Latest / Elon Musk Podcast / Tesla Q4 and full year 2025 Financial Results and Q&A

Transcript

Tesla Q4 and full year 2025 Financial Results and Q&A · Elon Musk Podcast

0:00
Elon. Thanks, Travis.
0:02
So we've updated the Tesla mission to amazing abundance.
0:07
And this is, this is intended to send a message of optimism about
0:12
the future. I think we're most likely headed
0:15
to an exciting, amazing era of, of abundance.
0:21
And, and I think with the, with the advent or with the, with the
0:25
continued growth of AI and robotics, I, I think we actually
0:29
are headed to a future of universal high income, not
0:32
universal basic income, but universal high income.
0:34
I mean, there's going to be a lot of change along the way,
0:39
but, but that is, that is the, that is what I see as the most
0:42
likely outcome. So, so I think that's that, that
0:46
that makes sense to update Tesla's mission to reflect that
0:50
that goal. And, and obviously along that
0:55
way, we're going to keep improving safety, driving down
0:58
the cost of goods and, and getting people access to
1:02
anything they need without compromise and still making sure
1:06
that the environment is great, nature is great and, and people
1:09
can have whatever they want, which seems like probably the
1:12
best future. So I'm open to other ideas, but
1:15
that sounds like, it sounds like it sounds like if you could say
1:19
what is the best future you could possibly imagine, I guess
1:22
it would be that everyone can have whatever they want,
1:23
including amazing medical care and, and, but we still keep, you
1:29
know, the beauty of nature and, and, and earth.
1:32
I think that's probably the best outcome.
1:36
And, and we're, we're seeing obviously the first steps along
1:40
that way this year for Tesla, first major steps as we increase
1:46
vehicle autonomy and begin to produce optimist robots at
1:50
scale. We're, we're making very, very
1:54
big investments. So this is going to be a very
1:56
big CapEx year as Bible will will get into that.
2:00
That is deliberate because we're making big investments for an
2:04
epic future. So I think all these investments
2:08
make a lot of sense. We'll continue to make sure that
2:11
when we do spend capital, it is spent very efficiently.
2:15
But, but it's, it's a lot of things, you know, major, major
2:19
investments in, in, in batteries and the entire supply chain of
2:22
batteries. So we're also going to be
2:28
significant manufacturers of, of solar cells and, and we're
2:33
making massive investments in AI chips.
2:37
So, but I think these will all make a ton of, ton of strategic
2:41
sense. And, and then I guess I have
2:45
like 1, I guess it's not like it's not exactly, it's, it's not
2:51
exactly bad news, but it's, it's a, it's, it's, it's what we're,
2:58
we're, it's, it's time to basically bring the Model S&X
3:02
programs to, to an end with an honorable discharge because
3:10
we're really moving into a future that is based on
3:13
autonomy. And, and so if, if you're
3:16
interested in buying a Model S&X, now would be the time to
3:19
order it because we, we expect to wind down S&X production and
3:24
next quarter and, and, and basically start production of
3:28
Model S&X next quarter. We'll, we'll obviously continue
3:32
to support the Model S&X programs for as long as people
3:36
have the vehicles. But, but we're, we're going to
3:39
take the Model S&X production space in our Fremont factory and
3:43
convert that into an Optimus factory, which will with the,
3:48
with the long term goal of having a million units a year
3:53
of, of Optimus robots in the current SX space in, in Fremont.
3:59
So that is slightly sad but but but it's but it is, it is it is
4:07
time to bring the SX programs to to an end and shift shift really
4:14
it's it's part it's part of our overall shift to an autonomous
4:17
future. And my profile picture on on
4:22
exit for a few months there. The future is autonomous.
4:30
And so let's say with respect to full self driving and robo taxi
4:37
people obviously following with very close attention the
4:40
progress of FSC and, and you can experience it for yourself.
4:44
If you've got, if you've got a Tesla, you can, you notice with,
4:46
really with, with every software update, the car gets better and
4:50
better at autonomy. And, and, and we're, you know,
4:56
we, we, we, we, we were able to do our first rides with no
5:03
safety monitor in the car in Austin.
5:05
These are paid rides. So these were just sort of
5:08
randomly selected paid rides with no safety monitor.
5:12
And, and, and I think maybe as of maybe yesterday or so, we, we
5:20
actually don't, we don't even have a chase car or anything
5:22
like that. So these are just cars with no
5:25
people in them and no one's following the car in, in Austin.
5:30
So we're, we're obviously are being very cautious about this
5:34
because we want to have, we want to have no, no, no injuries or
5:38
serious accidents along the way. So I think it makes sense to be
5:41
very cautious, but you'll see the amount of autonomy increase
5:46
dramatically, I think every every month essentially.
5:50
So and, and, and then there will also be an opportunity,
5:56
something we've talked about for a long time for existing owners
6:01
of, of Teslas to add or subtract their cars to the fleet, kind of
6:07
like how air Airbnb works where you can add or subtract your
6:12
house to the, the Airbnb inventory.
6:16
And, and I think probably the value of the, the Tesla is the
6:21
sort of partial, if you people adding or subtracting the cost,
6:25
the Tesla autonomous fleet, it's probably a little underweighted,
6:30
underweighted by a lot of people because we've got millions of
6:32
cars with AI4 that can do this. So, so it's that that it might
6:41
potentially, I think it will provide an opportunity for a lot
6:46
of customers to earn more by lending their car to the fleet
6:52
then their least cost to Tesla. Yeah, which is kind of, it's
6:59
kind of like you get, you get, you get, you get in that
7:02
scenario, you basically get paid to own a Tesla.
7:04
It's it's quite a good scenario. And, and we expect to have fully
7:12
autonomous vehicles in, you know, probably, I don't know,
7:22
somewhere between 1/4 and half of the United States by the end
7:24
of the year pending regulatory approval.
7:28
You know, a big factor would be if, if there's some kind of
7:31
federal preemption for autonomous vehicles in the
7:35
absence of that, you kind of have to go on a city by city or
7:38
state by state basis. But nonetheless, we, we even if
7:42
it is city by city, state by state, we, we expect to be in,
7:48
you know, I don't know, dozens of cities, dozens of major
7:52
cities by the end of the year. With respect to energy, the,
7:59
the, the Tesla energy team has done incredible work and the,
8:04
the growth rate on that is continuing to be very strong.
8:09
And, and we're, we're, we're building more manufacturing
8:11
capacity and, and expect that, yeah.
8:15
And energy will will have very high growth for really as as far
8:19
into the future as we can imagine the, the solar, the
8:23
solar opportunity is underestimated.
8:25
We think the the, the best way to add significant capability to
8:30
the grid is or normally to the grid if let's say it's powering
8:34
AI data centers is solar and batteries on Earth and solar and
8:40
space. So that's why we're, we're,
8:42
we're going to work towards getting 100 gigawatts a year of
8:47
of solar cell production integrating across the entire
8:52
supply chain from raw materials already finished solar panels.
9:01
Maybe a bit more about Optimus. We'll, we'll probably unveil
9:06
Optimus 3 in a few months. And I think it's, it's going to
9:14
be quite surprising to people. It's an incredibly capable
9:19
robot. And as I mentioned, we are
9:23
replacing the SX line in Fremont with a million unit per million
9:28
unit per year line of Optimus now, because it is a completely
9:34
new supply chain. It's just it's that there's
9:39
really nothing from from the existing supply chains that
9:43
exist in in in Optimus. Everything is designed from
9:47
physics post principles. So that means the the normal S
9:54
curve of manufacturing ramp will be longer for optimist than it
9:58
is for products that have at least some portion of an
10:03
existing supply chain. Like when everything's new, the
10:08
the production rate will be proportionate to the least
10:12
lucky, least confident part of the entire supply chain.
10:16
And if there's 10,000 things that need to go right, it's you
10:18
know, it only takes one to to be slowed to to lag that.
10:22
But so it will be sort of a stretched out S curve.
10:26
But I'm, I'm confident that we'll get to 1,000,000 units a
10:30
year of in Fremont of Optimus 3. And, and this, this atmosphere
10:39
really will be a general purpose robot that can learn by
10:43
observing human, human behavior. So you can like demonstrate a
10:45
task or initially verbally describe a task or show it a
10:49
task even showed a video and it will be able to do that task.
10:54
So it's a, it's going to be a very capable robot.
10:59
I, I think long term Optimus will have a very significant
11:05
impact on the USGDP. Like it will actually move the
11:09
needle on USGDP significantly. So in conclusion, you know,
11:19
there's still obviously many who doubt our ambitions for creating
11:22
amazing abundance, but we're we're confident it can be done.
11:27
And we're that we're making the right moves technologically to
11:30
ensure that it does. And Tesla's obviously not never
11:35
been a company to shy away from solving some of the hardest
11:38
problems. You know, though, it's, I think
11:42
that's kind of how you build value in a company.
11:44
You solve hard problems. So you don't know how you create
11:47
value by solving easy problems. So there's, there's a lot of
11:52
hard problems that the Tesla team is going to solve, but it's
11:55
an incredibly talented, hard working team.
11:59
And I'd like to thank actually everyone at Tesla for their
12:03
their incredible hard work. And it's an order to work with
12:06
such such a talented group. So thank you to everyone who is
12:09
supporting this mission. The future is more exciting than
12:12
you can imagine. Fantastic.
12:15
Thank you so much, Elon. And next we have some remarks
12:19
from Feb off. Go ahead.
12:22
Thanks, Travis. So Q 425 was an interesting
12:26
quarter in a couple of respects. On the autos front, while in Q3
12:30
we saw a surge in US demand before the IRA consumer credit
12:34
Cliff pulling in some demand from Q4.
12:38
In other parts of the world, we saw increase in demand leading
12:41
to record deliveries in smaller countries like Malaysia, Norway,
12:46
Poland, Saudi Arabia and Taiwan, while continued strength in the
12:50
rest of AIPAC and EMEA. We therefore ended 2025 with a
12:54
bigger backlog than in recent years.
12:56
Note that none of these countries have the latest
12:59
version of FSD supervised available yet.
13:03
On the storage front, we had yet another record in terms of
13:07
deployments. I would like to thank our
13:09
customers and Tesla in continuing this momentum.
13:14
On the automotive margins front, automotive margins excluding
13:18
credits improved sequentially from 15.4 to 17.9%.
13:24
The automotive gross profit was flat sequentially despite 16%
13:28
lower deliveries primarily due to regional mix as we had
13:33
proportionately more deliveries in APAC and EMEA.
13:38
As we look to 2026 with the progress that has been made with
13:42
Autonomy, our focus is on ramping production at all our
13:45
factories. Our biggest constraint globally
13:48
continues to be on the battery pack front.
13:51
While our teams have been creative in trying to resolve
13:53
the situation by now putting 4680 cells in non structural
13:57
packs, we continue to iterate improving things from here on.
14:02
FSD adoption continued to improve in the quarter reaching
14:06
nearly 1.1 million paid customers globally.
14:11
Of these, nearly 70% were upfront purchases.
14:14
It is important to note that beginning this quarter we are
14:18
transitioning fully to a subscription based model for
14:22
FSD. Therefore, net additions to this
14:25
figure will primarily be via subscription model and in the
14:29
short term will impact automotive margins.
14:34
On the energy front, we achieved yet another record in terms of
14:39
gross profit for the quarter and ended the year with nearly 12.8
14:44
billion in revenue, a 26.6%. Year, year over year growth.
14:49
This was the result of high deployments in all regions and
14:52
continued strength and demand for both Megapack and Powerwall.
14:56
As we look at 2026, our backlog remains strong, well diversified
15:01
globally and we expect increasing deployments with the
15:05
launch of Megapack 3 and Mega Block.
15:08
However, we expect margin compression from the increased
15:12
low cost competition impacts to market from policy uncertainty
15:16
and the cost of tariffs, services and others.
15:21
Margin declined from 10.5% to 8.8% primarily from higher
15:26
employee related costs for service centres.
15:29
As we start preparing for the ramp in activity from the growth
15:33
in the fleet size, we did see a momentum in margin.
15:37
We see we did see an improvement in margin from our supercharging
15:40
business which is included within services and other.
15:43
Additionally note that our robot taxi business related costs
15:47
while not material are also included within this.
15:51
Given that this we're still in the early phase of our fleet
15:53
deployment and are still doing a lot of validation testing, the
15:57
revenue and cost per mine metrics are not meaningful to
16:00
discuss at the moment. Then on total gross margin
16:06
front, you know we ended the quarter with over 20.1%,
16:10
something which we haven't achieved for over the last two
16:13
years. This improvement came despite
16:15
the impact of lower fixed cost absorption and the impact of
16:19
tariffs which were in excess of 500 million in Q4.
16:24
Operating expenses increased sequentially primarily from
16:28
increased stock based compensation for employees and
16:32
as we started recording charges on for one operation my store
16:37
under our 2025 CEO performance award that was deemed to be
16:41
probable over the war term. Additionally, I spend on AI
16:45
related initiatives and new products like Cybercap, Semi
16:49
Optimus and Megaprac etcetera continues to be on elevated
16:54
levels and we expect this trend to continue for the full year
16:57
2026. Net income was negatively
17:02
impacted from mark to market charges on a Bitcoin holding
17:05
which depreciated 23% as compared to the last quarter and
17:10
the impact of unfavourable impact of FX primarily from our
17:14
large intercompany boards. On the free cash flow front, we
17:19
ended up at 1.4 billion. You know, we did end up CapEx
17:24
being slightly below our previous guidance of 9 billion,
17:29
but like as Elon already mentioned.
17:32
This year is going to be a huge investment year from a CapEx
17:35
perspective. And at the moment, we're
17:38
expecting that CapEx would be in excess of 20 billion.
17:42
You know, we'll be paying for six factories, namely the
17:46
refinery, LFP factory, Cyber Cab, Semi, a new mega factory,
17:52
the Optimus factory. On top of it, we'll also be
17:55
spending money for building our AI compute infrastructure and
18:01
we'll continue investing in our existing factories to build more
18:05
capacity and then you know, also the related infrastructure along
18:10
with it. And we'll also further expand
18:12
our fleet of Robotaxi and Optimus.
18:16
While this may seem a lot, we believe this is the right
18:19
strategy to position the company for the next era and we make
18:22
such investments as you know mentioned in a very capital
18:25
efficient manner. Note that this does not include
18:29
potential investments in solar cell manufacturing or our
18:33
terrafab as we're still in early phase and we plan to provide an
18:38
update in future quarters. We're starting not the next
18:43
chapter, but a new book on the progression of this company.
18:47
2026 year would be when all of this began.
18:51
While at times it feels daunting, it is going to be the
18:55
most exciting change in Tesla's history and we could not have
18:59
even dreamed of embarking on this journey without the support
19:03
of our customers and our investors.
19:06
Thanks for again showing the confidence in US and let's get
19:09
ready for a future of amazing abundance.
19:11
Thanks. Great.
19:14
Thank you very much, Vaibhav. Now we're going to head over to
19:17
investor questions. As always, we will start with
19:19
questionsfromsay.com. The first question is today.
19:24
Today, there are approximately 90 million cars sold globally
19:27
each year. Does Tesla have a view, based on
19:30
its robotaxi ambition, what this number will be in five or ten
19:34
years? And how does this impact Tesla's
19:36
EV strategy to have more models? Yeah, Thanks, Travis.
19:41
As Elon said, the future is autonomous and obviously
19:44
autonomy and cyber cab are going to change the global market size
19:46
and mix quite significantly. I think that's quite obvious.
19:49
You know general transportation is going to be better served by
19:52
autonomy as it will be safer and cheaper and over 90% of vehicle
19:55
miles travelled are with two or less passengers now, which is
19:58
why we designed cyber cab that way.
20:00
In this new autonomous market, we at Tesla have the advantage
20:03
of efficiency, cost and manufacturing at scale that
20:05
really no one else has. And we've built that over the
20:08
last decades and we believe that that segment that we are
20:11
creating will grow a million zero for year.
20:19
Just to add to what Lars said there, the it's worth the, the
20:25
the point that last made, which is that 90% of miles driven are
20:28
with one or two passengers or one or one or two occupants.
20:31
Essentially is, is a very important one because that that
20:37
implies that that cyber, the cyber cab, which is a dedicated,
20:41
you know, 2 seater or dedicated robo taxi.
20:46
It's a little confusing with the terms robo taxi and cyber cab.
20:48
Sorry about the confusion. But, and in fact, in some
20:51
states, we're not allowed to use the word cab or taxi.
20:53
So it's going to get even more strange.
20:55
It's going to be like cyber vehicle or something cyber car.
21:01
But the, the, the, the cyber cab, which is a specific vehicle
21:05
model that we're making, does not have a steering wheel or
21:07
pedals. So this is clearly, you know,
21:10
there's there's no, you know, there's no fall back mechanism
21:15
here. It's like this car either drives
21:16
itself or it does not drive and we expect to start production in
21:21
in April as as always, it's an S curve of the the production rate
21:27
is an S curve. So starts off very slowly and
21:30
then grows exponentially. Then you hit the linear and then
21:32
ultimately you know it asymptotes at what are your
21:36
target volume is so but but we we would expect over time to
21:43
make far more cyber caps than all of our other vehicles
21:49
combined. Given, given that 90% of of
21:54
distance driven or distance being distance travelled exactly
21:59
no longer driving is is is one or two people, I think it's like
22:06
80% is just one. So it would mean that long term
22:14
Cyber Cab would make several times more Cyber Cabs per year
22:19
than all of our other vehicles combined.
22:23
Great. Thank you so much.
22:26
The next question a bit related, Are there still plans to launch
22:31
new models to address different price segments and vehicle
22:34
types, which could materially expand the Tam for Tesla?
22:38
Yeah, to to further on what we were just talking about, we've
22:41
launched our least expensive models ever over the last few
22:44
months and our continued expand that those models globally.
22:48
And over the last decade we have continually brought down the
22:50
cost of our vehicles without sacrificing range performance
22:52
for premiumness. And we'll continue to do that as
22:55
Vaibhav said, investing in our factories.
22:57
But these are all trade-offs of where we spend our time and
22:59
money. And to Elon's point just now
23:01
with Cyber cap coming, we are aiming to bring that Tesla
23:04
premium ride experience to our largest market yet.
23:06
They could be 5 or 10 times our current levels of production.
23:10
This new autonomous market, you have to start thinking about us
23:13
as moving to providing transportation as a service more
23:15
than the total addressable market for the purchase vehicles
23:19
alone. Of course, we do have plans to
23:21
have robotaxis in various shapes and sizes, but obviously Cyber
23:24
Cab will be the grand majority of that volume.
23:28
Yeah, the the vast majority of, of miles travelled will be
23:31
autonomous in the future. You know, I would say probably
23:36
less than less than I'm just guessing, but probably less than
23:38
5% of miles driven will be where somebody's actually driving the
23:43
car themselves in the future, maybe as low as 1%.
23:47
Great. The next question is
23:50
historically, Tesla has spoken about gross margin per model.
23:54
Are there standalone gross margin targets for the current
23:56
models excluding the benefits for FFDFSD sales?
24:01
You know, we've talked about this with the previous two
24:03
cautions, but transportation as we know is changing and I think
24:08
we cannot keep applying the same framework from a car sales model
24:12
to the future. What we are trying to do.
24:15
So it has to be looked at it more holistically.
24:18
In autonomy software will be the.
24:20
Driver for growth. From now on and as we aim to
24:24
maximize the global feed, we've been laser focused on cogs from
24:29
our side to make sure because that is something which we
24:32
manage. So we will keep focusing on
24:35
that, but I think we we need to look at it from a different
24:38
dimension. Yeah, like this Cyber Cab, that
24:41
is the whole design of cyber cab was it was to optimize the the
24:45
fully considered cost per mile of autonomous driving.
24:51
And you, it's a different design problem than if you're trying to
24:53
design cars for people who, who will be driving versus being
24:58
driven. And, and so, so Cybercav is
25:04
like, like I said, super optimized for minimum cost per
25:08
mile and, and also for a much higher duty cycle.
25:12
So it would, it would expect Cybercav to be used, you know,
25:18
probably 50 or 60 hours a week instead of the 10 or 11 hours a
25:23
week that a, a driven vehicle is used.
25:27
So, you know, typically people will might drive their car for
25:30
an hour and a half a day on average.
25:31
So it's like 10 hours per week out of 168.
25:36
But I, but I think an autonomous vehicle is likely to be used
25:41
probably five times as, as often, which which means that
25:45
you need to design the vehicle for a much more wear and tear
25:50
per unit time and much more resilience.
25:52
It's more like a like a commercial truck that that's,
25:57
you know, continuous operational close continuous operation is,
26:03
is how you design an autonomous vehicle.
26:08
And, and so we will have larger vehicles in the cyber cap in the
26:13
future that that are designed for full autonomy.
26:15
And we've, we've, we've actually shown pictures of this and in
26:18
fact have shown prototypes. So this is not exactly a secret.
26:22
In fact, we've given people rides in them.
26:24
So, you know, we're not, we're not, we're not keeping this,
26:28
hiding this light under a bushel here.
26:30
You know, it's like we're literally saying what we're
26:32
going to do and have said what we're going to do for a while.
26:36
So, you know, I, I really, I think long term we would, we
26:39
would really the only vehicles that we'll make will be
26:43
autonomous vehicles, with the exception of the next generation
26:47
Roadster, which we're hoping to debut in April.
26:54
Hopefully it's it's going to be something out of this world.
27:00
Fantastic. The next question we
27:02
unfortunately have to skip because it's not related to to
27:05
Tesla. And we would like to remind
27:09
folks who use the same platform to to please focus these
27:13
questions on on Tesla. So with that in mind, we're
27:17
going to move on to the next question, which is what is the
27:20
current bottleneck to increased robotaxi deployment and personal
27:23
use unsupervised FSD, the safety and performance of the most
27:27
recent models? Is it the safety and performance
27:30
of the most recent models or is it people to monitor the robot,
27:33
robot taxis in car or remotely? Or is there some other blocker I
27:39
don't know? Shilk if you want to kick off on
27:41
this one. Yeah, We have scaled the robot
27:45
axis service that's available to customers over the last year in
27:48
order to just learn the scaling problems without having to like
27:53
wait for the unsupervised basically like 2, two goals. 1
27:56
is like learn as much as possible from the fleet with the
27:59
safety monitors. And secondly, we laser focus for
28:02
the engineering team to solve the unsupervised FSD problem.
28:06
I think we did both like by the end of last year we, you know,
28:10
we had a long tail of issues that we are able to turn
28:12
through. And then in the last couple of
28:15
weeks we have started our unsupervised robot taxi service
28:19
to public customers in Austin. I think some customers took
28:22
pride last week and also in service countries today without
28:25
any real car, something like that.
28:27
Separately, we did scale the fleet size in the Bay Area and
28:31
in Austin and through that we learned, you know, issues with
28:33
charging and other issues that we would have seen once we sort
28:38
of scale the unsupervised fleet. So both are happening in
28:40
parallel. A variant of the software that's
28:44
used for the Robot Active service was shipped to customers
28:48
with V-14 and customers saw a huge jump in performance, like a
28:51
lot of, you know, happy feedback from customers.
28:55
So, and since then, we've improved the software
28:57
significantly as well. And customers will continue to
29:00
see with their own software releases that the software is so
29:02
good that you know they're like screaming to remove the tire
29:06
monitoring software because they're bored inside the car too
29:10
much. And to that a little bit with
29:14
what Ashok said about learning about our charging and service
29:17
needs, you know we're using our vast network of charging and
29:20
service centers that really only Tesla has in this space to jump
29:22
start our infrastructure build out needs to get ahead of
29:25
robotaxi autonomous vehicle demand.
29:27
And we excite that because of this network, we are the only
29:29
company capable of scaling at the rate that is needed for the
29:33
tsunami of autonomy that is coming.
29:35
Yeah. Great.
29:39
Moving on to the next question. After the unveil of the cyber
29:42
truck, Elon stated that if it didn't sell well, Tesla would
29:46
build a more conventional looking pickup.
29:48
How practical would it be to create this new design on the
29:51
cyber truck architecture and could it be conveniently built
29:54
on the existing production lines?
29:57
Actually, in its segment, Cyber Truck continues to be a leader
30:00
and is selling more than any other electric truck, electric
30:03
truck out there while our competition continues to pull
30:06
back. But to the question itself, from
30:08
a line standpoint, we always design our lines to be super
30:11
flexible. We built 3 and Y on the same
30:12
line. We built S&X on the same line,
30:14
still showing that we can do that.
30:17
The Cybertruck line was designed in the same way and is one of
30:19
our most fully ready for autonomy platforms.
30:24
Yeah. But yeah, we're, we're we will
30:28
transition the Cybertruck line to just a fully autonomous line.
30:33
And, and there's, there's obviously a market there for
30:36
cargo delivery, like you say, like localized cargo delivery
30:39
within, within a city, within a few 100 miles, something like
30:42
that. There's, there's a pretty,
30:45
there's a lot of cargo that needs to move locally within a
30:48
city and, and an autonomous cyber truck could be very useful
30:52
for that. Great.
30:56
Moving on to the next question regarding Optimist, could you
30:59
share the current number of units deployed in Tesla
31:02
factories and actively performing production tasks?
31:05
What specific roles or operations are they handling,
31:07
and how has their integration impacted factory efficiency or
31:10
output? Yeah, we're still very much at
31:15
the early stages of of optimists.
31:17
It's a, it's an R&D just on the R&D phase.
31:21
So we, we have had optimists do some basic tasks in the factory.
31:28
But, but, but as we iterate our new versions of Optimus, we, we
31:33
deprecate the old versions and, and so it's, it's, it's not, I
31:40
wouldn't say it's like, it's not, it's not, it's not in, in
31:44
usage enough factors in a material way.
31:46
It's more so that the robot can learn.
31:52
We, we wouldn't expect to have, you know, any kind of
31:56
significant Optimus production volume until probably the end of
32:00
this year. Great adding to like Optimus
32:05
Gen. 3 is an awesome robot that. It is awesome.
32:09
Yeah, it's an awesome robot that minimizes any differences, but
32:11
it's basically it looks like a human.
32:13
People could be easily confused that it's a human.
32:16
And this helps our strategy for the a, a too, because you can
32:19
learn from how humans do their do these tasks and it's a very
32:23
easy to teach the robot do in the same way as opposed to
32:26
previous robots. Yeah.
32:28
I mean, I guess one thing I should say like is, you know,
32:31
there's a lot of news of like, you know, various companies
32:35
announcing layoffs and whatnot. But you know, at our Tesla, you
32:40
know, factory in Fremont, we actually expect to increase
32:42
headcount over time and, and to significantly increase output
32:49
from from our factories. So, you know, we don't have any
32:53
layoff plans. We expect to actually increase
32:56
headcount. Great.
33:00
The next question, similar to the other autonomy questions but
33:04
slightly different. When is FSD going to be 100%?
33:08
Unsupervised, well it is 100% unsupervised in FSC is 100%.
33:14
I mean, we obviously have cars operating with no one in them
33:19
and no safety monitor and no follow car or anything like that
33:22
in Austin right now for, for customers.
33:26
We, you know, we're, we're being just very cautious with the
33:29
rollout. I mean, we, with each successive
33:33
version as we prove it out and we, we make sure that there are
33:36
no sort of unique issues in, you know, in, in particular cities.
33:41
Because like sometimes you get like some very, you know,
33:45
difficult intersection and it'll be an intersection where a lot
33:48
of humans have accidents. By the way, there's like some
33:50
some pretty nutty intersections where it worked with a lot of
33:54
humans make mistakes and have accidents in in various cities.
33:57
So we want to make sure that, you know, FSD can handle those
34:03
unusual intersections. Like if you take LA, for
34:08
example, where where Wilshire and Santa Monica combine is, is
34:11
like there's about, I don't know, 20 traffic lights and
34:15
people are constantly having accidents there.
34:19
So you want to make sure that FSD can handle unique things in
34:24
a particular city. So and then we're actually just
34:29
being paranoid about safety. But with each successful release
34:33
of FSD, we will reduce the amount of driver monitoring
34:38
that's needed proportionate to the safety of the FSD build.
34:46
Great. As it relates to Robo Taxi, what
34:50
has surprised you about the roll out so far?
34:53
We've talked about what's constrained the the fleet
34:56
expansion to date, but it it appears there are 200 vehicles
34:59
based on public tracking. Is that something that we can
35:01
confirm? I wouldn't say there's anything
35:06
that like really surprised us because because we are a large
35:08
fleet, we had all the metrics. So there was not sort sort of a
35:11
surprise. It was just continued work to
35:13
grind down on the long pair of issues, and that's what enabled
35:16
us to launch the unsupervised service in Austin.
35:22
Yeah. And, and I mean in, in, in terms
35:25
of robotaxi vehicles carrying paid customers.
35:32
I, I, I think we're well over 500 at this point between the
35:36
Bay Area and Austin. Yeah, there's, yeah, we're
35:40
varying amounts of like vehicles depending on the load.
35:42
But yeah, you can have like more vehicles during the peak times
35:45
and then a fewer vehicles in the off hours.
35:47
Yeah, this will probably, you know, double every month type of
35:52
thing. It's going to it's it's it's on
35:53
an exponential curve. I mean, one other thing people
35:59
forget that, you know, we've been deliberate on all this in
36:02
the sense that we have the supporting infrastructure
36:05
already been in place, whether it's service centers, charging,
36:09
yes, we'll have to augment as the fleet grows depending upon
36:13
the density of where the demand is and whatnot.
36:17
But it's not something like we just stumble upon it and we're
36:20
starting to. We've been at it for years.
36:24
Yes, not every city is designed the same way, same thing.
36:28
Our infrastructure is also not the same in every city.
36:31
But you have to give us credit that it's been a journey.
36:36
And like last said, if there's some company which can do it,
36:41
we've already been at it. So we should be able to deliver
36:43
much better. Great.
36:48
The next question is about chase cars, which we already covered.
36:51
So moving on to the last question, Elon, you've been
36:55
spending significant personal time on Tesla's chip design.
36:58
Yeah. What was the forcing function
37:00
behind this increased involvement?
37:01
And do you think external chip sales will represent a
37:04
significant portion of Tesla's valuation by the end of the
37:08
decade? Well, I mean, the, I tend to
37:12
spend time on wherever, whatever the most critical issue is for
37:15
the company and, and completing the AI5 chip design and having
37:20
it be a great chip is it's arguably number the number one
37:25
most critical thing to get done, which is why I'm spending more
37:27
time on that than currently anything else at, at Tesla spend
37:32
pretty much every Saturday on on this and, and every and, and a
37:38
chunk of every Tuesday. So it's like a, you know, if I'm
37:42
spending my Saturdays on something, it's, it's going to
37:45
be something pretty important. I do think AF5 will be a very
37:50
good trip and I'm I feel quite confident about the design at
37:56
this point. And then AI6 which will follow
37:59
that it'll be aspirationally would follow that in under a
38:04
year will will be yet another big leap beyond AI 5.
38:08
So I feel, I feel pretty good about our job strategy right now
38:14
And, but, but in terms of selling it outside of Tesla, we
38:19
first need to make sure we have enough chips for the all of our
38:22
vehicle production and all of our Optimus production and, and
38:27
then, and then we will actually use the AI5 chips in our data
38:31
centers. We already used the AI4 chips in
38:34
our data centers. So when we do training, it's a
38:36
combination of the AI4 chips and NVIDIA hardware primarily that
38:43
we do training with. So, but he said by the end of
38:49
the decade, I mean, it's, it's like things are changing so fast
38:53
that it's hard to imagine like what happens at the end of the
38:55
decade. I mean, we, we, we might, I
39:02
mean, the, when I, when I look ahead at say what, what's the
39:07
limiting factor for Tesla growth?
39:09
If you go say 3 or 4 years out, I think it actually is chip
39:14
production. Is there enough AI, OK, enough
39:19
AI logic and enough AI and enough memory, enough RAM for,
39:27
for our volume? And, and, and, and right now, I
39:32
see that as being the, the thing that probably limits our, our
39:37
growth in, in three or four years, which would imply that
39:41
we're, we're not selling chips outside of Tesla because we need
39:45
them. And in, in fact, I think I, I, I
39:54
think it's going to make sense. And, and this, this is
39:56
definitely going to be a sort of a controversial thing, but I
40:01
think Tesla needs to build a terrafab.
40:05
Did I mention this at the childhood meeting?
40:10
But but even when even when we look at the output of the best
40:14
case output of all of our key suppliers, and I'd say even
40:18
they're beyond suppliers are like strategic partners like
40:21
Samsung, TSMC and Micron, the and and we say like what's the
40:29
most you could possibly make then it's, it's not enough.
40:34
So, so we, I think in order to remove the constraint, the, the
40:40
probable constraint in three or four years, we, we're going to
40:44
have to build a Tesla Terafab, a very big fab that includes
40:51
logic, memory and packaging domestically.
40:57
And and that's, that's actually also going to be very important
41:00
to ensure that we are protected against any geopolitical risks.
41:07
I think people may be under weighting the, some of the
41:10
geopolitical risks that are, are going to be a major factor in
41:15
the in, in a, in a few years. So now, you know, a lot of
41:22
people are like this will say like, that's crazy.
41:24
Favs are really hard. I'm like, yes, I know favs are
41:27
really hard. I don't think they're easy, but.
41:33
We do. We do a lot of hard things.
41:35
You know, we didn't used to have car factories that we didn't
41:37
used to have battery cell factories or lithium refineries
41:42
or, you know, mega pack factories or, you know, all
41:46
these other things. We, we, we figured it out.
41:50
So, so I think it's, I think, I think we, if we don't do the
41:56
Tesla Terrafab, we're, we're going to be limited by supplier
42:01
output of, of, of chips and, and I think maybe memory is an even
42:07
bigger limited than than AI logic.
42:12
So you know, for example, we have, we have chip supply deals
42:18
with TSMC in Arizona and Samsung in Texas.
42:22
But, but currently there are no advanced memory fabs at scale in
42:27
the United States. There are 0, literally 0.
42:31
Hopefully you know Micron will have something going in a few
42:36
years because they are headquartered in Idaho.
42:38
You know whether they make a lot of potato chips, but then we
42:43
need to make computer chips too. So anyway, we're working, we're
42:51
working with our strategic partners on the chip front,
42:53
memory and logic. But I think, I think we, we,
42:59
we've got to also try, try our hand at building a large scale
43:04
fab that integrates logic, memory and and packaging.
43:10
And if we don't do that, we're just going to be fundamentally
43:12
limited by supply chain, especially if there's some Geo
43:15
Geo, if in a worst case geopolitical situation, it would
43:19
be quite, quite a severe situation.
43:20
So I think we, we quite frankly would be crazy not to try out
43:26
the Terra Terrafab. So yeah, right.
43:33
We'll have, we'll have, we'll have a bigger announcement on
43:35
this in the future. Awesome.
43:39
With that, we're going to move on to analyst questions.
43:42
The first analyst is Emmanuel from Wolf Research.
43:47
Emmanuel, please feel free to unmute yourself.
43:53
Great. Thank you so much.
43:54
It's Emmanuel Rosner from Wolf Research.
43:56
My first question is on the on the CapEx, you signal a pretty
44:00
large increase to over $20 billion for this year.
44:05
Was hoping to better understand where the investments are going.
44:08
Anyways, 2 dimension for us, which of the product line or
44:12
technologies account for the bulk of the increase?
44:14
And also do you view this as like one time in nature 2026 or
44:20
I guess how much of this is an ongoing level of high spending
44:24
for, for a number of years? And then just finally still on
44:27
that, with that level of spending you're going to be
44:30
burning cash. How what should we think about
44:33
the cash balance or any other way to to finance this?
44:37
Yeah. So, you know, Emmanuel, I tried
44:39
to put this in my opening remarks too, but I'll try and go
44:43
a little bit deeper. There's about 6 factories, which
44:48
we are starting production in this year.
44:52
So there's a lot of cash CapEx, which is going into that.
44:57
Then as we are trying to scale Optimus, we need a lot more
45:02
compute, so we're putting more money towards compute as well.
45:06
And then for training for. Training and then we're also
45:10
going to be spending money to expand the capacity at, you
45:15
know, our existing factories on top of it, you know, we're just
45:23
keep in mind that we're not none of these numbers which I shared
45:27
of of 20 million factors in anything to do with the solar
45:32
fab or the, the semiconductor chip fab.
45:36
Those would be, as Elon mentioned would come later on.
45:41
And you, you think your second part of your question was, is
45:46
this one off or would we expect more?
45:51
I think we're getting into this investment phase because we have
45:55
big aspirations. And when you look at it, some of
45:59
these aspirations are, I call them as infrastructure play,
46:03
especially if you have to do a chip fab and we have to do a
46:07
solar cell manufacturing fab. Those are infrastructure plays
46:11
and they're that that fun that funding takes a little bit
46:15
longer and you, you would be in an investment cycle for a little
46:19
bit longer initially. But third part of your question
46:24
was how are we going to fund it initially?
46:27
Obviously we have over 44 billion of cash and investments
46:31
on the books. So we'll use our internal
46:33
resources, but there are ways where we can fund it, especially
46:38
when we look at the robo taxi fleet because you know, anytime,
46:42
anytime you have a consistent stream of cash flow, you can go
46:48
and get money from the banks. And we have had conversations
46:52
with banks about it and that is something how we're going to do
46:56
it. And then on the infrastructure
46:58
play side, yeah, like I said, we don't have a number yet, but
47:05
given that it's it's an infrastructure play, it's a
47:08
longer tail, we will have to look at a little.
47:11
Bit more in terms of how we fund it.
47:14
Whether it's through more debt or other means.
47:21
Great. Our next question comes from
47:24
Andrew from Morgan Stanley. Andrew, please feel free to
47:28
unmute yourself. Great, Thanks so much for for
47:31
taking the question. I just want to start on the DXAI
47:36
investment that you guys announced today.
47:38
You know, you talked about there being some collaboration, you
47:41
know, between the company. So I'm just hoping to get more,
47:45
more information or, or if you're hoping that you could
47:47
shed more light on, on what that looks like and, and maybe how
47:50
the work XAI is doing can be leveraged at, at Tesla and and
47:54
vice versa. Yeah.
47:56
I mean, if you, if you looked at the disclosure which we also put
47:59
in there, we do talk about this is literally furtherance of our
48:04
Master Plan 4. And even today if you look at
48:08
Tesla vehicles, we are using Grok in there.
48:13
And as we look at things whether we can do it ourselves, yes,
48:19
there are a lot of things which we can do ourselves.
48:20
But if there are things which XCI can help accelerate our
48:25
progress, then why should we not do that?
48:28
And that is the reason why we've gone ahead with such an
48:32
investment because this this is one of the strategic initiative
48:38
because as it is, if you remember, I talked about how
48:42
many things which we are doing ourselves, if there are Ways and
48:46
Means we can find efficient ways for others to help us.
48:51
And next year literally fits into that bold.
48:53
So that's why we went ahead with.
48:56
It and we just had like a lot of investors ask us to do this.
49:01
There was like a lot of, you know, investor, but Tesla
49:07
shareholders said like we, we should invest in XAI.
49:09
So that's like we're just doing what shareholders would like
49:12
asked us to do pretty much. But but, but Greg Grok will be,
49:18
you know, very helpful in say, maximizing the efficiency of the
49:22
management of the of a large autonomous fleet.
49:25
So I mean, if you've got an autonomous fleet that's, you
49:28
know, in the future, 10 million vehicles or 10s of millions of
49:31
vehicles, then optimizing the efficient use of that fleet,
49:35
Grock will be, I think way, way better than any heuristic
49:40
solution or sort of manually managed solution.
49:45
And if you say managing, say a large team of optimist robots to
49:51
build a factory or build a refiner, you know, and, and say
49:54
a rare hypothetical, like this is hypothetical example, a rare
49:58
earth or refinery, which we do desperately need in, in America.
50:06
Then you, you say, well, like what's going to organize the
50:12
Optimus robots to to build that or refinery that would, you
50:16
know, you need kind of need an orchestra conductor.
50:19
And so then Grok would be kind of the orchestra conductor for
50:22
the Optimus robots to build the hypothetically.
50:25
And wherever it might, it might not be hypothetical in the
50:28
future. I'm just saying it's not
50:29
currently on our plans, but you know, we do we do need a lot
50:34
more or refining capacity in the US.
50:37
So then then what's going to manage?
50:41
They say, let's say 1000 optimist robots to to.
50:58
You're on mute right now, so I'm not sure if you're trying to ask
51:01
a follow up question. Ready.
51:07
We're going to move on to the next question, which was is
51:10
coming from Dan Levi at Barclays.
51:18
Dan, please feel free. Great, great.
51:21
Thank you. Elon, you, you talked about some
51:25
of the constraints on memory. Given the very tight supply, are
51:30
there any near term constraints on procuring memory?
51:33
And if there are, to what extent could you look at modifying the
51:38
functionality in the vehicle similar to what you did in 21
51:40
when we saw shortages on MC US? And maybe how are you thinking
51:44
about bridging in the next few years?
51:48
Well, the Tesla, the Tesla AI is, is very compute efficient
51:53
and very memory efficient. You know, so I think one of the
51:56
metrics once you consider for, for any given AI model is the
52:01
intelligence per GB, especially when you're constrained on RAM,
52:06
having a, an AI that has very high intelligence density per
52:11
GB. So you can say like what like
52:13
for a given number of gigabytes, how much functionality can you
52:15
get out of it? I actually think Tesla is ahead
52:19
of the rest of the world in, in intelligence density of AI byte
52:25
an order of magnitude or more like it's, this is going to
52:28
sound like a pretty bold statement, but I'm, I kind of
52:30
know what the, you know, what the intelligence efficiency of
52:35
the, of the big models are like rock and, and like the honest
52:38
and, and, you know, a bunch of the other models.
52:41
And I tell those AI is like in terms of its memory efficiency
52:44
of I think more than an order of magnitude better.
52:50
So, so, so that puts us in a pretty good position actually
52:57
for, for scaling. And we, we don't, we don't, we
52:59
actually do we, we do think that there's, we do have a solution
53:04
for logic and memory for let's say the next roughly 3 years.
53:14
But if you start going beyond three years and we look at the
53:17
scaling plans and how many fabs are getting built, and
53:20
especially if you factor in geopolitical uncertainty, you
53:25
know, there's always there's always risk that maybe the those
53:29
chips don't arrive, that people were expecting to arrive.
53:34
So that's, that's why I think we we need to have more fab
53:37
capacity in the US just in case, you know, chips don't stop
53:43
arriving for any reason. But you know, this is this is
53:48
really existential for for Tesla, because if you know,
53:52
Optimus is completely useless without an AI chip, it's not
53:55
like, you know, at least the cars we can put steering wheels
53:58
and pedals in or retrofit them if need, if need be.
54:01
But but Optimus is just a mannequin without, you know,
54:05
it's like the tin man for whatever from the Wizard of Oz,
54:09
but even worse. These the tin man could walk.
54:13
Optimus won't even be able to just sit there without an AI
54:16
chip. So we've, we've got a, a good
54:18
solution for a significant scale through, you know, for the next
54:23
roughly 3 years beyond that we, we will be supplier limited.
54:30
And so we've got to figure out some game plan to not not be
54:33
supplier limited. Great.
54:37
Our next question is going to come from George at Canaccord.
54:40
George, please feel free to unmute yourself.
54:44
Everyone, thank you for taking my my question.
54:47
So there's been a surge of startups, particularly from
54:50
China entering the humanoid market and wondering what the
54:54
long term competitive advantages that keep Tesla ahead are and
54:59
how based on what you've seen, will Optimus fundamentally
55:02
differ from these competitors? Thank you.
55:08
Well, I do think that the, by by far the biggest competition for
55:12
humanoid robots will be for China.
55:15
China is incredibly good at scaling manufacturing, actually
55:20
quite good at AI. As you can see from, you know,
55:23
the, the open source or not the open source, but the, the sort
55:25
of, I guess some of them are open actually.
55:28
But but basically the, the, the, the, the models that China's
55:31
distributing for free are actually quite good and they
55:35
keep getting better. So China's very good at AI, very
55:37
good at manufacturing and and we'll definitely be the the
55:42
toughest competition for Tesla. We to the best of our knowledge,
55:47
we don't see any significant competitors outside of China,
55:54
but China will definitely be be be tough competition.
55:57
There's there's no two ways about it.
56:01
So I was thinking like people sort of outside of China kind of
56:08
underestimate China. China's an ass kicker next
56:12
level. So you know, I guess you know,
56:18
we're going to build. We think we think optimists will
56:22
be much more capable than any robot that we are aware of under
56:25
development in China. So we think we'll be ahead in
56:32
terms of the real world intelligence, the
56:37
electromechanical dexterity, especially the hand design,
56:41
which is by far the hardest thing in the robot.
56:45
And in fact, I'd say there's there's there's really three
56:49
hard things about humanoid robots.
56:52
The building a an incredible hand that that has the same
56:59
degrees of freedom and dexterity as a human hand is an incredibly
57:02
difficult engineering challenge. Then there's the real world AI
57:08
and and scaling production. Those are the three hardest
57:13
problems by far for human or robot.
57:17
I think we're we're Tesla has, is the only company with that
57:21
actually has all three of those components.
57:26
So yeah. Great.
57:28
And our last question is going to come from Colin at
57:31
Oppenheimer. Colin, please feel free to
57:33
unmute yourself. Thanks so much.
57:37
Because you know you talked a lot about the CapEx spend, but
57:40
you know this is incredibly ambitious technology development
57:43
program that you're talking about.
57:45
Can you talk a little bit about the R&D spend and, and how
57:48
you're thinking about the synergies of of the different
57:50
components, You know, particularly in the hardware
57:52
side? You know, if you think about,
57:54
you know, batteries into, into chips into into memory and, and
57:57
the efficiency of the system and what sort of advantages you
57:59
think you'll end up getting out of, you know, some of these
58:02
purpose built devices that you'll end up integrating into
58:05
multiple end markets. Well, really all we're trying to
58:13
do is make sure that we, we can scale to very high volume with
58:19
autonomous vehicles with a human or robots and, and, and that we
58:27
address geopolitical risk, which I think, you know, there's so
58:36
many companies out there that are asleep with the switch with
58:38
regard to geopolitical, geopolitical risk.
58:41
They're like, or they just have their head in the sand and hope
58:44
nothing bad will happen. I'm way more paranoid than that.
58:49
I always think of Andy Grove's famous statement.
58:53
Only the paranoid survive. And why did he come up with that
58:56
statement at Intel? Let's think so.
59:03
I'm, I think there's a lot of wisdom in that statement.
59:08
So we're going to be paranoid and make sure that we can
59:13
continue to build batteries and robots and AI chips no matter
59:18
what happens, and companies that don't do that, a bunch of them
59:24
will cease to exist. Yeah.
59:26
I mean, remember, all this comes out of necessity.
59:30
It's not that we want to do it. It's just we.
59:33
We have no choice. Yeah.
59:35
I mean, we built the most advanced lithium refinery in the
59:38
world, by the way. So it's not just like I wrote
59:40
lithium refinery in Corpus Christi, it's not just a copy of
59:43
what others have done. It's it's an entirely new
59:45
process that is fundamentally more efficient and more advanced
59:47
than anything else in the world. The same is true of our of our
59:50
cathode refinery here in Austin. And and we wish others would
59:57
build this. Can, can other people please,
1:00:00
for the love of God, help the name of all that is holy and
1:00:03
others please build this, build this stuff?
1:00:06
It's not the first time you go. Exactly.
1:00:08
I mean, this is not the first time we've said something like
1:00:11
this. Like why do we have to put these
1:00:14
things in? Why can others not?
1:00:15
Also, please, can someone else pull these things?
1:00:18
I mean, this is very hard to build these things.
1:00:21
You know, we build them out of desperation, not for any, not
1:00:26
because nobody else is building lithium refineries and cathode
1:00:29
refineries. You know, we're pretty much the
1:00:33
not just the largest, but also the only lithium refinery and
1:00:37
cathode refinery in America. So, yeah, so we're making moves
1:00:47
to make sure that no matter what happens, tells it will prosper.
1:00:54
Great. Unfortunately, that's all the
1:00:55
time we have for Q&A today. We really appreciate everyone's
1:00:58
questions and we look forward to talking to you next quarter.
1:01:00
Thank you very much and goodbye very.
1:01:04
Cool.