The merchant work that turns a machine design into repeatable production, an accepted installation and supported operation.
AS OF 2026-09-13DATED · NON-EXHAUSTIVE
Research reviewed 2026-09-13. Available means documented access or production, not assured spare capacity for every specification. Tight refers to the scope identified in the node, not every robot component. Supply unclear means the reviewed evidence does not settle availability at the selected boundary. Expansion describes how additional usable supply can be delivered, not a promised lead time. Pricing power describes our assessment of the ability to defend price or terms; it is not inferred from technical importance alone.
A working prototype is not an operating business
Production, installation and operation are different promises
A design can work once and still be difficult to manufacture consistently. A manufactured robot can function correctly and still fail to fit the customer's process. An accepted installation can subsequently need maintenance, parts and help resolving failures. Those are distinct obligations, and different suppliers may accept them.
This branch maps purchased work, not every step that engineers perform. Contract production, application integration, independent testing and lifecycle service have identifiable suppliers and customers. Tooling, work instructions, calibration, operator training and field diagnosis sit inside those scopes unless a more specific market boundary becomes economically important.
The map does not require outsourcing. A machine producer or operator can retain any of this work. Nor does an external test provider own the permission to operate: the sold service is evaluation and evidence against an agreed scope, not a universal licence for autonomy.
Repeatability decides how much growth becomes valuable
Physical AI can expand the set of tasks worth automating. The commercial result depends on how much production and deployment work must be repeated for every machine, site and update. A reusable design can spread engineering cost; a succession of bespoke installations can keep revenue tied to specialist hours.
Waymo's May 2025 account of its Magna integration plant distinguishes installing the driving system from validating and commissioning the vehicle. It also describes planned production expansion. Neither an opened building nor a stated future capacity should be treated as current accepted output.
The commonality across families is often process knowledge and a support network. It is not a single factory through which every robot passes. The same contractor can serve delivery robots and road vehicles while using different tooling, qualification and customer agreements.
Supply-chain map
From a design to a supported installation
Production supplies machines; integration fits them to work; testing provides evidence; service supports the installed base. Each external purchase is optional even when the underlying work is necessary.
Arrows show what the purchased work supports, not a compulsory sequence or verified trades between companies. Testing can occur during development, production and installation. The operating boundary is not an additional supply node.
Contract machine production
Supply unclear
Role
Builds repeatable machines or integrated platforms for a customer
Expansion
Tooling, qualified inputs, trained processes and accepted production yield
Pricing power
Execution capability matters; customer concentration and utilization constrain returns
Why it exists
A machine developer may choose to own the design and customer relationship while purchasing production. A contract manufacturer supplies repeatable assembly, sourcing, process control and testing within an agreed scope. This is not the same transaction as selling a complete branded robot to an end user.
Relevant businesses
Magna: Serve's April 2024 agreement assigned Magna a contract-manufacturing role for delivery robots; it is dated contractual evidence, not a current output count.
Jabil: engineering, manufacturing and assembly services for autonomous systems and robotic products.
Worth understanding
The Serve announcement separates its manufacturing purchase from a technology-licensing relationship with Magna. That is a useful warning against treating every dollar between two companies as payment for the same node. Jabil's broader capabilities likewise do not imply that every listed activity is included in every manufacturing contract.
Our inference is that the scarce resource in a ramp can be a proven production process rather than floor space. Yield, testing, component availability and design stability determine how quickly accepted units emerge. The reviewed sources do not provide comparable spare capacity or first-pass yield for Physical AI programmes, so the supply label is unclear rather than assigning a standard expansion time.
A contractor can spread manufacturing knowledge across customers, but each programme may require dedicated tooling and engineering. Utilization and contractual treatment of materials, rework and design changes can matter as much as unit volume. More supplier revenue can also reflect expensive purchased components passing through its accounts rather than greater value added.
Application engineering, installation crews and repeatable commissioning methods
Pricing power
Process knowledge and acceptance responsibility can differentiate; bespoke work limits scale
Why it exists
A customer usually needs a task performed within an existing process, not a robot standing by itself. Integration combines equipment, tooling, software and site interfaces, then demonstrates the agreed operation. The supplied unit can be a workcell, a line or a commissioned application.
The supplier overlap with contract production does not justify merging these nodes. Producing many copies of a robot and making a customer's production process work are different promises. A builder can outsource manufacture while the customer separately selects a site integrator; an integrator can install equipment produced elsewhere.
JR Automation's description starts with analyzing the customer's operations and selecting an appropriate solution. ATS's handling discussion distinguishes adding robots to an existing line from an end-to-end installation. These are process-specific purchases, not an interchangeable pool of robot deployment hours.
Our inference is that domain expertise, reusable application designs and acceptance responsibility can support attractive work. The risk is that each project consumes nearly as much engineering as the last. A large backlog alone cannot distinguish strong demand from delayed execution. This read has not established comparable staffing utilization or commissioning queues, so it does not call integration universally tight.
Produces independent assessment and evidence within an agreed scope
Expansion
Qualified engineers, laboratory capability and scheduling
Pricing power
Expertise and recognized scope matter; not ownership of a permission gate
Why it exists
A producer or integrator may need an independent organization to test and evaluate its equipment. The purchased work can include performance, electrical, mechanical, electromagnetic or software-related assessment. The deliverable is scoped evidence and, where applicable, certification services—not a guarantee that any autonomous behaviour is acceptable anywhere.
Relevant businesses
UL Solutions: robotics advisory, testing, inspection and certification services across several equipment categories.
Intertek: robot and integrated-system risk assessment, testing and compliance evaluations.
Worth understanding
UL Solutions distinguishes machine-level evaluation and machine-family services. A reusable product family can reduce repeated work compared with treating every configuration as unrelated. Conversely, a changed design or intended use can create new evaluation work; the required scope must be determined for that system.
Simulation can supply development evidence but does not eliminate all physical testing. Nor does a component's certificate automatically cover the integrated machine. The branch keeps independent evaluation separate because the provider and commercial deliverable are distinct, not because a generic approval diamond should be monetized as a supply node.
Our inference is that expertise and recognized capability can support pricing, while scheduling and qualified labour can restrict local throughput. The reviewed service descriptions do not establish a node-wide queue or price trend. The next evidence needed is test scope, turnaround, repeat work and utilization—not a count of standards mentioned in a brochure.
Maintains and restores installed machine capability
Expansion
Field coverage, spares, diagnostics and technician productivity
Pricing power
Uptime value and installed-base knowledge; contract risk can offset recurring revenue
Why it exists
An installed machine has to remain useful as parts wear, conditions change and failures occur. Lifecycle service supplies maintenance, repairs, spares and technical support under a defined arrangement. The customer may buy individual interventions, a recurring service package or a broader performance obligation.
Relevant businesses
ABB Robotics: service agreements covering options such as maintenance, onsite support, spares and labour.
ABB's service offering includes fixed-fee options for parts and labour. This makes the distinction from an equipment sale concrete: the supplier accepts an ongoing cost obligation. Recurring revenue is attractive only if the fee covers the frequency and cost of interventions.
OEMs often have useful product knowledge and access to spares, but that does not make all maintenance captive. Customers and independent providers may retain work. The relevant market also changes with geography and response commitments: a service team near one dense installed base is not spare capacity for an unrelated fleet on another continent.
Our inference is that diagnostics, reliable parts and a dense field network can improve technician productivity and retention. More operating hours can create service demand; better reliability can reduce repairs. The unclear supply label reflects unmeasured coverage and workload, not doubt that service is commercially meaningful. Fleet software supports this work, but a dashboard is not itself the labour or parts needed to restore a machine.
This branch's four nodes share engineering and operational know-how, but their economics differ. Contract production depends on repeatable output and utilization; integration on accepted process performance; testing on scoped independent work; service on the installed base and the cost of fulfilling ongoing promises.
The unresolved supply evidence is local and contract-specific. A company programme should be examined through accepted units, first-pass yield, commissioning effort, service response and repeat business. Those observations can support a stronger economic judgment than assigning a generic shortage label to the fact that deployment is difficult.
This dated map selects economically useful boundaries. Company participation does not establish a trade relationship between the named suppliers, and a product announcement is not evidence of accepted machine output.