Here’s the questions I’d like to see TSLA answer, but I doubt that any of the analysts will ask them:
How many new, paying (not free trial) subscribers to FSD were there this quarter? Was there an increased paying subscriber rate this quarter? How many total paying FSD subscribers are there?
Did you move the safety monitors from the robotaxis to chase cars and give them a remote kill switch?
For Optimus, what percentage of revenue do you expect to go to xAI in comparison to the revenue going to Tesla?
Any questions you’d put to the TSLA management team if you could?
Easy in quotes, because it’s of course incredibly difficult to be able to get a bipedal robot to be able to do all those things. But it’s relatively easy, because that doesn’t involve manipulating any objects in the physical world. It doesn’t require the robot brain to be trained on what anything in the environment is like, other than the floor. It’s a much simpler task (but still incredibly difficult!) than teaching the robot to perform tasks with objects in the physical world, to understand how materials like cardboard or glass or organic material react when you apply force to them (like picking up a box or a cup or an onion).
We’ve had humanoid robots that could move themselves through space for quite a while (Boston Dynamics’ videos were always fun), because that’s a job that’s modest enough that it can be hand programmed in without even using AI. I think we’d all like to see if Optimus manage to successfully perform an actual task that people regularly hire humans to do.
I’ll expand on what Albaby said. Historically robots replaced humans for simple, repetitive tasks, or tasks that humans were bad at. As robots have gotten better they have taken over more and more tasks.
But the problem with replacing humans with robots at tasks humans are good at, is that by the time you do all the engineering and development necessary to create the robot, it is simply cheaper to use the human. Again, that line moves a little bit every year, but that’s the basic issue.
Tesla wants to create a single robot that do lots of tasks that humans are good at, as opposed to replacing a single task. Solve the problem one time, in other words. Thus far, though I’ve never seen a video of Optimus replacing human labor at anything. Logic says that it will eventually, but I’m curious what the path to profitability is.
That’s why I’d ask the board what the milestones are.
It’s been said before, but a lot depends on getting a good enough hand. To be sure, there are tasks which will call for a specialized hand substitute, but a good enough hand would enable a lot and current hands look clumsy compared to humans.
I’d say bolting together 4 square miles of solar panels in space to power a data center in orbit would be cool. And maybe they could take over the maintenance and care and all that stuff, so “robots in space” can jump right in with “data centers in space” run, of course, buy “AI in space”, which we can all wave to as we go past on the trip to Mars.
I think that’s the wrong order. I don’t need a dancing robot. I want something that can put away the groceries and move the clothes from the washer to the dryer.
TSLA has a P/E of 300 in part because of the anticipation of Optimus providing a huge revenue stream. IIRC, on the Q4 2024 earnings call Musk said it would be in the order of $10 trillion dollars. That’s a lot of dough.
But it matters a lot when that revenue arrives. If it can do practical tasks next year that’s a lot different investing case than ten years from now.
So, I’d like to see more videos of Optimus working in the Tesla factory and fewer videos of Optimus’ ballet moves.
Ah, but that’s the mistake I think a lot of people make about AI. We anthropomorphize. We tend to project human levels of capacity onto the AI tools, which leads us to make significant errors in assessing what they can do.
That’s how a bunch of lawyers nearly got themselves disbarred for using AI, for example. They asked AI to draft things that they submitted to court - and then were unpleasantly shocked to find that the AI had made up some of their cases from whole cloth. Pure fabrication. The judges were not pleased.
The reason they made that mistake is because AI was able to write prose so well - it could draft a very good pleading at the level of a solid mid-level associate - and once a human reaches the point where they can do that, of course they know enough not to make up cases.
But that’s not how AI’s work. They can be incredibly good at something, but have a defect or hallucination or missing spot in their abilities or whatever.
So among humans, it’s almost impossible to think that one that has matured enough to be able to breakdance would lack the ability to, say, make a slice of toast or load a dishwasher. But for AI, it can be a very long way away - because that dancing robot doesn’t necessarily even know that something like toast or dishes exist, let alone what their properties are or how to manipulate them in physical space.
100% this. The robot moving itself is the easiest part. Again - still extremely difficult, and no shade to the various companies who have solved that very complex engineering problem. But getting the AI in the robot’s head to learn how physical objects in the real world respond to being moved and squeezed and bent and turned - that’s the sort of stuff that humans have internalized by a few years of age, and which AI doesn’t have a data set to learn off of.
So they pretty much answered my first question about FSD (from the conference call):
FSD adoption continued to improve in the quarter, reaching nearly 1,100,000 paid customers globally. Of these, nearly 70% were upfront purchases.
So only around 330,000 paid FSD subscribers so far. Way less than I thought. Elon has a long way to go to get to 10 million subscribers.
Elon also mentioned the chase cars in Austin:
I think maybe as of yesterday or so, we actually do not even have a chase car or anything like that. These are just cars with no people in them and no one following the car in Austin. We are obviously being very cautious about this because we want to have no injuries or serious accidents along the way. I think it makes sense to be very cautious.
We’ll have to wait and see if any evidence shows up on Youtube of these Austin robotaxis with no safety monitor in them and no chase car.
So, not bad, 2 out of 3 questions answered.
Here’s a link to the conference call transcript (may be on the paid side):
Robotaxi deployment exceeded 500 vehicles in service and is reportedly doubling monthly, with substantial infrastructure and service network investment to support growth.
Note that Waymo currently has about 2500 vehicles total in operation. Also note that the public “robotaxi tracker” had less than half the number of Tesla vehicles in operation.
I don’t know why people were surprised on Models S & X being EOL’d. Musk hinted at this back in 2021:
“I mean, they’re very expensive, made in low volume. To be totally frank, we’re continuing to make them more for sentimental reasons than anything else. They’re really of minor importance to the future,” Musk said.
And back in July of last year, Tesla ended European orders for the vehicles.
Converting that old assembly line space to Optimus should really crystalize w where the company is heading.
Growth in FSD outpaced the growth in cumulative vehicles sold, so FSD is becoming more popular.
Yeah, but these two things are not the same. Waymo has 2500 autonomous taxis while TSLA has 500 taxis with probably 99% of them having a human driver in them.
Again, it’s the trend that matters. It took Waymo 21 months to go from rides with safety drivers in them to no drivers, it’s taken Tesla 7 months. Waymo has expanded to only 5 cities in 5+ years, and while they’re expanding faster now, they’re still vehicle-limited. And not all of Waymo’s vehicles are driving autonomously, as you pointed out in the Waymo Echo Park crash.
At what point in time do you estimate Tesla will have more robotaxi vehicles than Waymo?
Tesla will expand rapidly in Austin. California will take more time, as regulators will slow it down, but it’s going to happen, and fairly soon. Waymo has maybe 3 years to turn a profit on its autonomous taxi business - do you think they’ll get there? And if so, at what vehicle/ride volumes?
Not sure. I’m still waiting to see how TSLA robotaxis perform when operating truly autonomously. Waymo has demonstrated a great safety record while operating autonomously while we are still waiting to see how TSLA does.
I don’t think Waymo will be profitable within 3 years but Google can afford to keep burning cash on Waymo. Waymo is projecting to reach 1 million weekly trips by the end of this year which is more than double than they had last year. So, if they double weekly trips for the next 3 years then they are looking at 4 million weekly trips by end of 2028.
