Tesla has been working on developing their autonomous software for the best part of a decade and tomorrow at 1PM (AEDT), the world will be introduced to the first car from Tesla without a steering wheel or pedals.
Autopilot was first discussed way back in 2013, but plans to offer a driver assist system rapidly evolved. By 2015, Tesla shipped the first version of Autopilot, but by 2016 the game was on with the company shifting focus to chase the much larger ambition of delivering fully autonomous vehicles.
This is one of the most difficult technical challenges the world has ever chased, with our roadways providing an amazingly complex set of variables to accommodate and with human lives on the line, you need to be ultra committed to hunt this down.
The reason Tesla (and others) are working towards a future where cars drive themselves is that the opportunity is so immense. Moving to a world where cars can drive themselves offers our biggest hope for making our roads safer and eliminate unnecessary injury and death each and every year.
Commercially, the opportunity is also massive, with taxis and ride sharing services like Uber faced with a new existential crisis of having a competitor that features a dramatically lower cost base. If you can eliminate the driver and add another available seat in the car for passengers, the economics of that business is virtually unrecognisable.
With Autonomous Vehicles, there is the technology cost to consider, but in Tesla’s case, that’s a computer and some cameras, a few thousand dollars, once, compared to paying humans each and every time.
Full Self Driving (Unsupervised)
Tesla’s autonomous driving efforts have been criticised for being late and while we all wish the technology was done, the process to get us here has been one of iteration. There’s been a number of approaches to the challenge and as each roadblock is found, a new approach has been found to overcome it.
In late 2020, Tesla released FSD Beta to a small group of customers for the first time. At the time this was a massive step forward, but looking back now, compared to what we have today, 3 years later, it looks ridiculous.
The Autopilot team who are responsible for the development of FSD has remained small over the years, with Tesla believing that a small team of ultra-talented individuals being able to move faster than a large team that needs meetings about meetings.
The road to autonomy hasn’t been easy, however recent versions show real progress and give real hope that we’re not too far away from it becoming a reality.
With Tesla now prepared to unveil the dedicated robotaxi vehicle (CyberCab) on October 10, that suggests they are confident the software is approaching a point where the car can exceed human capabilities and enable a driverless vehicle to operate without human intervention.
The renamed FSD (Supervised) that we see today is what we get to see publicly, but internally the software in development will be many versions ahead. Its with these versions that Tesla may be seeing dramatically improved performance that suggests to them the time is right.
In early 2019, Tesla delivered an upgraded hardware suite (known as HW3). Since then, millions of Tesla vehicles have shipped and have been told their cars will one day drive themselves. Tesla needs to deliver on this promise and we’ll be keen to see if the Robotaxi is running the same HW4 that current vehicles are shipping with. If Tesla runs with a HW5 and a different number of cameras (even different placements) that will make many nervous and raise many more questions.
Bet the company
Tesla’s unique approach to delivering autonomous cars relies on collecting video data from the fleet and training AI models on this data. The storage and compute necessary to train these models requires significant year-on-year investment, measured in the billions of dollars.
Tesla will also spend Billions on manufacturing lines for their robotaxi vehicle and it’s only the massive financial opportunity that makes that level of investment make sense.
There are really 4 key factors at play when considering Tesla Robotaxi.
Software
Tesla’s Autonomy team need to complete the software for any of this to work. While customers pay for and use FSD (Supervised) today, this won’t cut it when it comes to a robotaxi, certainly one without controls.
As we’ve seen in the past, the end can look near, but the long tail of issues raises it’s ugly head again and the timeline moves out. If this was to happen once Tesla has commenced production of robotaxi, this has the potential for thousands of vehicles that can’t be driven, to be built and parked in a lot, waiting for a software update.
Hardware
The unveiling event will show the design of the car to the world at the 10.10 event. Using timelines from previous vehicle unveils, and subsequent delivery events, we are likely still 18 months away from the first Tesla-built robotaxi from hitting the streets and delivering paid customer rides.
Tesla is known for their low-cost manufacturing which is chiefly achieved by great engineering, but also manufacturing done at scale. This means Tesla is likely to invest heavily in an aggressive production ramp, with the total addressable market of a robotaxi future, measured in the many, many millions of vehicles per year.
The cost of goods to Tesla (remember they are unlikely to sell these), will be a capital investment that will by paid off over months, then each vehicle will tip into productivity and keep delivering rides for it’s usable life.
Regulation
Even if Tesla can complete the software and the hardware puzzles they are responsible for, Tesla needs tackle the challenge of gaining regulatory approval which is done at a state-by-state basis. This challenge won’t be a new one, with Tesla working their way across the US as they create Tesla Insurance products.
It’s clear that Tesla is a business that runs on data and lots of it. We should expect Tesla to collect and deliver copious volumes of data to regulators to show the robotaxi is provably safer than humans. It’s also likely that creates a compelling case to regulators as preventing the rollout of autonomous vehicles that are safer, effectively commits that state to continue losing lives to human error.
If the challenges faced by Uber are anything to go by, then we should expect the legacy transport sector to fight back hard and the job-focused unions to put up a massive fight to prevent this from occurring.
Cost
Let’s imagine Tesla takes care of the Software, Hardware and Regulatory approval to operate a Level 4/5 robotaxi service. The cost of delivering this service has to be at a price point, and an experience level that attracts customers to download the Tesla event.
It’s clear that many Tesla enthusiasts would love to use the service, but for this to be successful, you need to convince people who’ve never heard of Tesla, to get the app, add a payment method and start using the service regularly.
There’s infrastructure costs associated with running a transport service, particularly vehicle maintenance, but being EVs, one of the big questions for 10.10 is where these Robtaxis will charge.
It’s possible that Tesla has a list of real estate to acquire and add dedicated Supercharging and Maintenance Centers in each city they roll out to. Given the density of vehicles required to service customers in just a few minutes of ordering a ride, the quantity of vehicles needing to charge is unlikely to be viable alongside existing Supercharger demand from Tesla customers.
The other aspect to cost is the revenue share for customer-owned vehicles that join the service. Musk has said previously that you would be able to enrol your vehicle in the fleet and make revenue from the rides it delivers. He detailed that owners would have control over who accessed their vehicles – just friends and family, or riders rated X stars and above.
If any of the above don’t come together, there could be serious consequences to the value of Tesla with a significant chance that delays cause capital investments to be wasted.
How do I watch the 10.10 event?
Tesla will stream the event live on X. In America, the Tesla’s 10.10 event, will start at 7:00 PM Pacific Daylight Time (PDT).
Other timezones are listed below.
- Mountain Daylight Time (MDT): 8:00 PM
- Central Daylight Time (CDT): 9:00 PM
- Eastern Daylight Time (EDT): 10:00 PM
- Brasília Time (BRT): 11:00 PM
- Coordinated Universal Time (UTC): 2:00 AM on October 11
- British Summer Time (BST): 3:00 AM on October 11
- Central European Summer Time (CEST): 4:00 AM on October 11
- Moscow Time (MSK): 5:00 AM on October 11
- India Standard Time (IST): 7:30 AM on October 11
- China Standard Time (CST): 10:00 AM on October 11
- Japan Standard Time (JST): 11:00 AM on October 11
What about Optimus?
The event is called We, Robot for a reason. Tesla definitely thinks of their vehicles as robots on wheels and their other robot with an opportunity to change the world is Optimus.
First unveiled at AI Day in August 2021, the dancing human was considered a joke by many. Since then, Tesla took the same iterative approach and in December 23, delivered Optimus Gen2.
We could get our first glimpse at Optimus Gen3 at the 10.10 event, and an insight into how the same end-to-end software that will power the robotaxi, will also power the humanoid robot for commercial and private applications.
The robotics space has exploded in 2024 and Tesla needs to show recent progress if they want to continue to be seen at as keeping pace with other competitors. Musk has suggested Optimus’ hands will get a significant upgrade, moving to 22 degrees of freedom (i.e. 3 finger joints like humans) which will add to the dexterity and task lists the robots are capable of.
Personally I’d be shocked if we don’t see Optimus and the Robotaxi (Tesla’s two robots), interacting at the event. Optmius has long been thought of as the last mile solution for delivering groceries (computing with Uber Eats or Amazon deliveries), as well as charging the car. The first applications are likely to be commercial ones in Tesla’s own Gigafactories, but there’s no denying that an Optimus Supercharger attendant would be a better way to solve the charging question, than throwing humans at the problem.
