Time matters through what it accumulates
A physical-AI moat can depend on calendar time when a rival must accumulate operating evidence, qualify a process or secure constrained physical access before matching the service. Being old is not the advantage. The advantage is a valuable asset that develops through a sequence the rival cannot cheaply compress. A company must show what accumulated, why customers value it and what prevents a faster alternative from bypassing it.
Dierickx and Cool's asset-accumulation theory distinguishes resources built over time from resources available for purchase. It highlights time compression, interconnected assets and erosion as features of imitability. This is a theoretical lens, not an AI-era outcome study. Our application is a hypothesis: cheaper software development may increase the relative importance of some assets whose accumulation still depends on physical operation and observation.
To test that hypothesis, draw the rival's critical path. Which steps can run in parallel? Which depend on a prior result? Which are administrative habits that could change, and which involve a genuinely necessary observation period? A calendar estimate becomes useful only when it is tied to those dependencies. A slow incumbent process is an opportunity for a faster rival unless something valuable requires the delay.
Distinguish a demonstration from deployment proof
A robot demonstration shows performance under its demonstrated conditions. A commercial operating record must answer different questions: which tasks were attempted, which environments were encountered, how often people intervened and what happened when conditions changed. These are proposed evidence fields, not claims about an unnamed vendor. Record failed attempts alongside successful ones so a customer can assess the scope of the result.
For a hypothetical warehouse operator, a useful record might connect task attempts to site conditions, software versions, interventions and downtime. It should define a successful task before measuring it. Customer confirmation can strengthen the record, while missing periods should remain visible. A polished video cannot substitute for a denominator. Nor should a large number of operating hours obscure that every hour involved essentially the same easy task.
The barrier hypothesis concerns the cost of reproducing relevant experience, not the cost of recording time. Ask whether a rival could obtain equivalent evidence through another customer, a different deployment strategy or an accepted test protocol. If so, the lead may be shorter than the company assumes. Stronger claims explain why the experience is useful, difficult to substitute and actually available for the company to use.
Check what open learning makes reproducible
Open robotics research is a counterforce to blanket claims about proprietary experience. Open X-Embodiment assembled data from multiple robots and institutions and reported positive transfer across platforms. Its abstract supports the possibility of shared learning; it does not show that every robot can handle every industrial task. The strategic question is which part of a company's learning remains unavailable after competitors use the same public starting points.
A company can investigate that question with a baseline trained or configured using permissible public resources. Compare its result with the company's system on held-out tasks that matter to customers. Describe the environment, task distribution and human assistance. If the claimed advantage disappears, more accumulated hours are an inadequate explanation. If it persists, identify whether the cause is data, hardware, service execution or an interaction among them.
Synthetic environments may also be useful tests, but their fidelity must be demonstrated for the claim being made. A simulation result should remain labelled as simulation. Avoid turning a model's ability to propose trajectories into a statement about commissioned production performance. The moat story becomes clearer when concept evidence, controlled tests and customer operating outcomes are kept distinct and each is judged on its own terms.
Physical access can set the pace
Energy infrastructure illustrates why the physical setting matters. The IEA's 2025 Energy and AI report identifies long and complex grid queues and equipment lead times as risks to data-centre development. Those findings describe infrastructure constraints. They do not prove a particular operator owns a moat, and a power-purchase announcement alone does not establish usable capacity at the required site and time.
For a company-specific review, distinguish an expression of interest, a conditional agreement and an operating connection. Inspect location, capacity, term, dependencies and termination conditions. Ask what the service can deliver under the instrument it actually holds. The strategic object is usable access under meaningful constraints, rather than the impressive size of a planned facility or a queue position presented without context.
Every constraint also invites substitution. A rival might locate elsewhere, redesign for flexibility, use another supplier or serve a different customer segment. Test those routes before describing the resource as cornered. Physical scarcity can raise everyone's costs without benefiting the company that claims it. A barrier requires the company to possess a useful position relative to rivals, with evidence of control and a credible connection to customer benefit.
Maintain the asset after it is built
Elapsed-time assets can decay. A service record may become less relevant after a major hardware change; a skilled operating team can disperse; equipment access can expire. Review what carries forward between generations and what must be earned again. Rather than counting years since incorporation, maintain a dated map of the operating assets that still support today's offer and the assumptions that could invalidate them.
Choose evidence that makes deterioration visible as well as improvement. Examples include changing intervention rates, unresolved incidents, customer-confirmed service outcomes and the conditions attached to access rights. Protect sensitive operational details appropriately. The point is to let an authorised reviewer test the claim, not to publish every production record. A confidential claim can be documented without being described as publicly verified.
Today's Moat Index considers elapsed time and physical-AI replication within its v5.0 method. A broader next version and graph-backed recipes are coming. Use the current methodology to ask which actual asset a faster rival still needs to accumulate. Calendar time is a useful question when it reveals that asset; it is an empty slogan when it merely describes how long the company has existed.
Sources
Sources support the linked claims; our interpretation and checklists establish no outcome for your company.