Two drone-delivery announcements landed 48 hours apart, and both are built around the same round number.
On August 17, Uber announced a strategic partnership with Zipline, including an equity investment of undisclosed size, targeting one million drone deliveries per day by the end of 2029. On August 19, Amazon said Prime Air would reach nearly 500 US cities and towns by the end of 2026, and a company executive has projected one million deliveries this year.
The two "one million" figures differ by a factor of 365. That is the whole article, and almost no coverage of either announcement said it, because each was written without the other in the room.
Both numbers also need their provenance stated before anything is built on them. Amazon's 500 cities and its pricing come from the company's own release. The million-deliveries figure does not: it is a projection VP David Carbon gave in an internal March meeting, reported by CNBC. Amazon's own public language is vaguer — "hundreds of thousands of packages" this year and "thousands daily". Uber's million a day is a target three and a half years out, announced jointly with a private partner, with no committed capital figure attached. Neither number is guidance and neither is in a filing.
What a million a year actually is
A million deliveries spread across 2026 is about 2,740 a day across the entire United States.
Amazon publishes what it charges: free for Prime members on orders of $50 or more, $2.99 for Prime members below $50, $4.99 for non-members. So the entire programme's direct fee revenue this year, at the projected volume, is:
| Deliveries | Fee | Fee revenue | Share of TTM revenue | |
|---|---|---|---|---|
| All billed at the Prime rate | 1,000,000 | $2.99 | $2.99M | 0.00039% |
| All billed at the non-member rate | 1,000,000 | $4.99 | $4.99M | 0.00064% |
Trailing-twelve-month revenue is $775.7B, up 15.8%, from our stored quarters through the quarter ended June 30, 2026.
And that table is the ceiling, not the estimate. Every delivery in it is assumed to be billed. In reality the largest single category — Prime members ordering $50 or more — pays nothing, so actual fee revenue is somewhere below $3M and could be close to zero. Prime Air does not appear as a line in any Amazon segment disclosure, which is the strongest available evidence for how immaterial the revenue is: a business does not get its own line until it is worth one.
Against parcels it is no better, though here the arithmetic has to stay honest about its own precision. Amazon has said for years that it delivers billions of packages a year in the US and does not publish a precise annual figure. Against anything in the 5–7 billion range, a million drone deliveries is 0.014% to 0.020% of what the company ships — one to two hundredths of one percent. The right way to say that is as a range with the source of the imprecision named, not as a single decimal that implies a disclosure nobody made.
The sixfold that does not look like sixfold
Amazon's release calls the expansion "a sixfold increase from its current footprint." The same release says Prime Air currently operates 11 locations across seven states — Arizona, Florida, Kansas, Louisiana, Michigan, Nebraska and Texas — each covering roughly 175 square miles, and that it is going to nearly 500 cities and towns.
Eleven to five hundred is 45×. Sixfold is 6×. Those cannot both describe the same quantity, and no coverage we have seen reconciled them.
They describe different things, and the reconciliation is arithmetic rather than opinion. Amazon counts municipalities inside each site's coverage circle, not sites. A single 175-square-mile service area touches many towns. Third-party analysis of the existing footprint puts today's 11 sites at roughly 80 municipalities on that counting method — and 80 × 6 = 480, which is "nearly 500". The sixfold claim is internally consistent; it is simply not measured in the unit the "11 locations" figure uses.
That matters for reading the announcement. This is a sixfold expansion of covered towns, not a 45× expansion of infrastructure — the drone sites, the airspace approvals and the operations staff scale nowhere near the headline count. It makes the programme cheaper and more achievable than the raw jump implies, and correspondingly less of a step change than "500 cities" sounds. The ~80 figure is third-party analysis, not an Amazon disclosure, so the reconciliation is a strong inference rather than a confirmed one.
Uber's number is a different kind of number
One million deliveries per day by the end of 2029 is 365 million a year.
Zipline has flown, by its own account, more than two million deliveries in its entire history across commercial work with Walmart, Chipotle, Panera and Wendy's plus years of healthcare logistics in Africa. So the partnership's target is:
- 182× Zipline's cumulative lifetime volume — per year;
- half of everything Zipline has ever delivered, every day;
- 365× the volume Amazon's executive projects for all of 2026.
The second chart is that comparison, and the reason it is a chart is that the bars are unreadable without their printed totals — which is the point.
Priced at Amazon's published fees, purely to give the target a revenue scale, 365 million deliveries a year is $1.09B to $1.82B, or 2.0% to 3.3% of Uber's $55.2B trailing revenue. Uber has published no drone pricing, so those are Amazon's numbers applied to Uber's target, and they are illustrative arithmetic rather than a forecast. What they establish is the shape: a target that would be material to Uber, three years out, against a programme that is immaterial to Amazon today.
The number neither company published
The figure that would settle whether any of this matters is cost per delivery, against the cost of the van or the courier it replaces. Neither company has ever disclosed it. Zipline has not. Amazon has not, in thirteen years.
This is where most analysis quietly invents a number. We are not going to. What can be said is what the absence implies: a delivery method that is faster and more expensive is a marketing programme, and a delivery method that is cheaper at scale is a line item that eventually shows up in cost of sales. Until one of them publishes a unit cost, the case for drone delivery is unfalsifiable in both directions — which is exactly why the announcements are written in deliveries and cities rather than dollars.
Two ways to buy the same option, on two very different balance sheets
The structural difference between the two announcements is more interesting than either headline, and it is visible in the cash figures we already store.
| Amazon | Uber | |
|---|---|---|
| Model | Builds and operates its own drones and sites | Equity investment in an operator, integrated into Uber Eats |
| Capital | On balance sheet, inside a capex programme | Undisclosed minority stake; Zipline carries the fleet |
| Stated ambition | ~500 cities by end-2026; ~1M deliveries in 2026 | 1M deliveries per day by end-2029 |
| TTM revenue | $775.7B | $55.2B |
| TTM free-cash-flow margin | 2.14% | 18.33% |
Amazon's trailing free cash flow is $14.8B on the $691.3B of revenue in the same four quarters — a 2.14% free-cash-flow margin, because capex is eating the company. Our stored FCF series runs three quarters behind the revenue series, so this window ends with the quarter to September 30, 2025; the margin is computed against the revenue of those same quarters rather than against the newer $775.7B, which would flatter nothing and mislead slightly.
Uber's is 18.33% — $10,124M of free cash flow on $55,226M of revenue, the trailing figure the company made a milestone of three weeks ago.
That gap is the argument for each company's chosen structure. Amazon is adding another owned, capital-hungry logistics asset to a business whose capex is already the swing factor in its cash generation — and it can afford to, because Prime Air is small enough to disappear inside the rounding. Uber is buying the same optionality with someone else's fleet and someone else's certification burden, which is the only version of this that a platform converting 18% of revenue to cash should want. Neither has told anyone what a delivery costs.
What to watch
- Whether Amazon repeats the million-delivery figure in a public filing or earnings call. An executive's internal projection reported by CNBC and a disclosed number are not the same thing, and the difference will show up first in whether it is ever repeated on the record.
- The municipality count Amazon uses when it claims completion. If "nearly 500" is measured in towns inside coverage circles, the honest scoreboard at the end of 2026 is the number of sites, which is what actually has to be built.
- Any unit-cost disclosure from either side. Cost per drone delivery against cost per van delivery is the only figure that converts either programme from a scale story into an economics story.
- Uber's first Uber Eats drone deliveries, expected later in 2026 in Zipline's existing markets around Houston and Dallas. The gap between that start and 1,000,000 a day is three years and five orders of magnitude.
- Whether Prime Air ever gets a line. Amazon breaks out AWS, advertising and third-party seller services. The quarter Prime Air appears in a segment disclosure is the quarter it stops being a rounding error.
Sources and provenance. Amazon's expansion to nearly 500 US cities and towns, the 11 current locations across seven states, the ~175 square miles per site, the next launches (Chicago, Atlanta, Cleveland, Syracuse, Boise), the "sixfold increase" language and the $2.99/$4.99/free pricing are all disclosed in Amazon's release of August 19, 2026, and corroborated by CNBC and GeekWire coverage. The projection of one million deliveries in 2026 is claimed, not disclosed: it is attributed to VP David Carbon in an internal March meeting and reported by CNBC; Amazon's own public language is "hundreds of thousands" this year and "thousands daily". The estimate that today's footprint spans roughly 80 municipalities is third-party analysis, not an Amazon figure, and the sixfold reconciliation rests on it. The Uber–Zipline partnership, the undisclosed equity investment, the one-million-deliveries-per-day target by end-2029 and the plan to start Uber Eats drone deliveries later in 2026 around Houston and Dallas are claimed — announced jointly by the two companies on August 17, 2026, with no committed capital figure. Zipline's ">2 million lifetime deliveries" and ~$7.6B valuation are the company's own figures. Amazon's US package volume is not disclosed at annual precision; the 0.014%–0.020% range is bounded against a 5–7 billion parcel assumption and is deliberately given as a range. All fee-revenue figures are derived — arithmetic on Amazon's published prices and the claimed delivery volumes — and the Uber revenue-scale figures apply Amazon's prices to Uber's target because Uber has published no drone pricing. Amazon's $775.7B trailing revenue (+15.8%), $14.8B trailing free cash flow and 2.14% margin, and Uber's $55,226M revenue and $10,124M free cash flow (18.33%), are derived from the quarterly series stored on this site; Amazon's free-cash-flow window ends with the quarter to September 30, 2025 and its margin is computed against the revenue of those same quarters. Neither AMZN nor UBER has a published R40 model, so no model reconciliation was available for this piece.
