Robotaxi depot charging: ports, uptime, and the hands-free roadmap

A robotaxi depot is a charging problem disguised as a parking lot. The vehicles run around the clock, return with low state of charge on a schedule set by demand rather than by shift, and cannot connect themselves to anything. This guide covers how to size ports and power, what uptime means when the customer is an algorithm, and how to build a depot today that can go hands-free later.

The scale is here

In August 2026, Nevada authorized roughly 7,100 robotaxis in Clark County across four operators: Tesla with 5,000, Waymo 1,000, Uber with Aviari 1,000 and Zoox 100, as tracked by The Charge Port's robotaxi tracker and reported by TechCrunch. One metro, one summer. Each authorized vehicle is a vehicle that must charge at a depot, and each depot is a site the utility was not planning to serve.

The vehicles cannot plug themselves in

The defining constraint of AV depot charging is simple and often forgotten in RFPs: robotaxis do not plug themselves in. Fleet operators handle charging and maintenance with depot staff, as Joule Labs explained in 2026 and Electrek's robotaxi guide confirms. That means every design decision runs through a human with a cable.

Three consequences follow. Port layout must minimize attendant walking distance, so pull-through bays with the port on the attendant's side beat angled parking. Session starts are batched, because one attendant plugs in vehicles as they arrive, so peak power demand is smoother than a shift-change depot but never zero. And the cost of a failed port is an attendant's time plus a vehicle's revenue, so uptime and mean time to repair are operational metrics, not contract decoration.

Size from sessions, not vehicles

A 300-vehicle robotaxi depot does not need 300 ports. Vehicles return at different times, and each session is short: an 80 kWh pack arriving at 20% and leaving at 90% takes 56 kWh, which is 25 minutes at 150 kW. Over a 20-hour operating window, one port can serve many vehicles.

The sizing sequence:

1. Energy per day. Vehicles × kWh per session × sessions per vehicle per day. For 300 vehicles at 56 kWh and 1.25 sessions, about 21,000 kWh.
2. Required power. Energy ÷ operating window ÷ efficiency. 21,000 ÷ 20 ÷ 0.92 is about 1.1 MW average, with a peak allowance for arrival clusters.
3. Ports. Sessions per day ÷ sessions per port per day. If a port can turn a vehicle every 35 minutes including plug time, that is about 34 sessions per port per 20-hour day, so 375 sessions need 11 ports at full utilization. Real depots run ports at 50 to 60% utilization to absorb arrival clusters, so 20 to 25 ports.
4. Power sharing. With dynamic allocation across ports on one block, 25 ports at 150 kW nominal do not need 3.75 MW installed; they need the average plus headroom, which a 1 MW+ block with storage provides.

The autonomous fleet calculator runs these numbers for a given port count and session profile.

Why mobile power fits AV depots

Robotaxi depots are leased, urban and temporary by design; operators expand, move and consolidate as service areas change. Permanent switchgear in a landlord's building is capital that stays behind when the depot moves. Utility timelines make it worse: first-time site electrification runs 8 to 18 months per Oxmaint's 2026 planning guide, and utility upgrades 3 to 18+ months per Fleet Rabbit's 2026 guide.

Mobile blocks with on-board storage sit on the pad, feed 150 kW ports with power sharing, recharge from whatever service exists or from on-site generation, and leave on a trailer when the lease ends. There is no trenching and no switchgear inside the building, which matters when the lease forbids both.

Uptime when the customer is an algorithm

A robotaxi fleet dispatcher does not tolerate a "port temporarily out of service" notice; it reroutes vehicles, and the depot's effective capacity drops. Procurement guidance in 2026 calls for monthly per-charger uptime of 97 to 99% with penalties and treats mean time to repair as the SLA that matters, per Charged Fleet's reliability feature and Evaisun's RFP checklist. For an AV depot, add three requirements:

• Telemetry into the dispatch system over OCPP 2.0.1 and a REST feed, so the fleet knows port status before a vehicle is routed to it.
• Response measured in hours, not business days, because the depot is open at 3 a.m.
• Spares staged regionally, because a replacement target of 72 hours is only credible if hardware exists within driving distance.

The MegaWatts SLA is published with these terms so they can be compared line by line.

The hands-free roadmap

Attendant plug-in is today's reality. The path to hands-free charging is moving, and depots built now should not block it.

Wireless charging for AV and robotaxi fleets is being developed at 7.2 to 50 kW by WiTricity, according to a 2026 MarketsandMarkets wireless EV charging insight and EV Infrastructure News' report on WiTricity's Florida R&D. Electreon acquired InductEV in March 2026, per the same MarketsandMarkets insight, consolidating inductive charging capability aimed at fleets. Robotic plug-in systems are the other path and suit higher-power DC sessions.

What to do now:

• Pull-through bays on level pads so a wireless pad or a robotic arm can be added without re-striping.
• DC power at the bay, not only at a wallbox, so higher-power automated systems have a supply.
• ISO 15118 Plug & Charge so authentication does not depend on an attendant tapping a card.
• A block controller that treats a wireless pad as another dispenser, which is a software question to ask the vendor before signing.

An example depot

A 400-robotaxi depot on 20-hour operation, sized with the public calculator, resolves to roughly 22,000 kWh a day, about 1.2 MW required, two 700 kW or one 1 MW+ block, and on the order of 25 ports at 150 kW with power sharing. See the labeled example. Real numbers for a real depot come from a Site Power Plan in 48 hours, including port layout for attendant flow and a wireless-ready pad plan.

Attendant flow, in numbers

The attendant is the throughput limit of a robotaxi depot until charging is automated, so design for the attendant first.

A plug-in takes about a minute if the port is on the correct side of a pull-through bay and the cable is on a retractor; three minutes or more if the vehicle must be repositioned or the cable dragged. Over 375 sessions a day that difference is 12 hours of attendant time. Bay layout is therefore a staffing decision.

Group ports in clusters of eight to ten so one attendant can plug in an arrival cluster without walking the whole depot. Put a status light on each port visible from the lane entrance so the attendant sees a fault before a vehicle is routed to it. And give the dispatch system the port map over the telemetry feed, so it sends vehicles to the cluster with open ports rather than to the nearest bay.

Mistakes to avoid

Sizing ports to fleet size. A 400-vehicle depot with 400 ports pays for 375 idle ports most of the day. Size to sessions and utilization.

Ignoring arrival clusters. Demand-driven fleets return in waves after commuter peaks. The power system needs storage or headroom for the wave, not just the daily average.

Permanent switchgear in a leased building. Robotaxi depots move. Keep the capital on trailers.

Treating uptime as a monthly average. A port that fails at 2 a.m. and is fixed at 9 a.m. shows, for example, 99% uptime and cost the depot its overnight wave. Require MTTR by fault class, as covered in what a written charging SLA should include.

Blocking the hands-free path. Angled bays, AC-only power at the bay and card-tap authentication all make wireless or robotic charging a rebuild rather than an add.