current sales levels. Amazon could stock roughly 50,000 extra batteries costing $2007 each to allow swapping8 between deliveries. In total, drones and batteries would cost roughly $80 million, bringing the total capital cost to about $130 million, relatively low for a transformation of retail logistics.
Part 3: Estimated Operating Costs of Prime Air
ARK estimates that Prime Air’s biggest operational expense will be labor. Amazon operators will have to monitor drone activity, occasionally intervening manually for remote landings for safety reasons. If they were to work in two eight-hour shifts, each responsible for 10-12 drones, roughly 6,000 operators would be needed.9 Salaries could be roughly $50,00010 a year per operator, totaling $300 million per year in annual labor cost.
To facilitate remote landings, drones would need to stream video back to Amazon’s facilities. ARK assumes that data will cost roughly $6.30 per GB per month, driving total bandwidth cost to roughly $25 million per year.11
Maintenance and fuel make up the additional operating cost. Maintenance on infrastructure could be roughly 10% of the capital cost, and maintenance on drones close to 20%12 of the purchase cost. Total maintenance costs therefore could approach $15 million per year. In addition, the retail price of electricity should be roughly 10 cents per kWh, and if every delivery were to deplete a full battery, fuel would cost roughly $4 million per year. In total, operating costs could amount to almost $350 million per year.
With about $130 million in capital costs and $350 million in operating costs, Amazon’s cost per package for delivery would be roughly 88 cents.13 At a 15% discount rate, Amazon could deliver 400 million packages for $1 profitably.14 That said, Amazon could choose to deliver the service free of charge for Prime members.
If ARK’s assumptions are off base, most variables will not change Amazon’s cost per package meaningfully, with the exception of labor cost (as shown below). If the number of drones each operator could control were to halve or if drone operators’ salaries were much higher than anticipated, Amazon’s costs could exceed $1 per package.
- North American sales are roughly 60% of Amazon’s revenue. Analyst estimates are predicting over $100 billion in revenue for 2015. Source: Yahoo Finance. ↩
- Source: https://blog.compete.com/2013/04/17/attention-retailers-amazon-prime-members-are-located-in-aisle-5/. Amazon Prime sales are 12-20% of total sales, so assume price per order is closest to $60. ↩
- Source: http://www.theverge.com/2013/12/2/5166948/here-are-the-things-amazon-needs-to-get-its-delivery-drones-off-the-ground ↩
- Source same as footnote 3. ↩
- Amazon has 75+ fulfillment centers in the US. Many of these are in the same town or area, so assume 50 of this will be retrofitted to run Prime Air. Source for number of fulfillment centers: http://www.businessinsider.com/how-many-fulfillment-centers-does-amazon-have-in-the-us-2015-3 ↩
- The octocopter shown in Amazon’s Prime Air video looks similar to a DJI spreading wings S1000+, which costs $1900 retail. Assume with electronics and software the cost to Amazon would be around $2000 each. ↩
- The DJI Phantom 2 battery is about $130. ARK assumed that Amazon’s drones would need twice the range of these machines and there is a 20% markup included in the $130 price. (130 x 2 x (1-20%) ~ $200) ↩
- Swapping depleted batteries in drones for fully charged batteries. ↩
- Please note that ARK’s assumption for drones per operator is very conservative. Some experts have estimated that 20 to 30 drones per operator would be manageable. ↩
- Job postings on Indeed for unmanned aerial vehicle operators range from $30,000 to over $100,000. ARK thinks it could be closer to the lower end at $40,000, or $50,000 fully loaded with benefits. The FAA no longer requires drone operators to own a pilot’s license, but they must obtain a certificate. The FAA plans to draw up a process for that certificate. ↩
- A Verizon plan for 100 GB of data per month costs $790 (link below). Assuming a 20% wholesale discount, data would cost about $6.30 per GB. ARK assumes landings take 2.5 minutes, so with over 400 million landings per year this would be almost 90 million minutes of streaming per month. ARK assumes that 0.5 Mbps is needed to stream the video. This would mean over 300 terabytes of data per month, which would cost close to $25 million per year.
- ARK came to this assumption after researching crash rates for drones. While many crashes are caused by human error and autonomous software should improve this rate, assuming each crash requires repair & maintenance this could amount to almost 20% of the purchase price of the drone per year. ↩
- This has been updated from ARK’s 2014 estimate due to our revised, more conservative assumptions. ↩
- This would yield an IRR of over 50% and a payback period of less than three years. ↩
ARK's statements are not an endorsement of any company or a recommendation to buy, sell or hold any security. For a list of all purchases and sales made by ARK for client accounts during the past year that could be considered by the SEC as recommendations, click here. It should not be assumed that recommendations made in the future will be profitable or will equal the performance of the securities in this list. For full disclosures, click here.