Agility Doubles Digit Deployment Across GXO Warehouses

Humanoid robotics has spent years producing impressive videos, but warehouse operators pay for repeatable work across shifts and locations. On October 26, 2026, GXO and Agility expanded Digit from one established operation to two additional US fulfillment sites, with deployment planned for early 2027. The agreement matters because it tests whether a humanoid can move beyond a carefully tuned first facility without restarting the learning process at each building. Agility says Digit has accumulated more than 65,000 hours of real-world operation, while the existing GXO deployment has moved more than 100,000 totes. Those figures do not prove attractive economics or unrestricted human collaboration. They do give the companies a stronger base than a staged demonstration. The next question is whether operating experience can be packaged into a repeatable deployment method.

What Actually Happened

GXO and Agility announced a new commercial agreement on October 26, 2026. According to the GXO announcement, Digit robots will support e-commerce fulfillment at facilities in Statesville, North Carolina, and Lebanon, Tennessee. Deployment is expected to begin in early 2027 with Digit 4, followed by a migration to Digit 5 when the new generation becomes commercially available. The robots will handle repetitive material-moving work rather than a broad list of human tasks. That narrow scope is important. Warehouse automation succeeds when a system performs a defined job with predictable throughput and integrates with the surrounding flow. The deal expands a paid operating relationship instead of creating a new pilot with an unfamiliar customer. It therefore offers a cleaner test of whether Agility can repeat what worked at GXO’s Georgia site.

Agility’s parallel commercial update added the wider context. The company reported three additional manufacturing customers and an active pipeline of more than 30 prospective customers across automotive manufacturing, logistics, e-commerce, agricultural equipment, consumer products, and defense. It described the GXO expansion as a progression from one deployment to more facilities within the same customer. Agility also said Digit has logged more than 65,000 operating hours in real-world programs. Those are company-reported figures, not an independent audit, but they are more useful than a choreographed task count because they indicate exposure to faults, maintenance, floor variation, and operational interruptions. They also give future customers a baseline for questions about reliability and service staffing. The event date passes the gate because both the expanded agreement and the commercial metrics were released on October 26.

A filing tied to Agility’s planned public listing provides another concrete measure. The SEC-filed investor call transcript says Digit moved more than 100,000 totes at GXO’s Flowery Branch, Georgia facility before the expansion to two new sites. GXO described the 2024 relationship as the first formal commercial deployment of humanoids in live logistics operations under a robots-as-a-service agreement. The new locations broaden the site count but do not disclose the number of robots, contract value, hourly price, or target throughput. Those missing details limit any claim about scale. The announcement proves geographic expansion and continued customer commitment. A renewal or expansion by the same buyer is stronger evidence than an unrelated demonstration. It does not yet prove that dozens or hundreds of units can operate profitably across a network.

Why This Matters More Than People Think

The hardest step in industrial robotics is often the move from a working cell to a repeatable deployment. A first site receives intense engineering attention. Teams map the floor, tune perception, adjust workflows, train staff, and solve edge cases. Those costs can be justified as development. A second and third site reveal whether the solution has become a product. If deployment time falls, remote support replaces on-site experts, and the same robot handles similar work with limited retraining, Agility can spread engineering costs across more units. If every building requires custom code and constant staff, revenue can grow while margins remain weak. GXO offers a useful test because the operator understands its own processes and can compare performance across facilities. The expansion therefore measures repeatability, not merely robot capability.

Digit’s human-shaped mobility is valuable only if it avoids expensive building changes. Warehouses already contain aisles, racks, totes, doors, and work heights designed for people. A robot that uses those spaces can target tasks where fixed automation or conveyors are difficult to justify. Yet humanoid form also introduces balance, actuator, battery, and safety challenges. Wheeled autonomous mobile robots are simpler for many horizontal moves. Robotic arms are faster for fixed picking stations. Digit must win tasks where legs, reach, and general movement offset higher mechanical complexity. GXO’s choice of repetitive material handling suggests that the current value lies in connecting existing process steps rather than replacing an entire warehouse system. Expansion to two sites will show whether that middle layer contains enough work to support a fleet.

However, critics argue that site count can overstate commercial maturity when unit count and economics remain private. Two buildings may host a small number of robots with heavy vendor support. Company-reported operating hours do not reveal productive utilization, intervention rates, mean time between failures, or the cost of spare parts. Moving 100,000 totes sounds large, but the number needs context such as totes per hour, shifts worked, and comparison with human or conventional automation. The risk is that investors treat continued experimentation by a sophisticated customer as proof of a finished labor substitute. GXO has reason to test emerging systems even before they meet standard automation returns. The next phase must produce operating data that separates a strategic innovation program from a scalable service business.

The Competitive Landscape

Agility competes with humanoid developers including Tesla, Figure, Boston Dynamics, Apptronik, Unitree, UBTECH, and AgiBot. It also competes with less glamorous machines that already move goods. Locus, Geekplus, Amazon Robotics, and other autonomous mobile robot vendors offer mature fleet systems. Fixed industrial arms handle high-speed repetitive work. Conveyors move large volumes with predictable uptime. Agility’s advantage is a narrow history of paid humanoid operation in live facilities and a robot designed around warehouse material flow. Tesla and Figure can draw on deep capital and manufacturing talent. Chinese suppliers can pressure hardware prices and ship large volumes. Agility must turn its operating lead into deployment methods, safety evidence, service infrastructure, and customer renewals before rivals catch up or simpler automation covers the same tasks.

The historical parallel is the early adoption of industrial robotic arms in automotive plants. The first machines were expensive, narrow, and separated from workers, but they delivered value on tasks that were repetitive and hazardous. Adoption grew as integrators standardized cells, safety systems, maintenance, and programming. Humanoids may follow a similar path, though the useful unit is likely a workflow rather than a robot. Buyers need tote types, staging rules, exception handling, charging, fleet management, and worker interaction to operate as one system. A capable biped without those pieces remains a research platform. GXO and Agility are effectively building an integration template through repeated use. If the template transfers across sites, the market can grow faster than robot capabilities alone would suggest.

Digit 5 is central to the competitive claim because the new sites plan to migrate from Digit 4. Agility says the newer robot is designed for cooperative safety near people and a broader set of industrial workflows. The Digit 5 launch materials describe changes intended for commercial scale, but the robot still needs field proof. A generational transition can improve reliability and task coverage, or it can introduce new faults into a stable deployment. Competitors will compare their own demonstrations with Agility’s operating record, while customers will care about uptime and service response. Procurement teams will also compare warranty, maintenance, and replacement terms across vendors. The migration gives Agility a chance to show that accumulated data improves a new product rather than remaining tied to an older fleet.

Hidden Insight: The Product Is the Deployment Recipe

The non-obvious asset may be the deployment recipe rather than the biped itself. Every hour in a live warehouse produces information about floor conditions, tote placement, worker behavior, charging, network gaps, and failure recovery. Agility can convert those observations into site surveys, default configurations, simulation cases, operator training, and remote diagnostics. A competitor can copy a visible mechanical feature, but it cannot instantly copy years of exceptions encountered in production. This advantage lasts only if Agility captures the experience in software and process. If knowledge stays with individual field engineers, expansion will require staffing to rise with every site. The company’s commercial value depends on turning local fixes into reusable product behavior so each new building takes fewer days, fewer interventions, and less custom code.

GXO also gains an asset from the relationship. A logistics operator that learns where humanoids work can redesign processes before the technology becomes common. It can identify which customer contracts, building layouts, and labor patterns create the best returns. That knowledge can improve bids and capital plans even if GXO eventually buys robots from several vendors. The operator also supplies a demanding environment where Agility can find defects that laboratory tests miss. This exchange explains why early deployments can continue before public unit economics look compelling. GXO receives operational learning and a potential automation option. Agility receives data, credibility, and a route to more sites. The partnership works when both parties capture value during the learning period, not only after robots reach a theoretical end state.

The planned shift from Digit 4 to Digit 5 creates a test of software continuity. Customers expect a new robot to inherit task knowledge, fleet tools, safety procedures, and integrations from the old one. If each hardware generation requires a new deployment, the product cycle becomes costly for both vendor and buyer. Agility needs stable interfaces between warehouse systems, task orchestration, and robot control so improvements below that line do not break operations above it. This is similar to enterprise computing, where customers tolerate hardware changes when applications and management tools keep working. A smooth migration would show that Digit is becoming a platform with durable operational interfaces. A difficult migration would reveal that the current success depends on generation-specific engineering.

Multi-site operation will also expose fleet economics that a single showcase can hide. Spare parts can be pooled across locations, remote support can serve several teams, and training materials can be reused. Those effects lower cost as deployments spread. The opposite can happen if failures vary by building or travel consumes field-engineering time. Agility must decide how much inventory, service staff, and monitoring to place near customers before fleet size supports it. Robots-as-a-service shifts more of that operational burden to the vendor, which can speed adoption but also concentrates maintenance and financing risk on Agility. The company’s planned public listing raises pressure to show growth, yet disciplined expansion matters more than raw site count. Each new facility should improve the service system for the next one.

What to Watch Next

Within 30 days, watch for contract and investor disclosures that add unit counts, order value, deployment fees, or service terms. Agility’s public-company process may produce more detail than a private robotics vendor usually shares. GXO may explain which workflows at Statesville and Lebanon will use Digit and how it will measure success. The most useful baseline would include current labor method, expected tote throughput, operating hours, and intervention rules. Without a baseline, later performance claims will be hard to assess. Also watch whether the companies identify safety boundaries for Digit 4 before the planned Digit 5 migration. The difference between fenced, separated, and cooperative operation affects both workflow value and regulatory expectations. Insurance and worker-training requirements would provide another concrete measure of deployment complexity.

Within 90 days, early site preparation should reveal how portable the deployment is. Look for changes to floors, racks, tote presentation, charging areas, wireless coverage, and warehouse management software. A short installation with limited physical modification would support the humanoid value proposition. Extensive redesign would weaken it. Agility should also show how it trains GXO staff, handles exceptions, and supports robots remotely. Competitors may announce broader task demonstrations, but the relevant comparison is paid uptime on a defined process. If Agility publishes productive hours, interventions per shift, or maintenance intervals, buyers can begin to model cost. Energy use and spare-parts consumption would make that model more complete. If reporting stays at cumulative hours and tote counts, uncertainty around operations will remain high.

Within 180 days, the two sites should be close to launch or in early operation. Watch whether both begin on schedule, how many Digit 4 units arrive, and when Digit 5 becomes available. A clean migration with stable warehouse integrations would validate Agility’s platform claim. Delays, narrow operating windows, or heavy on-site support would show that expansion remains engineering-led. GXO’s decision to add more workflows or facilities would be the strongest commercial signal because it reflects internal performance data that outsiders cannot see. The market should also compare Agility with conventional automation bids for the same work. Humanoids do not need to beat every robot. They need to win enough variable, human-designed tasks at a cost and reliability level that supports repeated orders.

Screenshot this: Digit’s real product test is not walking through a warehouse, but repeating one paid deployment across buildings without repeating all the engineering.