Autonomous Delivery Is Already at Scale — The Numbers Most People Don't Know
The Scale Nobody Is Talking About
Starship Technologies has completed 9 million autonomous deliveries.
Inceptio Technology has driven 250 million autonomous trucking miles.
These are not projections. They are not pilot programmes. They are not press releases about future capabilities. They are operational figures from systems running today, reported by ARK Investment Management LLC (2026), Starship Technologies (2025), and Inceptio Technology (2025).
The reason most people do not know these numbers is that autonomous delivery happened quietly — on university campuses, in suburban neighbourhoods, on Chinese highway freight corridors — before the media cycle had time to frame it as a revolution. By the time anyone started paying attention, the operations were already at scale.
This article documents what the deployment data actually shows, without projection or speculation. The numbers are sufficient on their own.
Browse all commercial delivery robot platforms currently tracked by Geppetto: Delivery Robot Category →
The Three Categories of Autonomous Delivery
Autonomous delivery is not a single technology. It is three parallel deployment tracks, each with distinct leaders, geographies, and unit economics.
1. Sidewalk and Campus Robots
Ground-based autonomous delivery robots — operating on pavements, in pedestrian zones, and across university campuses — have accumulated the largest single-operator delivery count of any autonomous logistics category.
| Operator | Cumulative Deliveries | Notes |
|---|---|---|
| Starship Technologies | 9,000,000 | Pavement/campus robots, US & Europe |
| Meituan | 4,900,000 | Campus and urban robots, China |
| Coco Robotics | 500,000 | Pavement robots, US cities |
| Avride | 200,000 | Campus and urban, US |
| Serve Robotics | 100,000 | Pavement robots, LA via Uber Eats |
Source: ARK Investment Management LLC, Big Ideas 2026; Starship Technologies 2025.
Starship's 9 million deliveries represent a lead that is not remotely close. The company has been operating commercially since 2018 across Milton Keynes (UK), Tallinn (Estonia), and dozens of US university campuses. Its cumulative figure is the clearest single data point demonstrating that sidewalk autonomous delivery is not an experimental technology — it is a proven logistics operation.
For a direct comparison of the leading US platforms: Starship vs Serve Robotics →
Platform profiles: Starship Technologies Delivery Robot → | Serve Robotics Serve 4 → | Kiwibot →
2. Drone Delivery
Autonomous drone delivery has accumulated significant volume concentrated in medical logistics (Zipline) and consumer food delivery (Meituan, Wing, Manna).
| Operator | Cumulative Deliveries | Primary Use Case |
|---|---|---|
| Zipline | 2,000,000 | Medical/commercial, Africa & US |
| Meituan | 780,000 | Food delivery, China |
| Wing (Alphabet) | 750,000 | Consumer goods, Australia & US |
| Manna | 200,000 | Food delivery, Ireland |
| FlyTrex | 200,000 | Consumer goods, US |
| Matternet | 60,000 | Medical logistics, Switzerland & US |
Source: ARK Investment Management LLC, Big Ideas 2026.
Zipline's 2 million deliveries are the most operationally significant figure in drone logistics. Unlike food delivery operators whose volume is concentrated in permissive regulatory environments, Zipline has proven the model in multiple jurisdictions including regulated US airspace. Its transition from medical-only to commercial consumer delivery in the US (launched 2024) represents the most consequential expansion in drone logistics to date.
Meituan's combined ground robot (4.9M) and drone (780K) figures make it the largest autonomous delivery operator by total volume globally — a fact that receives minimal coverage in Western technology media.
Agricultural drone context: DJI Agras T50 →
3. Autonomous Trucking
Autonomous freight haulage has accumulated its largest volumes not in the United States, but on Chinese highway corridors.
| Operator | Cumulative Autonomous Miles | Geography |
|---|---|---|
| Inceptio Technology | 250,000,000 | China highway freight |
| Pony.ai | 4,200,000 | China & US |
| Aurora | 3,800,000 | US commercial freight |
| Kodiak | 3,000,000 | US highway |
| Gatik | 2,100,000 | US middle-mile |
Source: Inceptio Technology 2025; ARK Investment Management LLC, Big Ideas 2026.
Inceptio's 250 million miles are in a different order of magnitude from every other autonomous trucking operator. This is not a technology demonstration — it is a commercial freight operation at highway scale, generating the kind of real-world performance data that cannot be replicated in test environments.
Aurora's position is distinct: in April 2024, Aurora became the first company to operate fully driverless commercial freight trucks on US public roads, on routes between Dallas and Houston. Its 3.8 million miles are at lower absolute volume than Inceptio but represent the most advanced US regulatory milestone in autonomous trucking.
For a data-driven analysis of what this means for the profession most directly affected: Will Robots Replace Truck Drivers? →
The Geography of Deployment
The most important structural fact in the autonomous delivery data is one that the headline numbers obscure: the majority of this volume is outside the United States.
Inceptio's 250 million autonomous trucking miles are operated in China. Meituan's combined 5.68 million robot and drone deliveries are in China. The global volume leader in sidewalk robots (Starship) is headquartered in Estonia and built its operational base in the UK and northern Europe before expanding to US campuses.
The United States leads in autonomous vehicle technology development by most measures — sensor systems, AI frameworks, regulatory sophistication, and capital investment. It does not lead in autonomous delivery deployment. The deployment leaders are operating in geographies with either lower regulatory friction (China's approach to autonomous freight), purpose-built infrastructure (Zipline's fixed-wing logistics networks in Rwanda and Ghana), or contained operating environments that sidestep public road regulation entirely (university campuses).
This matters for forecasting. When US regulatory frameworks for drone delivery and autonomous trucking mature — a process that is visibly accelerating, with FAA BVLOS rulemakings and NHTSA autonomous vehicle guidance both advancing — the technology and operational playbooks will already exist at proven scale. The US deployment curve will not require a technology development phase. It will require a regulatory unlocking phase, after which deployment can ramp rapidly against an existing global operational base.
For context on the delivery driver profession and displacement risk: Will Robots Replace Delivery Drivers? →
The Cost Collapse
Deployment at scale has a direct consequence for unit economics. The ARK 2026 analysis documents cost trajectories that, if sustained, restructure the economics of every business that relies on physical delivery.
Autonomous trucking:
- Human-driven truckload: $0.07 per ton-mile
- Autonomous truckload: $0.03 per ton-mile
- Reduction: ~57%
Last-mile delivery:
- App-dispatched human delivery (current): $15 per order
- Drone or robot delivery (trajectory): under $1 per order
- Reduction: ~93%
Source: ARK Investment Management LLC, Big Ideas 2026.
The last-mile figure is the more structurally significant of the two. Trucking cost reduction of 57% is substantial — it reprices freight economics and changes the competitive position of every logistics operator. But it operates within an existing framework of how goods move.
A 93% reduction in last-mile delivery cost does not operate within the existing framework. It changes what gets ordered, how often, and in what quantities. At $15 per delivery, consumers consolidate orders, add items to justify the delivery fee, and think twice about ordering low-value items. At under $1 per delivery, those constraints disappear. The marginal cost of an additional delivery approaches zero, which means the marginal decision to order approaches zero. Same-day delivery at near-zero marginal cost is not a faster version of current delivery. It is a different consumer behaviour model.
The businesses that model their supply chains, inventory positioning, and customer acquisition costs against a sub-$1 last-mile world before it arrives will have a structural advantage over those that reprice reactively.
For a full breakdown of autonomous vehicle cost economics: Robotaxi Cost Per Mile 2026 →
The Trucking Quiet Revolution
Of the three autonomous delivery categories, autonomous trucking has received the least proportionate coverage relative to the scale of its operational progress.
Aurora's April 2024 fully driverless commercial launch on US public roads was a regulatory milestone with no precedent in autonomous vehicle history. The vehicles operated without a safety driver, on live public highway, carrying commercial freight for paying customers. That this event passed without sustained mainstream coverage reflects a structural bias in technology journalism toward consumer-facing products over industrial infrastructure.
Inceptio's 250 million autonomous trucking miles in China represent a commercial operation of comparable scale to mid-tier logistics companies. The system operates on production freight routes — not test tracks, not geofenced demo corridors, not supervised pilot schemes. The commercial validation work that Aurora, Kodiak, and Gatik are conducting in the US has already been conducted, at much larger scale, in China.
The highway is already automating. It is automating at a pace and scale that most public commentary has not registered, because trucks do not generate the visual drama of humanoid robots, and Chinese freight corridors do not feature in Western technology media.
Geppetto currently tracks 8 commercial delivery robot platforms across ground, aerial, and logistics categories. Browse the full delivery robot directory →
What Comes Next
The deployment trajectories across all three categories point toward near-term milestones that will make the current scale figures look like early baselines.
Sidewalk and campus robots are expanding beyond their current operational strongholds. Serve Robotics' integration with Uber Eats represents the first large-scale attempt to embed autonomous delivery robots into an existing consumer delivery platform with network scale — removing the need to build demand from scratch. If the unit economics hold in open urban environments (a question the operational data from LA deployments is answering in real time), the transition from campus to city becomes a distribution and regulatory problem, not a technology problem. Best Delivery Robots for Business 2026 →
Drone delivery is approaching the regulatory inflection in the US that ground robots already navigated. The FAA's Beyond Visual Line of Sight (BVLOS) rulemaking framework, combined with Wing's operational experience in Texas and Virginia and Zipline's commercial launch in Utah, is generating the regulatory precedents that will underpin broader US urban drone delivery. The international operational data — particularly Manna's density in Irish urban environments and Wing's suburban deployments in Australia — provides regulators with real-world safety performance records that accelerate US approvals.
Autonomous trucking is in route expansion mode. Aurora's commercial launch covers Dallas-Houston; additional routes are scheduled. Kodiak and Gatik are expanding their route networks. The constraint is not technology readiness — it is the pace at which insurance frameworks, interstate regulatory harmonisation, and shipper contracting can adapt to a fully driverless operational model. These are solvable problems with known timelines, not open research questions.
The Cricket's View
Starship has completed 9 million deliveries. Inceptio has driven 250 million autonomous trucking miles. These numbers exist. They are not projections. They are not pilots. They are operations running today.
The reason most people don't know them is that autonomous delivery happened quietly — in China and on university campuses — before anyone was paying attention. The transition to autonomous logistics did not announce itself. It just started.
The question is no longer whether autonomous delivery works at scale. The question is how quickly the regulatory and commercial infrastructure in each geography adapts to an operational reality that has already arrived elsewhere. The countries and companies that treat that adaptation as urgent are operating in the correct frame. The ones still debating whether autonomous delivery is technically feasible are operating one transition behind.
Frequently Asked Questions
How many autonomous deliveries have been completed globally?
As of 2025–2026 data, cumulative autonomous ground robot deliveries exceed 14.7 million across all operators, with Starship Technologies (9 million) and Meituan (4.9 million) accounting for the majority. Drone deliveries exceed 3.9 million cumulative, led by Zipline (2 million), Meituan (780,000), and Wing (750,000). Autonomous trucking miles exceed 263 million, with Inceptio Technology (250 million) representing the dominant share. Sources: ARK Investment Management LLC Big Ideas 2026; Starship Technologies 2025; Inceptio Technology 2025.
Which company has completed the most autonomous deliveries?
Starship Technologies leads cumulative autonomous ground robot deliveries at 9 million as of 2025. Across all modalities combined, Meituan's combined ground robot (4.9 million) and drone (780,000) volumes make it the largest single operator by total autonomous delivery count globally at approximately 5.68 million.
Where has autonomous trucking been deployed at the largest scale?
Inceptio Technology, operating in China, has accumulated 250 million autonomous trucking miles on commercial highway freight routes — approximately 60x the combined autonomous miles of the leading US operators (Aurora, Kodiak, Gatik, Pony.ai). In the United States, Aurora became the first company to operate fully driverless commercial freight trucks on public roads in April 2024, on routes between Dallas and Houston.
What is the projected cost of autonomous last-mile delivery?
ARK Investment Management LLC (2026) projects autonomous drone and robot delivery costs of under $1 per order, compared to approximately $15 per order for app-dispatched human delivery — a reduction of approximately 93%. Autonomous truckload freight is projected at $0.03 per ton-mile versus $0.07 per ton-mile for human-driven operations, a reduction of approximately 57%.
Why has autonomous delivery scaled faster outside the United States?
The primary factors are regulatory framework and operating environment. China's approach to autonomous freight regulation has enabled commercial operations at highway scale without the phased driverless testing requirements applied in US jurisdictions. Starship's UK and European deployments benefited from pavement robot regulations that were defined earlier and more permissively than equivalent US local ordinances. University campus deployments in the US succeed because they operate on private or semi-private infrastructure outside public road regulation. The US leads in technology development but lags in deployment density, primarily due to regulatory timeline differences rather than technical capability gaps.
What delivery robots are currently available for commercial deployment?
Geppetto tracks 8 commercial delivery robot platforms including the Starship Technologies Delivery Robot, Serve Robotics Serve 4, and Kiwibot. Browse the full delivery robot category →
How does drone delivery regulation work in the United States?
US drone delivery operates under FAA Part 107 rules, with commercial Beyond Visual Line of Sight (BVLOS) operations requiring individual waivers or operating under emerging BVLOS rulemaking frameworks. Wing has received FAA Air Carrier Certification and operates in Virginia and Texas. Zipline launched US commercial consumer delivery in 2024 under Utah operational approvals. The FAA's BVLOS rulemaking process, expected to produce final rules in the 2025–2026 timeframe, is expected to significantly reduce the per-operator regulatory burden and enable broader urban deployment.
Further Reading
- Amazon's Robot Density Record
- Robotaxi Cost Per Mile in 2026
- How Much Does a Delivery Robot Cost in 2026?
- Wright's Law and Robot Cost Decline
Further Reading
- Amazon's Robot Density Record
- Robotaxi Cost Per Mile in 2026
- How Much Does a Delivery Robot Cost in 2026?
- Wright's Law and Robot Cost Decline