2026-06-18 — views

Physical AI Energy 2026 — Waymo LIDAR Power Draw vs Tesla Camera Efficiency: The AV Fleet Charging and Grid Impact Benchmark

Waymo's LIDAR draws est. 100-150W vs Tesla cameras at under 8W. AV EV trucking could cut freight CO2 by est. 75% when diesel gives way to electric.

Article 204 in the Physical AI Benchmark Series — Physical AI Energy and Climate 2026: Waymo EV Fleet Charging vs Tesla FSD Efficiency vs AV Route Optimization — The AV Energy and Grid Impact Benchmark

The energy dimension of Physical AI is commercially significant and structurally underreported. Every autonomous vehicle is a rolling electrical system — not just a drivetrain, but a sensor suite, a compute cluster, and a communications node, all drawing power continuously. The sensor architecture choices that AV companies make have direct and measurable consequences on per-mile energy consumption, fleet range, charging infrastructure requirements, and, at scale, on the electrical grid itself.

Three dynamics define the AV energy story in 2026: (1) sensor energy overhead — the measurable power draw difference between LIDAR-heavy and camera-only architectures; (2) AV driving efficiency advantage — evidence that autonomous trajectory optimization reduces per-mile energy consumption versus average human driving; and (3) fleet charging at scale — the grid demand, infrastructure investment, and V2G opportunity created by commercial AV fleets operating on predictable schedules.

The climate implications run through all three. Transportation accounts for est. 27% of US greenhouse gas emissions (EPA, 2022). The transition of both passenger robotaxis and long-haul freight trucks to electric platforms is one of the largest potential decarbonization levers in the US economy. AV technology either accelerates or complicates that transition depending on how sensor architecture, fleet charging strategy, and route optimization interact with grid infrastructure.

Section 1 — Physical AI Energy: Three Dimensions

The energy analysis of Physical AI systems requires examining three distinct dimensions that operate simultaneously.

Dimension 1: Sensor Energy Overhead

AV vehicles are EVs that carry additional electrical loads beyond the drivetrain — sensor suites that consume power continuously whether the vehicle is at highway speed or stopped at a light. The architecture choices made by different AV companies create measurable differences in this overhead.

LIDAR units — the laser-ranging sensors that form the backbone of most commercially deployed AV stacks — consume est. 15-25W each (est.). A Waymo Gen 5 Jaguar I-PACE carries 5+ LIDAR units plus cameras and radar, producing est. 100-150W of continuous sensor power draw (est.) while operating. Over a 6-hour commercial shift at highway speed, this represents est. 0.6-0.9 kWh of additional sensor energy draw (est.). For a Jaguar I-PACE with a 90 kWh battery, this represents est. 0.7-1.0% additional energy consumption attributable to sensors alone (est.).

Tesla’s camera-only FSD approach produces a starkly different sensor power profile. Eight cameras consuming est. under 1W each results in est. under 8W total camera power draw (est.) — compared to Waymo’s est. 100-150W LIDAR overhead. This difference is structural and durable: camera technology power consumption is low regardless of improvements, while LIDAR power consumption only declines as technology advances.

Dimension 2: AV Driving Efficiency Advantage

The average human driver accelerates and brakes sub-optimally. Aggressive acceleration wastes energy overcoming inertia; reactive braking wastes the kinetic energy that regenerative braking could have recovered. NREL (National Renewable Energy Laboratory) research on eco-driving shows that optimized driving profiles can improve EV efficiency by est. 5-15% (est.) relative to average human driving behavior.

Waymo and Tesla FSD both use trajectory optimization that produces smoother, more energy-efficient driving: optimized acceleration profiles, anticipatory deceleration, and reduced velocity variance in stop-and-go traffic. This efficiency gain is real, measurable, and offsets some or all of Waymo’s sensor energy overhead.

Tesla FSD’s efficiency benefit operates through a specific mechanism: FSD’s smooth trajectory on familiar routes — learned from millions of miles of training data — produces driving profiles that reduce energy consumption relative to human driving. Predictive deceleration into stops optimizes regenerative braking capture. Speed consistency reduces unnecessary kinetic energy cycles.

Dimension 3: Climate Context

Transportation accounted for est. 27% of US greenhouse gas emissions (EPA, 2022). Within transportation, passenger vehicles and light trucks account for est. 57% of transportation emissions, and freight trucks account for est. 23%. The AV transition to electric platforms — both passenger robotaxis and AV freight trucks — represents a significant potential decarbonization pathway.

The caveat is grid dependency. The climate benefit of AV EVs depends on the electricity grid mix powering their charging. Charging EVs on a coal-heavy grid produces more lifecycle CO2 per mile than some hybrid vehicles. The US average electricity CO2 intensity has been declining, and California — where Waymo’s largest operations are concentrated — has one of the cleanest grids in North America (est. 98% clean energy). But the climate benefit of the AV EV transition is grid-mix dependent and varies significantly by operating location.

Waymo’s Gen 5 and Gen 6 fleets are 100% electric. Waymo Gen 5 fleet = Jaguar I-PACE (all electric). Waymo Gen 6 = purpose-built EV designed from the ground up as an electric vehicle. Waymo operates charging depots in each commercial city: in Phoenix, the fleet recharges at a covered depot in Chandler, AZ between commercial shifts. Charging infrastructure investment is an explicit per-city capital expenditure — est. $10M-$30M per city (est.) for depot construction and charging equipment.

Tesla’s fleet of 6M+ FSD-capable vehicles is predominantly electric (Model 3, Model Y, Model S, Model X) with some ICE vehicles carrying legacy Autopilot. Tesla Semi is fully electric: EPA-certified 1.7 kWh/mile at maximum payload (80,000 lbs GCW). Tesla’s planned Megacharger network for Semi (est. 1,000+ megawatt-level chargers planned by 2025-2026 est.) provides the charging infrastructure for electric long-haul freight.

Aurora’s current commercial fleet runs on diesel — Peterbilt 579 tractors with Aurora Driver software and Cummins diesel engines. Aurora’s AV technology is fuel-agnostic: it can be applied to EV freight trucks as electric Class 8 platforms become commercially available. The energy story for Aurora today is driving efficiency on diesel; the transformative energy story is the future transition to EV freight.

Section 2 — Waymo Fleet Energy: EV Charging at Commercial Scale

Energy dimension	Waymo approach	Scale implications	2028 outlook
Fleet vehicle energy consumption	Jaguar I-PACE: 90 kWh battery; EPA range est. 234 miles; with Waymo sensor overhead (est. 100-150W continuous) (est.), effective commercial range is reduced; Waymo commercial vehicle utilization: est. 16-20 hours/day (est.) with charging during off-peak hours	At 16 hours/day operation + est. 25 mph average commercial speed (urban stop-and-go), a Waymo I-PACE drives est. 400 miles/day; this requires 1-2 charging cycles per 24-hour period; Waymo charging logistics require vehicles to leave service for est. 1-2 hours per charging cycle	Gen 6 purpose-built EV design will likely optimize battery size + sensor power efficiency; expect improved range + reduced relative sensor overhead in Gen 6
Commercial charging infrastructure	Waymo maintains private charging depots in each operating city; vehicles return to depot between peak demand periods (typically late night / early morning) for charging; Waymo uses Level 2 + DC fast charging at depots	Per-city charging infrastructure: est. 200-500 charging ports per city at scale (est.) depending on fleet size; this is equivalent to a mid-size commercial EV fleet operator; Waymo’s charging infrastructure is a city-entry cost that must be incurred before commercial operations can start	Charging depot construction time + permitting is a secondary bottleneck on Waymo city expansion (in addition to mapping + regulatory); Gen 6 fleet scale increases the charging infrastructure requirement proportionally
Sensor power vs EV range tradeoff	LIDAR sensor power consumption represents a measurable energy overhead on Waymo’s EV platform; LIDAR technology is improving (solid-state LIDAR typically consumes less power than spinning mechanical LIDAR); Waymo’s custom LIDAR (Honeycomb + Laser Bear Honeycomb) has been refined over multiple generations with improved power efficiency	The sensor-EV range tradeoff is a design constraint that Waymo manages through battery sizing + charging schedule; with declining LIDAR power consumption per generation, the tradeoff improves over time	Solid-state LIDAR is expected to reduce sensor power consumption significantly vs. spinning mechanical LIDAR; Gen 6 and future Waymo vehicles likely use lower-power solid-state LIDAR
V2G opportunity	Waymo’s fleet has predictable deployment schedules (commercial hours, typically 7am-2am in most markets); during off-peak hours (2am-7am), the fleet is stationary and charging; this creates a V2G (vehicle-to-grid) opportunity where Waymo fleet batteries could discharge back to the grid during peak demand events	V2G at Waymo fleet scale: est. 2,500 vehicles x 90 kWh x 50% usable V2G capacity (remaining 50% reserved for next shift) = est. 112 MWh of V2G capacity per night (est.); this is meaningful grid balancing capacity for cities like San Francisco	Waymo has not publicly announced V2G implementation; V2G requires bidirectional charging hardware and grid interconnection agreements; a future capability but not current practice
AV driving efficiency	Waymo’s trajectory optimization produces smoother driving than average human behavior: optimized acceleration, more predictable deceleration, reduced stop-and-go velocity variance; NREL eco-driving research suggests est. 5-15% efficiency improvement from optimized driving profiles (est.)	For the Jaguar I-PACE, a 10% efficiency improvement from Waymo driving optimization would add est. 23 miles of effective range per charge (est.); this partially offsets the sensor energy overhead	As AV driving policies improve with more training data, energy efficiency of the driving trajectory is expected to improve further; Waymo’s driving is already in the efficient driving zone

Section 3 — Tesla FSD Energy: Efficiency Advantage Without Sensor Overhead

Energy dimension	Tesla FSD approach	Efficiency implications	2028 outlook
Camera-only sensor power	Tesla’s 8-camera FSD system consumes est. under 8W total sensor power draw (est.) vs Waymo’s est. 100-150W (est.); this difference in sensor power draw represents a meaningful energy efficiency advantage for Tesla’s camera-only approach at equivalent driving speed and distance	At 20 hours/day commercial operation (Tesla Robotaxi scenario), the difference in sensor power draw between Tesla FSD and Waymo LIDAR-based system is est. 0.6-0.9 kWh/day per vehicle (est.); for a Tesla Robotaxi fleet of 1,000 vehicles, this represents est. 600-900 kWh/day savings vs. a hypothetical LIDAR-based equivalent	Camera sensor power will remain low regardless of improvements; Tesla’s energy overhead advantage from sensor architecture is structural and durable
FSD driving efficiency	Tesla FSD’s trajectory optimization on trained routes produces smooth acceleration and deceleration profiles; Tesla has claimed FSD-engaged driving improves energy consumption vs. typical human driving; regenerative braking usage is optimized by FSD (predictive deceleration into stops vs. reactive braking)	Tesla FSD’s energy efficiency improvement is not separately quantified in Tesla’s public disclosures; anecdotal and community reports suggest FSD-engaged trips commonly achieve better Wh/mile than human-driven trips on the same route; the efficiency improvement is route-dependent	FSD efficiency improvement expected to increase as neural net policies incorporate energy optimization more explicitly; energy efficiency optimization is a training signal that can be directly added to FSD’s reward function
Tesla Semi energy efficiency	Tesla Semi: EPA-certified 1.7 kWh/mile at max payload (80,000 lbs GCW); diesel equivalent: est. 6-7 kWh/mile energy equivalent (at 6-7 mpg diesel, est.) (est.); Semi’s efficiency advantage: est. 75% lower energy consumption per mile compared to diesel equivalent on an energy-equivalent basis	The Tesla Semi’s energy efficiency improvement is primarily from electric drivetrain efficiency (not AV driving behavior): EV efficiency is est. 85-90% drivetrain-to-wheel vs diesel est. 40-45%; regenerative braking recovers additional energy on downhill segments; the AV driving efficiency improvement is additive on top of the EV drivetrain advantage	Tesla Semi EV efficiency advantage grows with grid decarbonization; as grid electricity carbon intensity falls, the lifecycle CO2 of Tesla Semi vs. diesel truck improves further
Route optimization efficiency	Tesla’s Navigation on Autopilot optimizes routes for time and energy; FSD uses predictive traffic data to avoid congestion stops (which waste energy on braking and re-acceleration); at the fleet scale, Tesla’s navigation optimization for energy efficiency is one of the few fleet-scale energy optimization tools available without additional infrastructure	If Tesla’s 6M+ vehicle fleet uses energy-optimized routing, the aggregate energy savings could be significant; even est. 2-3% fleet-wide energy improvement (est.) across 6M vehicles driving est. 15,000 miles/year each represents est. 1.8-2.7 TWh of saved energy per year (est.)	Fleet-scale energy optimization could enable Tesla vehicles to collectively reduce energy consumption by coordinating routes to reduce the aggregate fleet’s stop-and-go driving
Tesla Supercharger + Megacharger infrastructure	Tesla has the largest EV charging network in North America (Supercharger network for passenger vehicles; Megacharger for Semi); for a Tesla Robotaxi fleet, the Supercharger network provides charging infrastructure without requiring Waymo-style private charging depot construction	Tesla Robotaxi can leverage Tesla’s existing Supercharger network for opportunistic charging (not just depot charging); this reduces the per-city charging infrastructure capital expenditure for Tesla Robotaxi vs. Waymo	Tesla Supercharger network expansion increases Tesla Robotaxi charging flexibility; Megacharger network expansion enables Tesla Semi + future AV truck charging

Section 4 — AV Route Optimization and Grid Impact

System dimension	Current state	Energy/grid implication	Policy/infrastructure need
Braess’s paradox in AV routing	Braess’s paradox: in transportation networks, adding a new road (or optimal route) can paradoxically increase total congestion if all drivers take the newly optimal path simultaneously; AV systems all optimizing routes simultaneously could trigger convergence on the same “optimal” routes, causing those routes to become congested — defeating the optimization	Traffic simulation studies suggest that if a large fraction of vehicles are AV/connected and use the same routing algorithms, convergence on identical routes is possible; the result could be degraded traffic flow in corridors that all AVs identify as optimal while other corridors are underutilized	Traffic management centers in AV-dense cities must coordinate routing recommendations across AV fleets to prevent Braess’s paradox; V2X (V2N specifically) can enable city traffic management centers to recommend diverse route distributions
Fleet charging grid demand	At Waymo fleet scale in a single city (est. 700 vehicles in SF, est. 1,100 in Phoenix), the fleet charging demand during overnight hours is: est. 700 vehicles x 90 kWh x charging efficiency (est. 85%) = est. 74 MW peak charging demand for SF fleet alone (est.); at national AV fleet scale (est. 1M AV commercial vehicles by est. 2030) (est.), simultaneous charging creates a significant grid demand event	Grid demand from AV commercial fleet charging: if 1M AV commercial vehicles (robotaxi + AV trucks) charge simultaneously overnight, the aggregate demand is est. 90-150 GWh per night (est.); this is manageable if staggered (smart charging) but creates a problematic demand spike if simultaneous	Smart charging coordination (utilities + AV fleet operators agreeing on staggered charging schedules) is essential for grid stability at AV fleet scale; this requires utility rate structures that incentivize off-peak fleet charging
Solar + AV charging synergy	Waymo’s Phoenix fleet operates in the highest-solar-radiation market in the US (Chandler, AZ); daytime charging from solar generation is possible if Waymo’s depot includes solar panels or contracts for solar power purchase agreements	Phoenix solar + AV fleet charging: daytime off-peak charging at solar-powered depots reduces AV fleet carbon footprint and captures low-cost solar electricity; this is commercially attractive in Arizona AV operations	Solar depot + smart charging investment is incremental to Waymo’s depot capital expenditure; commercially attractive in high-solar markets (Phoenix, Las Vegas, Los Angeles)
AV trucking energy impact	Aurora’s current fleet is diesel (Peterbilt 579); the energy efficiency advantage of AV trucking with diesel trucks (vs human-driven diesel) comes only from driving efficiency optimization; when AV trucks transition to EV, the combined AV + EV efficiency advantage is significant	Tesla Semi AV (hypothetical, when FSD-for-Semi is developed): 1.7 kWh/mile x AV efficiency improvement (est. 5-10%) (est.) = est. 1.5-1.6 kWh/mile (est.); vs diesel equivalent est. 6-7 kWh/mile energy equivalent = est. 75-80% energy per mile reduction (est.); at US freight industry scale (est. 50B truck-miles/year) (est.), this is a massive energy and CO2 reduction potential	AV EV trucking as a climate policy lever: the combination of AV + EV for long-haul trucking could reduce US freight CO2 emissions by est. 40-60% (est.) vs current diesel fleet; this aligns with EPA’s Phase 3 GHG standards for heavy-duty vehicles (finalized 2024)
Humanoid robot energy	Tesla Optimus: battery est. 2.3 kWh (est.) per unit; est. 8-hour operation per charge (est.); recharging time est. 1-2 hours (est.) at factory charging stations	At 50,000 Optimus units deployed in Gigafactories (if Tesla’s 2026 target is met), the aggregate daily charging demand is: 50,000 x 2.3 kWh / 8-hour shift = est. 14 MW continuous draw (est.) during operation; plus est. 50,000 x 2.3 kWh = est. 115 MWh per day for recharging (est.)	Humanoid fleet energy is manageable at current scale but creates meaningful charging infrastructure requirements at 50K+ unit scale; Gigafactory solar + charging coordination needed

Section 5 — AV Energy Benchmark Scorecard

Energy dimension	Waymo	Tesla FSD / Robotaxi	Aurora (AV Trucking)	2028 outlook
Sensor energy overhead	High: est. 100-150W LIDAR + sensor suite (est.); meaningful per-mile energy overhead vs camera-only	Low: est. under 8W camera suite (est.); structural energy efficiency advantage over LIDAR-based systems	Medium: Aurora’s LIDAR overhead similar to Waymo’s but on a larger vehicle (diesel Class 8 where sensor power is a smaller fraction of total energy)	Solid-state LIDAR reduces Waymo’s overhead; camera systems continue low power; gap narrows but does not close
Driving efficiency improvement	High: Waymo’s trajectory optimization is among the best in the industry; smooth deceleration + anticipatory stop-avoidance; est. 5-15% efficiency improvement vs average human (est.)	High: Tesla FSD trajectory optimization produces energy-efficient driving profiles; regenerative braking optimization; est. similar efficiency improvement to Waymo (est.)	Moderate: Aurora’s AV driving optimization on highway (primarily speed/following distance optimization); est. 5-10% fuel efficiency improvement vs human driver highway average (est.)	All AV systems will improve driving efficiency as policies mature; convergence toward similar improvement percentages
Fleet charging infrastructure	Proprietary depots (capital expenditure per city); covered charging at each operating location; smart charging needed at scale	Supercharger network leverage (no depot needed for Robotaxi in Supercharger-dense markets); Megacharger for Semi; significant infrastructure advantage vs Waymo’s depot model	Not applicable currently (diesel); when transitioning to EV freight, Megacharger-scale infrastructure needed	Tesla’s charging infrastructure advantage grows; Waymo must invest in depots per city; Aurora EV trucking needs new Megacharger-equivalent freight charging network
V2G potential	High: commercial fleet with predictable off-shift scheduling is ideal V2G resource; 90 kWh x 2,500 vehicles = est. 225 MWh total fleet capacity (est.)	Very High: 6M+ vehicle fleet with V2G capability would be among the largest grid battery resources available; Tesla has V2G hardware in some markets; Robotaxi fleet V2G during off-shift could be significant	Low currently (diesel); EV freight V2G = very high long-term (large battery capacity per truck)	V2G is a 2028-2030 implementation story for all three; Tesla has largest V2G potential by fleet size
Climate / CO2 reduction	High: all-EV fleet in commercial operation; grid-powered vs gasoline (gasoline rideshare competitor); lifecycle CO2 reduction significant where grid is clean (CA: est. 98% clean energy grid for EV charging)	Very High for Semi: est. 75-80% lifecycle CO2 reduction per mile vs diesel (est.) (at US average grid); FSD passenger vehicles: lifecycle CO2 reduction vs gasoline ICE est. 40-60% at US average grid (est.)	Moderate currently (diesel AV); transformative when EV: est. 40-60% CO2 reduction per freight mile vs diesel at US average grid (est.)	AV EV trucking is the largest single CO2 reduction opportunity in US freight if Aurora/Waymo Via adopt EV platforms

Overall Verdict

The Physical AI energy story is positive for climate but more complex than a simple “EVs are clean” narrative. Tesla’s camera-only FSD architecture has a structural sensor energy efficiency advantage over Waymo’s LIDAR suite — but Waymo’s driving trajectory optimization is equivalent in quality, and the sensor overhead is manageable relative to the 90 kWh battery capacity. The most impactful energy story in Physical AI is not the robotaxi efficiency battle: it is the potential decarbonization of long-haul freight trucking.

Aurora’s current diesel fleet generates zero net CO2 benefit versus human-driven diesel — but when AV technology migrates to EV freight platforms (Tesla Semi or EV Class 8 equivalents), the combination of AV driving efficiency plus electric drivetrain efficiency represents the largest single decarbonization opportunity in US transportation since the introduction of catalytic converters. The physics are unambiguous: diesel drivetrain efficiency of est. 40-45% versus electric drivetrain efficiency of est. 85-90% is a structural advantage that no diesel efficiency improvement can overcome. AV driving optimization adds est. 5-15% on top of that.

The grid impact dimension is the caution flag. Smart charging coordination, V2G program development, and routing diversity management are not optional features for a large-scale AV deployment — they are prerequisites for a positive grid outcome. Without them, large commercial AV fleet charging creates demand spikes that stress grid infrastructure and, depending on the marginal generation source, could temporarily increase CO2 intensity per mile driven. The Physical AI energy opportunity is real. Realizing it requires infrastructure investment and policy coordination alongside the vehicle technology itself.

Note: Figures labeled “(est.)” are directional estimates based on publicly available information as of mid-2026. Sensor power consumption, vehicle energy consumption, fleet sizes, and infrastructure costs are not fully publicly disclosed by any of the companies mentioned. This article does not constitute investment advice.