Inside GXO's plan for up to 1,000 Apptronik humanoids: picking, packing, and a $15-an-hour target
GXO and Apptronik: 1,000-robot deal
GXO has kind of been the logistics industry's biggest humanoid test lab for almost three years. Now it's laying out what a scaled humanoid workforce could actually look like.
The contract logistics giant signed a letter of intent with Apptronik today to expand trials of the Apollo humanoid starting in 2027, then progress toward a global rollout of up to 1,000 robots over four years. First commercial deployments are slated for North America in late 2027.
The commercial model will be robots-as-a-service.
GXO has said it expects to have nearly 20,000 robots working across its network by the end of 2026. Nearly all of that is conventional warehouse automation. About 40% of its warehouses already have heavy robotics fit-outs, CEO Patrick Kelleher said in an August interview with FreightWaves. Humanoids are the layer GXO hopes to add on top, doing the manual work that fixed automation still leaves for people.
45+ humanoid robot pilots
GXO has done a lot of homework. The company has completed more than 45 humanoid pilots across vendors including Agility, Reflex Robotics and Apptronik.
"We've done 45 pilots on humanoids so far. We have an additional pilot launching in Europe very shortly," Kelleher said on GXO's Q2 earnings call in August.
Under the LOI, Apollo will be working across a range of warehouse workflows, including five that GXO and Apptronik name specifically. Each one tests a different part of the robot.
- Goods-to-person picking is the closest thing to a fixed workstation. Automated systems bring totes or shelves to the robot, and the robot picks individual items out of them. Mobility barely matters here. Hand dexterity, grasp planning and cycle time are everything, because the humanoid is effectively replacing a human picker at a station designed around human reach and speed.
- Person-to-goods picking flips that: the picker walks the aisles to the inventory. Now Apollo needs navigation, localization in dynamic spaces shared with people and forklifts, and the ability to pick from shelving at varying heights. It's a much harder autonomy problem.
- Case and tote movement is the classic first humanoid job, and it's mostly a payload and endurance problem. Agility's Digit has been doing a version of this at GXO's Flowery Branch, Georgia site for more than two years, moving totes between AMRs and conveyors.
- Packing is arguably the hardest of the five. It involves deformable items, bags, boxes of varying sizes and fill decisions. That's exactly the type of fine, bimanual manipulation humanoid makers are training models for right now.
- Sortation sits in the middle: repetitive, fairly structured, but high-volume and speed-sensitive.
That's a much broader scope than GXO's other humanoid programs. Digit's upcoming expansion to Statesville, North Carolina and Lebanon, Tennessee in early 2027 focuses on repetitive material handling for e-commerce fulfillment. The Apptronik trials explicitly target picking and packing, which is where most warehouse labor actually goes.
GXO and Apptronik have already done a fair amount of groundwork. Earlier this year Apollo completed a U.S. pilot with GXO focused on autonomous picking of small consumer goods. And the two companies have run data collection and live autonomous evaluations on more than 50 logistics tasks at Apptronik's Robot Park in Austin.
Economics & ROI of humanoid robots in logistics
Kelleher has been unusually specific about the economics.
In the same FreightWaves interview, Kelleher said a humanoid’s operating costs (excluding the robot itself) run $10 to $15 an hour today and should fall below $10 within two years, while the price of the robots should drop by half. All in, he expects humanoids to cost $15 to $20 an hour within two years.
That’s the level he said GXO needs for return on investment.
That target is benchmarked against human labor, and it's not a high bar compared to fully loaded warehouse wages in the U.S. or Western Europe. The catch is productivity. Kelleher has pointed to dexterity and task speed as the biggest remaining gaps. A robot that costs $15 an hour but picks at half human speed is effectively a $30-an-hour picker.
The 4-year deal rollout timeline lines up with Kelleher's forecast. If unit costs halve and productivity climbs over the next two years, robots arriving in late 2027 and beyond land right when the math is supposed to start working.
RaaS helps too. GXO isn't buying 1,000 robots up front. It pays for robots as a service, and each use case has to meet jointly agreed productivity, safety and operational KPIs before it scales. That lets GXO scale robot by robot and workflow by workflow.
Apollo 2, maybe Apollo 3 and 4 too
The release just says "Apollo," but this is clearly going to be Apollo 2, at least initially.
Apptronik launched Apollo 2 at the end of June, alongside its nearly 90,000-square-foot Robot Park facility. It comes in two configurations: bipedal, and a wheeled base designed to fit within existing industrial mobile robot safety standards. Apptronik has said Apollo 2's data will feed development of a future Apollo 3.
The original Apollo was spec'd at 5-foot-8 and about 160 pounds, with a 55-pound payload and a four-hour swappable battery. Apptronik hasn't yet published a numeric spec sheet for Apollo 2.
For warehouse work, the wheeled-or-legged question matters. Warehouses have flat floors, so legs mostly matter for stairs, mezzanines, narrow gaps and tight spaces, and wheels are easier to certify for safety and more energy-efficient. A wheeled Apollo that fits existing mobile robot safety standards could be the faster path to deployment in live facilities alongside people.
Safety is a big part of the trial scope: the trials will validate safety, security and technical requirements for scaled deployment.
My guess: the robots that arrive in late 2027 won't be the ones in Robot Park today. A four-year deal almost certainly spans at least 1 hardware generation, and more likely 2 or 3.
Robot Park isn't just a test facility. It's a data factory. Fleets of Apollo 2 robots collect real-world task data there and at customer sites, and that data trains Google DeepMind's Gemini Robotics models under Apptronik's research partnership with DeepMind.
GXO is one of those customer sites. So every GXO trial generates exactly the type of data needed to make Apollo better at GXO's work. That's a meaningful advantage when the hard problems are long-tail picking and packing, where performance is mostly limited by how much relevant training data a model has seen.
"Our expanded partnership with GXO shows what's possible when you pair a leading logistics provider with robots designed from the start for thoughtful collaboration with people," Apptronik CEO Jeff Cardenas said.
Apptronik has the funding to build out that loop. It raised $935 million in its Series A and extension, and now has nearly 500 employees.
GXO is making multiple bets on the future
GXO isn't picking a single winner. It now has scaled-deployment paths with two humanoid makers, aimed at somewhat different work.
Agility's Digit, a bipedal robot focused on tote and material handling, is GXO's longest-running humanoid deployment. GXO was Agility's first commercial customer under the industry's first humanoid RaaS deal in 2024, and Digit 5, launched in September, promises cooperative safety for working near people without cages. Agility began trading on Nasdaq as AGLT last week.
Apptronik's Apollo, in bipedal and wheeled versions, is now aimed at picking, packing and sortation as well as case and tote movement, with up to 1,000 units on the table.
Reflex Robotics, which makes a wheeled humanoid, signed a RaaS pilot with GXO in 2024 for tote transfers and picking.
For a 3PL with more than 1,000 facilities and hundreds of customers, that makes sense. Different sites and workflows will favor different form factors, and it's still very early in the market. Eventually, however, you'd assume that GXO would pick just 1 vendor.
Sources
- GXO/Apptronik press release, Nov. 12, 2026
- GXO is expanding Digit to 2 more warehouses (HumanoidMegaHub)
- Apptronik Announces Robot Park (Forbes)
- Apptronik Scores $935 Million (Forbes)
- AGLT: First Pure-Play Humanoid Robot Company Starts Trading On Nasdaq (Forbes)
- GXO CEO interview (FreightWaves)
- Agility and GXO expansion, Q2 call quote (The Robot Report)
- Apptronik Apollo 2 and Robot Park (The Robot Report)
Written by
I tell stories with data and build startups. I've raised angel and VC funds, led software companies and startups, and chronicled the rise of the mobile economy. Currently I write for Forbes, host several podcasts, and am starting up HumanoidMegaHub. I was TUNE's mobile economist and Singular's VP of Insights, built the research team at VentureBeat, and managed teams creating software for partners like Intel and Disney. In past lives I built websites, e-learning platforms, and mobile apps, and, a long time ago, was named to Folio's top 100 of the media industry's "most innovative entrepreneurs and market shaker-uppers.”