Car v Cloud (Autonomy and business models)

This is not a post on cloud enabled cars or how autononous driving will use the cloud. Its a collection of thoughts on how  and when computing business model got disrupted by the cloud and similar analogies might hold for the automobile industry. A lot has been discussed on the business models of Tesla, Waymo, Uber, Lyft and perhaps what I am stating here is already observed and stated by others. I thought I will share some of my observations based on experience an EV which has autonomy as a core tech differentiator (Tesla) and economics of the EV model. This was prompted by a financial analyst session that I attended hosted by Merill Lynch on the “The future of driving” and debates with my good friends Atul Kapadia and Mike Klein.

Lots of good observations by Daniel Daniel of Merilly Lynch, but a few I will share here to set the stage for my analyses and observations of this topic. The annual automobile ownership costs and cost per mile are thus (based on 15K miles/year)

Sedan (57c) SUV (68c) Minivan (61c)

Constrating that to a Model X (75D) assuming 15K miles/year and free supercharge (unique to Tesla) works out to 69c/mile (included in that is cost of capital, electricity cost, nominal maintenace/year and insurance). So most of the cost. The model 3 by contrast (LR , rear wheel) comes to 39c/mile. My benchmark is, if I can get to say ~25c/mile, I might be open to a new model of using a car. Maybe its lower – but that is the starting point for this discussion. So what does it take to get to 25c/mile. For reference Uber and Lyft charges me >$1/mile. I would rather drive a 25c/mile than >$1/mile is its available at the click of a button from my Phone – i.e. a 4x reduction in cost of ‘shared’ transportation. That is very disruptive to current Uber/Lyft models.

A Simple EV Model is available for any or all of you to review. I have compared a Model 3 owner (LR) vs a model 3 shared (SR) vs Model X and Camry (owned and shared). Some assumption of fuel cost, maintenance and insurance costs are included as well. Feel free to copy and use it or modify to your hearts content. Simple EV model:

Current ICE cars that are lot cheaper to buy (20-30%) for similar size and features – but EVs certainly are more limited in range. ICE that have range (Toyota Camry is perhaps a good benchmark is 30+% cheaper and achieves great mileage – no wonder Toyota has postponed the EV transition).

So if you are driving 15K miles a year and keep the car for 10+ years, its better to own than use Uber. Much like if you have cloud compute with reserved instances for 3 years esp for larger instances of VM, its better to have your own infrastructure than a cloud instance. (yes, availability and other functionality is better in the cloud – but for the purpose of this – just comparing cost at the VM layer). Amazon’s infrastructure economics works when there is re-purpose and better utilization of the infrastructure.

So when does the car ownership model will be subsumed to a subscription model like cloud computing? Definitely if you are driving less than 15K miles and esp short distances where Uber or Lyft is viable and maybe less than 6 years before you replace the car. That is not the majority of US car buyers today. But some other dynamics are at play like cloud computing that could well play out here. EV cars and autonomy do add cost. EV has the other problem of battery life cycle and battery life (age). So lets assume a 100KWh battery and 80% charge per cycle (effectively 67% as you do not want to go below 20% either). With gas at $4/gallon and 27mpg for an SUV vs 40mpg for a sedan (sparing all the cost assumption details), it seems like if a car or car service can have

  1. Share the capital spend thus share the car with anbody or a few (different models will appear)
  2. Full autonomy so it can go from Point A to B and thus enable sharing
  3. Simple access (like the model 3 iphone or card access) – make it impersonal (don’t leave your personal stuff in the car)
  4. Be ‘usable’ for 2-3x mileage than normal driving – e.g. 480K miles in 8 years i.e. better use of capital spend.
  5. Available at points of use or within walking distance say 250 meters or comes to you on call.
  6. Sufficient energy for daily driving distance and nightly or daily charge up.

Then the extra cost of an EV and autonomy can be justified. Simple math shows at $50K (pre-tax) model with 75KWh battery with 450Kmiles driven over an 8 year period (125 miles per day) will cost 26c/mile. The SUV economics are different – as the transport and passenger profile are different (like needing 256 GB VM machines vs 4TB VM machines – usage model is different to justify that cost).

So the challenges or in other words oppotunity is to have a $50K car with Autopilot and ideally 100KWh battery. Combine that with simple to walk in and drive (model 3 already achieves including driver personalization), will use autopilot to primarily autonomously let a car drive from person A to B. The individual use of Autopilot is a whole different matter frought with different acceptance profile, legal etc.

Of all the car companies, seems like Tesla is further along the key parameters above. If they can be achieved, I would be open to not have my own Tesla and go for a shared car managed by the cloud at 25c/mile vs my current >50c/mile. That is $300/month of ‘auto subscription’.

So you may ask – the same model will apply to gasoline cars with autonomy. Sure. Like in cloud – I am not just buying a VM, i want the fancy new functionality. I want my driving experience to be simpler (use autonomy in stop and go traffic), like the performance characteristics (acceleration etc when I need it). I have also made everything more efficient (expecting electric cars to have longer mechnamical life) except for the battery.

Why is this not the future of Tesla network or Uber or Lyft. I think the nuance is a slightly different business model. Tesla network could – but who will pay for the upfront capital and where will be the fleet park (lot of cars). For Uber or Lyft, they solved the problem with Taxi service alternative, but its not clear they have addressed the future business model. As Mike Volpi says here – the treat the car OEMs as ‘metal benders’ like the cloud companies did initially with Dell or HPEs and out of which ODM model formed. They can morph to a new model – but even with them, who is going to pay for the upfront capital and parking (nightly or other gaps in the day). You need a lot of cars to be paid by somebody to make this work. That is how the cloud infrastructure was built out. Who has the largest fleet that is under utilized. Maybe its Hertz, Avis etc. Even for Hertz and Avis a massive fleet upgrade and associated capital cost is expensive. Maybe there is a hybrid model that is evolved from the current Turo model and a lease model.

At the end its a financial engineering that fits the most common usage model. That will be the winning business model much like AWS was able to leverage the unused capacity and technology built it out for their own use to deliver and automated, easy to consume cloud computing.

Something to think about. What if I can subscribe to 8c/mile auto service. That is > 10x of Uber or Lyft and that is $100/month charge for 15K miles. Somebody is going to solve it and either a NewCo or evolved business model from Tesla.

Love to hear your thoughts.