2026-06-18 👁️ — views

AV Fleet Depot Operations — Waymo's Hidden Scaling Bottleneck vs Tesla's Asset-Light Model

Waymo depots charge, calibrate, and clean every vehicle daily — a capital-intensive ceiling Tesla's consumer model sidesteps by distributing costs to owners.

Article 106 in the Physical AI Benchmark Series — AV Fleet Depot Operations: How Waymo Charges, Maintains, and Calibrates Its Fleet, Why Depots Are a Hidden Scaling Bottleneck, and How Tesla’s Consumer Model Eliminates the Problem

When analysts discuss the autonomous vehicle ramp, the conversation typically centers on software capability — miles per disengagement, network weather performance, edge case handling in complex urban environments. What receives far less attention is the unglamorous logistics infrastructure that makes commercial AV service possible: the depot.

Every Waymo vehicle that picks up a passenger in Phoenix or San Francisco does so after spending the night at a dedicated facility where it was charged, inspected, cleaned, and validated for the next day’s service. Every sensor array that confidently perceives the world around a moving robotaxi does so because a technician at a depot verified its calibration within the past week. Every interior that a new passenger steps into has been cleaned since the previous service period ended.

Depots are the physical backbone of commercial AV fleet operations. They are also a direct ceiling on fleet scale — and therefore a ceiling on weekly rides, revenue, and the pace of geographic expansion. Understanding depot operations is essential for any serious Physical AI benchmark analysis, because it quantifies one of the most capital-intensive and time-consuming constraints on the commercial AV ramp.

Tesla’s consumer ownership model approaches this problem from a structurally different direction. When a Tesla owner’s vehicle participates in a future robotaxi network while the owner is at work, the depot functions are distributed across millions of residential garages and Tesla service centers already in existence. The comparison illuminates a fundamental architectural difference in how the two leading AV programs plan to scale.

Section 1 — What a Commercial AV Depot Does

A commercial AV depot is not a parking lot. It is a purpose-built operational facility that performs seven distinct functions on every vehicle in the fleet, on a continuous cycle, to keep that fleet deployable.

The highest labor burden by frequency is interior cleaning. A commercial AV serving back-to-back passengers in an urban environment must be cleaned multiple times per day — and in the absence of an onboard attendant, all cleaning happens at the depot between service periods. This is structurally different from a traditional taxi or rideshare, where the driver can observe and manage interior condition in real time.

Sensor calibration is the highest-stakes function from an operational integrity standpoint. A misaligned lidar or miscalibrated camera array can degrade the vehicle’s perception system in ways that are not immediately obvious during normal operation. Commercial AV operators must have robust calibration protocols and the equipment to execute them — specialized targets, measurement rigs, and certified technician sign-off — all housed at the depot.

The combination of these seven functions means that depot capacity directly determines how many vehicles a commercial AV operator can maintain in service-ready condition at any given time. You cannot have more active vehicles than your depot infrastructure can service.

Section 2 — Depot Economics for Waymo

The cost structure of AV depot operations is rarely discussed in public filings or investor presentations, but it is significant and recurring. The estimates below are based on commercial real estate benchmarks, automotive service industry labor rates, and EV charging infrastructure cost data as of mid-2026. All figures should be treated as directional estimates.

The sensor replacement cost line item deserves particular attention. Commercial AV vehicles operate in dense urban environments where minor incidents — a scrape against a curb, a shopping cart collision in a parking lot, a low-speed fender contact — are unavoidable at commercial fleet scale. Each incident that damages a lidar unit can generate a $5K–15K replacement cost, plus the labor and time to recalibrate the vehicle after replacement. At fleet scale, incident-driven sensor replacement becomes a predictable and substantial recurring expense.

Waymo’s Gen 6 vehicle design is intended partly to reduce this cost burden. By integrating sensors more deeply into the vehicle body and using a simplified sensor architecture relative to earlier generations, Gen 6 reduces both the exposure surface for sensor damage and the cost per replacement event (est.). This is a meaningful operational improvement, but it does not eliminate the category — it reduces the per-incident cost while incidents themselves remain an unavoidable feature of dense urban commercial operation.

Section 3 — Depot as a Scaling Bottleneck

The economic estimates above quantify the cost of depot operations. The strategic impact is about velocity: how fast can a commercial AV operator add vehicles to its active fleet, and what limits that rate?

The 6–18 month depot build lead time is the critical constraint. It means that even if Waymo’s software capability improves dramatically tomorrow — even if every remaining technical milestone is cleared simultaneously — the geographic ramp of its commercial service would still be gated by how fast it can identify sites, negotiate leases, build out charging infrastructure, hire and train staff, and reach operational readiness in each new city.

This is not a solvable software problem. It is a physical-world logistics problem that scales with the number of cities targeted and the capital available to build infrastructure ahead of revenue. For an operator that has spent over $10 billion to reach the current fleet size (est.), the additional capital requirement for multi-city depot infrastructure represents a meaningful constraint on expansion velocity.

The city-by-city nature of the bottleneck is particularly significant for the competitive dynamics of the AV race. Waymo cannot benefit from a national launch event the way a software product can. Each city is an independent infrastructure build, an independent permitting process, and an independent operational ramp. The first mover advantage in any given city is partly a function of who builds their depot there first.

Section 4 — Tesla’s Consumer Model: The Depot Problem Doesn’t Exist

Tesla’s approach to autonomous vehicle deployment — at least for its consumer-owned vehicle fleet — sidesteps the depot problem almost entirely. The reason is structural: when a vehicle is consumer-owned, the depot functions are distributed across the owner’s existing infrastructure and Tesla’s already-built service network.

The phrase “asset-light” is accurate for the consumer-owned vehicle scenario. When a Tesla owner opts their vehicle into the robotaxi network during working hours, they are deploying an asset they already own, charging it with infrastructure they already pay for, and maintaining it through service relationships they already have. The incremental cost to Tesla of adding one more vehicle to its robotaxi network is close to zero from a depot perspective.

This is a fundamentally different unit economics profile than Waymo’s. Waymo must spend $15K–30K per vehicle per year in depot costs before that vehicle generates a single dollar of ride revenue. Tesla, in the consumer-owned model, externalizes those costs onto vehicle owners who have already chosen to absorb them.

The critical caveat is the Cybercab. Tesla’s purpose-built robotaxi vehicle — designed without a steering wheel or pedals for human control — cannot be owned and operated by a consumer in the traditional sense. A Cybercab fleet requires dedicated cleaning, maintenance, and charging management. The moment Tesla deploys Cybercabs at scale, the depot clock starts ticking. The Cybercab is the moment Tesla’s asset-light advantage for the consumer-owned fleet stops applying, and some version of the Waymo depot problem arrives for the purpose-built fleet.

How Tesla resolves this for Cybercab — whether through distributed mobile service, hub-and-spoke depot networks, or franchise arrangements with existing service providers — will be a defining operational question for its robotaxi economics.

Section 5 — The Depot Benchmark Metric

For Physical AI benchmark purposes, depot operations can be quantified as a distinct dimension alongside miles per disengagement, fleet size, and geographic coverage. The following metrics form a depot benchmark framework.

Phoenix is the most instructive data point. It is Waymo’s most mature operational market, where the fleet has been running the longest and where the ratio of vehicles to depot capacity is highest. The depot utilization metrics from Phoenix represent the current best-case benchmark for commercial AV depot efficiency — a benchmark that will need to scale significantly to support the 100,000-vehicle fleet Waymo has discussed as a medium-term target (est.).

Reaching 100,000 vehicles at current depot density would require approximately 100–200 depot facilities across the United States (est.). At 6–18 months per depot build, and with the real estate and permitting constraints of dense metro areas, this represents a multi-year infrastructure program running in parallel with technology development. The AV ramp is not purely a software ramp. It is a simultaneous physical infrastructure ramp, and depots are the rate-limiting step.

Section 6 — Depot Operations as a Physical AI Benchmark Dimension

The Physical AI benchmark series measures the pace at which autonomous systems transition from laboratory capability to real-world commercial scale. Most benchmark dimensions focus on software and perception: how good is the system, how many miles has it logged, how does it handle edge cases?

Depot operations represent a different category of benchmark: the physical-world infrastructure required to deploy and sustain commercial AV service. It is a dimension that matters as much for long-run scaling velocity as any software improvement, and it is a dimension on which the two leading programs differ fundamentally.

Waymo is building a vertically integrated commercial AV service. Every part of that service — vehicle, software, depot, dispatch, customer interface — is owned and operated by Waymo or its Alphabet parent. This gives Waymo maximum control over service quality and operational standards, but it means that every scaling step requires Waymo capital and Waymo operational execution.

Tesla is building a platform. The consumer-owned vehicle fleet is the platform’s base layer. Each Tesla sold with FSD capability is a potential revenue-generating node in the network — one that the owner funds, charges, and maintains. The platform operator (Tesla) collects a network fee without bearing the per-vehicle depot cost. This is a structurally different business model with structurally different scaling economics.

The depot bottleneck is not the whole story of the AV race, but it is an underappreciated chapter. Understanding it clarifies why geographic expansion for commercial AV services is measured in years per city rather than weeks per city, and why Tesla’s per-vehicle economics — if FSD technology reaches the reliability threshold required for unsupervised operation — could scale much faster than any depot-dependent commercial fleet.

Note: All cost estimates, fleet size estimates, depot count estimates, lead time estimates, and operational projections in this article are directional estimates based on publicly available information and industry benchmarks as of mid-2026. Figures labeled “(est.)” should not be treated as confirmed data. This article does not constitute investment advice.

Sources

• Waymo One operations and fleet — Waymo ↗
• Tesla Supercharger network and service centers — Tesla ↗
• AV fleet operational costs — McKinsey Center for Future Mobility ↗
• Waymo Gen 6 vehicle — Waymo blog ↗
• Tesla Cybercab announcement — Tesla ↗

Tags