RESEARCH MEMO · Robotics & Autonomous Driving
Agility Digit 5 Enters Auditable Operations
Robotics & Autonomous Driving | Research Memo | Commercial Evidence Strengthened
Named customers, 65,000-plus field hours, 100,000 tote moves and roughly 98% task accuracy move Digit 5 beyond product demonstration—while more than $300 million of conditional orders still is not revenue.
Cumulative real-world operation
GXO tote moves at roughly 98% accuracy
Orders subject to contractual milestones
Thesis strengthened | Operating evidence upgraded | Scale revenue and unit economics unproven
01 · RESEARCH BRIEF
The one-minute brief
Agility disclosed named customers, cumulative operating hours, task volume, accuracy, safety architecture and multi-year orders in one evidence set. That materially improves the quality of humanoid-commercialization evidence and raises our research priority. But orders remain milestone-dependent, 2027 delivery is a plan and factory capacity is only design capacity. Acceptance, revenue, utilization, maintenance cost and customer ROI are the next gates.
Audio transcript
Agility Digit 5 matters not because it performed another demonstration, but because the company disclosed named customers, more than sixty-five thousand field hours, one hundred thousand tote moves, roughly ninety-eight percent accuracy and more than three hundred million dollars of conditional orders. That moves humanoid robotics closer to auditable operations. Orders are still not revenue, and design capacity is not production. The next tests are delivery and acceptance, intervention rates, maintenance cost, customer returns and cash flow. Commercial evidence is materially stronger; scale profitability remains unproven.
Known facts and open questions
- Research action
- Higher research priority, not an investment conclusion
- Maturity
- 61/100 | Directional assessment
- Validated
- Named customers, field hours and repeat tasks
- Unproven
- Order conversion, scale margins and customer ROI
Auditable operations
Sustained disclosure of customers, hours, tasks and accuracy
Orders to revenue
Delivery, acceptance, expansion and renewal
Viable economics
Hourly cost, maintenance, margins and cash flow beat alternatives
02 · THESIS → EVIDENCE → UPDATE
What changed in the thesis?
Field evidence
- Prior thesis
- Humanoid demonstrations were plentiful; comparable operating data were scarce.
- New evidence
- Digit has logged more than 65,000 hours and completed 100,000 GXO tote moves at roughly 98% on-task accuracy.
- Updated view
- Field usability improves and Digit moves into priority research coverage.
System capability
- Prior thesis
- A single robot action does not prove safe, full-shift operation.
- New evidence
- Digit 5 adds an independent safety controller, human detection, Agility Arc and NVIDIA Halos with IGX Thor.
- Updated view
- Competition expands from the robot body to safety, orchestration, charging and fleet operations.
Commercial conversion
- Prior thesis
- Orders, capacity and delivery plans must remain separate from revenue and cash flow.
- New evidence
- Orders exceed $300 million but are milestone-dependent; 2027 deliveries and 10,000-unit annual capacity remain plans or design capability.
- Updated view
- Commercial evidence improves, but scale revenue and profit do not yet warrant an upgrade.
03 · EVIDENCE & ANALYSIS
Evidence and analysis
01｜What happened
Agility introduced Digit 5 and named customers including GXO, Schaeffler, Amazon and Toyota Motor Manufacturing Canada. It reported more than $300 million of orders as of May 2026 alongside field-operation, task and safety data.
02｜Why the evidence matters
A demonstration, an order headline or a capacity target alone does not establish commercialization. Named customers, cumulative hours, repeat tasks and accuracy appearing together bring outside researchers closer to auditing whether the robot is doing sustained work.
03｜The product is becoming an operating system
Digit 5 raises payload by 40% to 50 pounds, runs for about 90 minutes and charges in about nine minutes, with a design goal above 20 productive hours per day. The value increasingly lies in whether safety, human collaboration, fleet orchestration and autonomous charging can support a full shift.
04｜What remains unproven
The $300 million order figure is not revenue; 65,000 hours do not reveal distribution by robot, human interventions or maintenance cost; 98% accuracy cannot be generalized across tasks; and 2027 delivery plus 10,000-unit capacity remain forward-looking.
04 · INVESTMENT IMPLICATIONS
Industry and asset implications
Humanoid robotics
Valuation should shift from demonstrations toward hours, tasks, failures, renewals and cost.
Enabling infrastructure
Safety compute, edge AI, fleet software and charging become more measurable picks-and-shovels layers.
Public markets
Agility is not currently a directly listed security; supplier read-throughs such as NVIDIA remain industry evidence, not automatic revenue attribution.
05 · VALIDATION & RISKS
What to verify next
Delivery and acceptance
On-time 2027 delivery with disclosed customer acceptance, expansions and renewals.
Failure signal: Delays, cancellations or missed contractual milestones.
Operating reliability
Disclosure of uptime, mean time between interventions, failure rate and maintenance time.
Failure signal: Heavy human support prevents full-shift autonomy.
Customer economics
Hourly cost and payback outperform labor or fixed automation alternatives.
Failure signal: Deployments remain small pilots without expansion.
Financial conversion
Orders become revenue with hardware, software and service margins plus manageable cash use.
Failure signal: Manufacturing investment rises without matching revenue or margins.
What would change our view?
Company disclosures include forward-looking targets and orders subject to milestones. Field hours and one customer's task data do not prove cross-industry replication. Manufacturing scale-up, reliability, service burden and cash consumption remain material constraints.
06 · FAQ
Key questions
Has Digit 5 reached scaled commercialization?
No. Field evidence and conditional orders are stronger, but scaled delivery, revenue and profit remain unproven.
Why is more than $300 million of orders not revenue?
Orders can depend on delivery, performance and acceptance milestones. Revenue follows only after performance and accounting recognition conditions are met.
What is the most important new evidence?
Operating hours, task volume, accuracy and named customers appear together—stronger evidence than a demonstration or capacity target alone.
How can public investors participate?
Agility is not currently directly listed. Supplier mappings are research leads and still require company-specific revenue and profit verification.
07 · TERMS & SOURCES
Terms, sources and related research
Key terms
- Conditional orders
- Orders that may convert only after delivery, performance or other contractual milestones are met.
- Mean time between interventions
- How long a robot operates before human help is required, a core reliability metric.
- Design capacity
- Maximum intended factory output—not actual production, sales or revenue.
- Customer ROI
- Whether labor savings or added output cover purchase, maintenance and operating costs.
This report is based on Agility Robotics' official disclosure. Orders, deliveries, capacity and productive-hour targets are conditional or forward-looking and are not recognized revenue or proof of scale profitability. For research and education only; not individualized investment advice.