The reasons for autonomy that Elon outlined at the “We, Robot” event are his same talking points that he’s used over the past decade. Autonomy will be cheaper, safer, and more sustainable. It’s hard to argue against most of these benefits.

Cheaper

There are two sides of the cost benefit: hard and soft costs.

Hard costs refer to the average cost per mile, which depends on how many hours and miles a vehicle operates each day. I agree with Musk’s math that the average car is used about 10 hours per week (in the US), which equates to 9% of the week’s waking hours. If a vehicle could be used as an autonomous ridesharing driver, it’s reasonable it could be operational 75% of the waking hours (maintenance and charging happen during sleeping hours). In other words, these vehicles will have 7-8x higher utilization, which should imply a 65-75% reduction in cost per mile. While this is compelling math, the benefits will only accrue to vehicles in a fleet. As a point of reference, there are about 260m cars and light trucks on the road in the US, of which only 2m are used for Lyft and Uber.

The soft costs are all related to time savings. In my view, this represents the largest cost savings from autonomy. Being able to get 10 hours back in your week, regardless of how you use it (productivity, entertainment, or sleep), has significant value. A fun thought experiment: the average person in the US makes about $30/hour. We value our personal time greater than our work time. If you value personal time at $40/hour, that implies an extra $21k of time value per year gained by not having to drive.

Safer

Humans are becoming increasingly more distracted behind the wheel with auto related deaths increasing in each of the past 11 years (since the iPhone went mainstream) by 2.5%. The idea that autonomous vehicles are unsafe largely stems from the media’s emphasis on the few tragic accidents resulting in injuries or death. It’s important to note that all of today’s autonomous systems are in testing and will continue to improve.

More sustainable

This one is debatable given autonomous vehicles can potentially improve traffic flow by reducing human error, optimizing driving efficiency, and enhancing communication between vehicles and infrastructure. However, they could also contribute to the problem by increasing the number of vehicles on the road and miles driven. The net effect will depend on how well these vehicles are integrated into the transportation system and the ratio of shared to personal vehicles on the roads.

My long-standing optimism about Tesla’s opportunity has been tested by the painful reality that this is taking much longer than I anticipated. Patience has become a core value for investing in Tesla. The bad news is when it comes to autonomy, this is likely going to take longer than I would have thought a month ago.

Before I get into why my timing perspective has changed, I want to highlight the possibility that autonomy could arrive much sooner than I believe. My thinking is there’s a 20% chance that FSD has a “ChatGPT moment” in the next year. In theory, the way the neural networks learn, FSD is ripe for a breakthrough. The reason why I have a low probability on this outcome is I would have expected exponential progress 6 months ago. Although there’s encouraging data on miles driven for FSD v12, the trend line for disengagements remains lumpy.

What’s new in my thinking that this will take longer is related to the Robotaxi geofenced rollout, which is expected to start next year in California and Texas. While at the surface this may seem like a positive for FSD, my conversations with industry participants this week suggest the geofencing approach is an acknowledgment that it’s taking longer than expected for FSD to gain enough traction to satisfy NHTSA. The idea is if Tesla were making substantial progress, they might have skipped the geofencing exercise all together.

The counterpoint is Tesla is starting the Robotaxi fleet in California and Texas to initiate the approval process with NHTSA, gaining approval state by state. While this is a logical view, I don’t see the reason why the Robotaxi commercial path needs to follow Waymo’s. Musk’s first principle approach typically yields a unique path to solving a problem. The fact that he’s taking Waymo’s path makes me believe unsupervised FSD is still years away.

A final headwind to approval is NHTSA. It’s no secret that Tesla and NHTSA have not seen eye to eye over the past 5 years, as the company has made beta versions of FSD available to around a half million consumers. NHTSA has launched nearly 40 investigations into Tesla since 2016. To put it further into perspective, tragically, around 20 people have lost their lives using some form of Autopilot or FSD beta. Tesla would argue that in the first quarter of 2024, there was one crash for every 4.31 million miles driven with Autopilot engaged, compared to one crash for every 1.58 million miles for vehicles not using Autopilot.

I’ve modeled the US Robotaxi fleet through 2030. Here are my key base assumptions:

• The market currently sees about 4B rides annually, and I conservatively expect this to remain flat through 2030. As a comparison, Lyft and Uber’s US businesses are projected to grow ~20% this year.
• The average ride price is $22.
• Tesla will charge $16.50 per ride, offering a 25% discount.
• Tesla’s take rate will be 15%, lower than Lyft and Uber’s 25%, incentivizing third parties to add vehicles to Tesla’s fleet.
• The Robotaxi fleet will have a 90% operating margin, excluding potential sales of the Cybercab, which I estimate will plateau at 40k vehicles annually.

Offering a product priced 25% below competitors in a hyper-price sensitive market should capture significant market share. At 50% market share by 2030, this would generate approximately $4.5B in operating income, representing around 20% of estimated 2030 operating income.

Falling below the fold of the “We, Robot” event was how autonomy will impact Tesla’s business. At its core, autonomy’s biggest lever for Tesla investors is market share expansion. For every 5% increase in US market share, Tesla could sell an additional 775k cars. As a reference point, the company is projected to deliver ~2.05m cars globally next year, with ~900k in the US.

I believe only two companies in the Western world can solve autonomy: Tesla and Waymo (Google). The central unknown around Tesla’s ability to leverage full autonomy for market share is related to their appetite to license the technology to other automakers. Musk has made it clear that talks with OEMs are underway. If Tesla pursues a licensing model for FSD, my math above is largely irrelevant given the benefits of FSD will be across the industry. Tesla, of course, would make a lot of money licensing this tech. We’ll save that sensitivity analysis for another day.

Disclaimer