There are Waymo Jaguars galore learning London at the moment, instantly recognisable by their awkward lidar hats and customary glum safety driver stuck in traffic behind the wheel. Wayve is out there too, Baidu is circling (for now?), and the main obstruction between them and my next Instagram story seems to be TFL bureaucracy (shock). Elsewhere the future is already fully operational across cities from Beijing, Seoul, Dubai, Phoenix, Austin, San Fran etc. Waymo is at 500k rides per week!
Whether Waymo’s lidar-and-HD-maps orthodoxy beats Wayve’s end-to-end vision bet is a good question, but one I have no alpha on whatsoever. As an early-stage investor I’m now more concerned by the question of what the automotive industry’s transition to autonomy teach us about every other industry that’s about to be automated? Cars are the first place we get to watch a physical AI transition happen at scale.
Enabling technologies create second order opportunities
Claude-enabled esoteric historic analogies in articles are a current pet peeve, but the past would indicate that enabling technology arrives, costs collapse, access democratises, and then entirely new institutions get built to finance, insure and service the thing.
Sail gave us the chartered monopolies like the East India Company and marine insurance at Lloyd’s. Rail enabled railroad-era e-commerce business like Sears and institutions like Moody’s as the bond market matured. And auto’s helped give us modern consumer credit as insatiable demand created the need to figure out who could actually afford one.
With history pointing to value in the boring bits beyond the enabling technology, we’ve spent a while mapping the $2–3trn services and infrastructure economy that sits around the automobile: repair shops and aftermarket parts, salvage, cleaning, parking, roadside assistance, wholesalers, auctions, dealerships, leasing, consumer finance, expense cards, loan collection, charging, telematics.
Now, what does this map look like in 10 years? And what are the key predictive levers we need an opinion on?
Well two important assumptions in our current universe are humans owning the cars and humans crashing them too.
On Ownership
I’ve seen this debate framed as binary a couple of times: either autonomy pushes us to an aircraft-style fleet model, or humans carry on valuing individual ownership and not much changes.
Obviously there will be a spectrum, and anyone who has watched enterprise AI adoption up close knows how much inertia exists. Perhaps fortunately, the limitations of the technology require marginal geo-fenced expansion over time rather than a big bang moment, but just look (and welp) at the latest FT sentiment data in London.
How we interact with these vehicles will also be determined by politics. Black cab drivers who spent years on The Knowledge are not going quietly, but more interestingly Uber is lobbying US cities on the principle that robotaxis should be required to operate on hybrid networks alongside human drivers, i.e. on Uber, so they can take a clip. A playbook mirroring the oligopoly Sabre and Amadeus built over flight distribution systems, a toll booth airlines spent decades and billions trying to dislodge
For these large fleets the key assumption underpinning economic viability is utilisation. As a proxy, if your neighbourhood doesn’t have enough demand to support Uber, I don’t think Waymo is coming any time soon, however revolutionary AVs could be for suburban parents and the elderly.
Which is why Tesla’s alternative universe is perhaps complementary: you own the car, and it earns money on the network while you sleep. Assuming you aren’t precious about a stranger being in your car, of course.
Whether AV’s get brought in fleets or individually via retail has major implications for segments like dealerships and consumer financing. But once you have an asset earning revenue there are two interesting questions: utilisation and residual value
Utilisation
A private car is used about an hour a day. AV fleets will need to target 12 hours+, else you have an expensive capital asset sitting idle. That clearly changes assumptions on maintenance intervals, claims frequency, depreciation and financing structures.
High utilisation is important to drive return on capital calculations predicating these fleets being financed. A calculation that must also calibrate for things like annual miles per vehicle, speed & miles per trip, down time (charging, cleaning, maintenance), vehicle purchase cost & useful life etc. There is a useful model to play with here.
Ryanair is not obsessed with turnaround time and maximising plane uptime for the Twitter bit. They addressed these factors through standardisation onto 737’s to kill complexity and vertically integrating into maintenance.
Interestingly it seems Waymo is doing the opposite. Their fleet uses 3 different OEMs across a mix of Jaguar I-Paces, Zeekr-built Ojai vans and Hyundai Ioniq 5s in validation. They have also anointed Moove as their preferred partner to handle robotaxi service operations.
However, the AV service market seems far from settled and several other vendors are innovating (Robodock, Aseon Labs, Terrawatt, Rocsys etc.). Interestingly Aseon report the operational overheads of Waymo’s centralised depo model consumes seven hours per vehicle per day, representing 70% of total cost of service at scale and about 45% of Waymo’s California miles are non-passenger miles. Their solution to the problem is building distributed robotic micro-depots.
The technical complexity of repairs (sensor re-calibration, LiDAR alignment, software diagnostics) also make generalist repair shops less relevant. What does the Kwikfit of this era look like? What about CCC?
We will have fewer collisions, but likely more write offs as the complexity of repair increases. Is that bullish or bearish for Copart salvaging?
In the US players like Revv have an interesting angle addressing the shortage of technician capabilities with ADAS and in China JD.com launched a ‘robot ambulance’ service in Beijing earlier in the year. It offers maintenance and repair solutions for humanoid robots, quadruped robots, AI companion robots. Their service covers basic repairs, fault diagnosis, battery replacement and recharging, testing and certification, cosmetic maintenance, and equipment recycling.
Everything is fintech
Notably, while most of the aviation industry leases its aircraft, Ryanair’s strong balance during covid enabled it to move further toward an owned model. And the aircraft leasing industry is a sleeper. Borrow cheap, buy in bulk at OEM volume discounts, lease to operators. Public players like Aercap are generating high teens return on equity. So, is there an Aercap for robotics?
Similar to the land of GPUs, residual value is the key challenge. A 737 depreciates to roughly 15% of value over 25 years and can be re-marketed to airlines around the world. Residuals so reliable that AerCap sold $1.4bn of assets last quarter at 1.7x book equity! A fleet of Waymo Zeekr RTs is worth wildly different amounts depending on whether the software licence is transferable, whether regulation impairs a key market, whether a new sensor generation obsoletes the suite, and whether the OEM keeps supporting the platform. This creates a collateral problem, something EV leasing companies have been caught offside on as the battery cost curve nuked residual values. The Waymo’s of the world can raise enough equity in the near-term to defer these questions, but my assumption is robot hardware is going to be cheap. Led by Unitree, the Chinese supply chain is doing to actuators what it did to drones, solar and EVs. But perhaps emerging innovators like Cenotian have figure it out.
EquipmentShare is an interesting player to watch. Started in 2015 as “Airbnb for construction equipment” enabling P2P rental with a 20% take rate and IPO’ed earlier this year. P2P didn’t scale (you can’t build national coverage through fragmented owner-lenders), so they pivoted to owned fleet plus physical locations. The P2P DNA survived as the OWN Program: where third-party investors own the equipment, ES operates it, investors take a revenue share when the machine rents. As of Q2 they had $9.8bn of original equipment cost under management with $5.5bn being under the OWN Program. The key to this success is their telematics platform, enabling real-time location, utilisation and maintenance data across 300k+ connected assets. Institutional investors can underwrite because there’s continuous performance data. They’ve had some issues post-IPO, but intrigued to see if they lean into robotics overtime.
And of course insurance is a classic discussion here. We saw during covid as miles collapsed, claims fell and advocacy groups found insurers took in $42bn in excess premiums in 2020 and returned only $13bn. Waymo + Swiss Re have shown across 25m+ fully autonomous miles: 88% fewer property damage claims, 92% fewer bodily injury claims vs human baselines. Already Lemonade is doing FSD-mile discounting for Tesla, a 50% cut on FSD miles. First move toward the endgame.
If not now, when?
AV’s won’t arrive everywhere at once. They’ll come geofenced, in commercially viable postcodes, borough by borough. Likewise, broader autonomy will start in workflows where throughput is not the binding constraint (high-value, low-turn processes) and where you can retrofit existing capex rather than replace it. Tractors, excavators, ship loaders etc. So timing is the question, but fortunate favors the early?
If you’re building around these themes or better informed than me, I’m on jamie@triplepoint.vc.
Some other thoughts:
- Do longer commutes become more viable? What happens to associated real estate values? Does traffic just kill that utopia? Do we need more tunnels?
- Thousands of hours are saved from commuting, do these just become more monetisable minutes for the attention economy, i.e. go long ads / ?
- 29,918 killed or seriously injured (KSI) casualties in the UK from road collisions in 2025. How much of a benefit is this to NHS capacity and A&E demand?