Amazon's Project Tetromino: What Fully Automated Delivery Stations Mean for DSPs
An internal plan reported by Business Insider sketches robot-run delivery stations starting with a 2028 pilot. Amazon says the numbers are wrong. Either way, the timeline tells delivery partners something useful.
What the report says
In late August 2026, Business Insider reported on an internal Amazon document describing Project Tetromino, a plan to automate final-mile delivery stations with AI and robotics. The codename is reported to be a nod to Tetris. The goal is a station where packages are sorted and staged for delivery vans with minimal human handling.
According to the document as reported, the rollout would look like this:
| Phase | Reported scope | Reported cost |
|---|---|---|
| 2028 | Initial pilot | $103 million |
| 2029 | About five additional sites | Roughly $85 million each |
| By 2030 | About ten more locations | Not stated |
The document reportedly puts total capital spending above $530 million by 2029, and projects the new stations would process packages about 2.5 times faster than Amazon's current delivery station designs.
What Amazon disputes
An Amazon spokesperson told Business Insider that Project Tetromino is an early-stage concept, and said the financial projections and timeline in the document are inaccurate and do not reflect current plans. Amazon added that it keeps testing technology aimed at employee safety and customer experience. Treat every number above as a reported internal estimate, not a commitment.
Where Boxbot fits
The report names Boxbot, a robotics startup, as a possible technology partner. Boxbot's system moves packages from conveyors onto trays and uses software to sequence them by route. The startup claims this can make vehicle loading up to ten times faster. The document only suggests Boxbot could be part of the system; no deal has been confirmed publicly.
Why the timeline matters more than the robots
The interesting part for anyone running a station today is not the technology. It is the schedule. If everything in the document happened exactly as written, the first automated station would open in 2028 and roughly 16 sites would exist by 2030. That leaves the 2026 and 2027 peaks, and most of 2028, running on the same manual process stations use now: parcels come off the truck, people sort them by route, and at many stations outside Amazon's network, drivers finish the sort themselves before they can load.
The report also confirms a direction. A company with Amazon's volume is willing to spend tens of millions of dollars per site to take sorting off people's hands. That tells you where the cost of manual sorting sits in a station's economics.
Large networks automate centers first
Big carriers have automated sortation centers for years. The delivery station, the last stop before the van, is where manual work is still common, because each site is small, layouts vary and volume swings with the season. A plan like Tetromino is expensive precisely because it is trying to solve that last step at scale.
What independent stations can do now
Project Tetromino is Amazon's own project for Amazon's own stations, where Amazon associates already do the sort. For independent last-mile stations and delivery partners that sort their own freight, for example for regional e-commerce carriers, you do not need a $100 million budget to take the sort off drivers. A single automated line, sized to your routes and your space, covers most small and mid-size stations:
- Size it to your window. A SortLease S15 line is rated at up to 3,600 parcels per hour. Add lines in parallel or stack them vertically when volume grows.
- Pay for use, not equipment. SortLease installs, runs and maintains the line, and you pay per parcel from $0.10. Purchase is quoted per configuration if you prefer to own.
- Share a site. Three or four delivery partners can share one sorter through the shared hub model.
- Keep your carrier relationship yours. The line reads printed labels and does not connect to carrier systems. We never contact your carrier without your permission.
With last-mile driver pay rising (see Amazon DSP raise 2026 and 2027 for a benchmark), every minute of hand-sorting costs more. See who this fits on our station sorting page. Compare the two options in our manual vs automated sorting cost comparison, run your own numbers in the driver sorting cost calculator, or see an installation in progress at our Palm Springs page.
Frequently asked questions
What is Amazon's Project Tetromino?
It is an internal Amazon plan, reported by Business Insider in August 2026, to fully automate final-mile delivery stations with AI and robotics so packages are sorted and staged for vans with minimal human handling.
When would Project Tetromino start?
The reported document describes a pilot in 2028, about five more sites in 2029 and about ten more by 2030. Amazon says the timeline is inaccurate and calls the project an early-stage concept.
How much would Project Tetromino cost?
The document reportedly budgets $103 million for the 2028 pilot and roughly $85 million for each 2029 site, with total spending above $530 million by 2029. Amazon disputes these projections.
Is Boxbot part of Project Tetromino?
The document suggests Boxbot's technology could be used, but no partnership has been publicly confirmed.
Does this change anything for DSPs in 2026 or 2027?
Not directly. Even on the reported schedule, the first automated station opens in 2028, so most stations will keep sorting by hand through the next peak seasons. At Amazon stations, sorting stays with Amazon. Independent stations that sort their own freight can automate sooner with a single pay-per-parcel line.
Sources
- Yahoo Finance / Stocktwits: Amazon reportedly plans automation initiative for delivery stations (citing Business Insider)
- GeekWire: Amazon eyes 'fully automated' delivery stations to bring robotics to the last mile
- Omega: Amazon's Project Tetromino targets full delivery station automation by 2030
SortLease is not affiliated with the companies or agencies named here. Third-party facts are summarized from the sources above; SortLease capacity figures are rated specifications, not measured results.