Welcome to Tesla's first quarter 2021 Q&A webcast. I'm joined today by Elon Musk, Zachary Kirkhorn, and a number of other executives. Our Q1 results were announced at about 1:00 P.M. Pacific Time in the update deck we published at the same link as this webcast. During this call, we will discuss our business outlook and make forward-looking statements. These comments are based on our predictions and expectations as of today. Actual events or results could differ materially due to a number of risks and uncertainties, including those mentioned in our most recent filings with the SEC. During the question and answer portion of today's call, please limit yourself to one question and one follow-up. Please press star one now if you would like to join the question queue. Before we jump into Q&A, Elon has some opening remarks. Elon?
Great. Thank you. Q1 2021 was a record quarter on many levels. Tesla achieved record production, deliveries, and surpassed $1 billion in non-GAAP net income for the first time. We've seen a real shift in customer perception of electric vehicles, and our demand is the best we've ever seen. To be totally frank, we're used to seeing a reduction in demand in the first quarter, and we saw an increase in demand. That exceeded the normal seasonal reduction in demand in Q1. Model 3 became the best-selling midsize premium sedan in the world. In fact, I should say the best-selling luxury sedan of any kind in the world. The BMW 3 Series was, for the longest time, the best-selling premium sedan. It's been exceeded by the Tesla Model 3.
This is only three and a half years into production and with just two factories. For Model 3 to be outselling its combustion engine competitors, I think, is quite remarkable. In the past couple of quarters, we delivered roughly a quarter million Model 3s, which translates to an annualized rate of half a million per year. When it comes to Model Y, we think Model Y will be the best-selling car or vehicle of any kind in the world, and probably next year. I'm not 100% certain next year, but I think it's quite likely. I'd say more likely than not that in 2022, Model Y is the best-selling car or truck of any kind in the world. With regard to Full Self-Driving, Full Self-Driving beta continues to make great progress. This is definitely one of the hardest technical problems that exists, that's maybe ever existed.
Really, in order to solve it, we basically need to solve a pretty significant part of artificial intelligence, specifically real-world artificial intelligence. That sort of AI, the neural net needs to be compressed into a fairly small computer, a very efficient computer that we've designed, but nonetheless, a small computer that's using on the order of 70 W or 80 W. This is a much harder problem than if you were to use, say, 10,000 computers in a server room or something like that. This has got to fit into a smallish brain. I think with the elimination of radar, we're finally getting rid of one of the last crutches. Radar was really making up for some of the shortfalls of vision, but this is not good. You actually just need vision to work. When vision works, it works better than the best human.
Because it's like having eight cameras. It's like having eyes in the back of your head, the sides of your head, and just three eyes at different focal distances looking forward. Processing it at a speed that is superhuman. There's no question in my mind that with a pure vision solution, we can make a car that is dramatically safer than the average person. It is a hard problem because we are actually solving something quite fundamental about artificial intelligence, where we basically have to solve real-world vision AI. We are. Key to solving this is also having just a massive data set. Just having well over 1 million cars on the road that are collecting data from very sort of corner case, rare situations.
Sort of like so many weird things in the world, like a truck carrying a truck, or a car with a kayak on the roof where the kayak has a little weight dangling from the front of the kayak in front of the car, and yet the car must ignore this and just look at the road. It's really quite tricky, but I am highly confident that we will get this done. This quarter, and I think we'll continue to see that a little bit in Q2 and Q3. Q1 had some of the most difficult supply chain challenges that we've ever experienced in the life of Tesla. Insane difficulties with supply chains, over the whole range of parts. Obviously, people have heard about the chip shortage. This is a huge problem.
In addition to that, for example, we had quite a bit of difficulty scaling drive unit production in China, because we were unable to get critical engineers there because of COVID quarantine restrictions. Which meant that Tesla worldwide was dependent on drive units made at our factory in Nevada. That was a very challenging situation. I think we're mostly out of that particular problem, but those are just two of many challenges. The team's really done an incredible job of dealing with really severe supply chain shortages. With respect to the Model S and X, there were more challenges than expected in developing the Plaid Model S, what's called the Palladium program, which is the new version of Model S and X, which has a revised interior, and a new battery pack, and new drive units, and new internal electronics.
Has, for example, a PlayStation 5 level infotainment system. There are just a lot of issues encountered ensuring that the new factory was super safe was quite hard, because we're packing more energy in a smaller space. It took quite a bit of development to ensure that the battery of the new S, X is safe. Then we're trying to get all of these new cars slowly for the past few months. We're just stacking them up in the yard and just making refinements to the cars that we've built. We do expect to ramp Model S production and start delivering them probably next month. Then to be in sort of fairly high volume production for the S in Q3, and to start delivering the Model X in Q3 as well.
I think as we ramp up, I think probably the demand for the new S, X will be quite high. It's really just going to be a question of ramping supply chain and internal production processes. Probably, we're going to aim to produce over 2,000 S, X per week. Perhaps, if we get lucky, upwards of 2,400 or 2,500. This, again, is contingent on global supply chain issues, which there are just a lot of factors outside of our control here. I do think these things will get sold, so it's just a matter of time, and then we'll be doing well over 2,000 S, X per week. It's a great car. It actually costs us a little bit less to produce, but it is a superior product. In conclusion, there's a lot to be excited about in 2021 and 2022.
We're building factories as quickly as we can. Both Texas and Berlin are progressing well, and we expect to have initial limited production from those factories this year, and volume production from Texas and Berlin next year. At the same time, we are continuing to ramp production of Model Y in Fremont and Shanghai. In the background, we're continuing development work on the Semi, Cybertruck, the Roadster, and other products. Thanks to everyone at Tesla who have made this year a huge success. Now on to questions.
Thank you very much. We have some remarks from Zachary Kirkhorn as well.
Okay.
Yeah. Thanks, Martin. Thanks, Elon. Congratulations to the Tesla team on breaking multiple records in the first quarter of 2021, as Elon had mentioned, which is typically the most difficult of the year for many reasons. To summarize the quarter, I think it's best understood by three key items. First, we successfully launched and began the ramp of Model Y in Shanghai, achieving positive gross margin in the first quarter of production and receiving a great reception from the market. Second, as Elon mentioned, although we began the production process for the Model S during the quarter, we have not yet begun customer deliveries. The reduction in Model S and X deliveries from Q4 to Q1 were a meaningful headwind to free cash flows and profit generation.
For example, we incurred an estimated $200 million of direct P&L impact relating to this program in Q1, the majority of which is reflected in COGS. That's before even considering the impact of lost revenue and profits as a result of the transition. As Elon mentioned, we expect the first deliveries to begin shortly. Third, as we continue to work through the instability of the global supply chain, particularly around semiconductors and port capacities. While the Tesla team, in partnership with our suppliers, did tremendous work keeping our factories running, we did experience high expedite costs in the quarter, and they were also higher than they were in Q4, with some minor interruptions to production over the course of the quarter. We believe that this landscape is improving, but it does remain difficult, and it's an evolving situation.
If we double-click within net income, auto gross margin, excluding credits, improved sequentially and year-over-year. This is in spite of the cost mentions for S and X and expedites and a reduction in global ASPs as our cost structure as a company is reducing at an even faster pace. As we look out over the course of the year, we feel optimistic about our gross margin strength, particularly as some of these headwinds we're experiencing start to be resolved. On services and other margins, these have recovered and are trending towards profitability, aided by strength in the used car business, operational improvements in service, and additional service revenue opportunities that help absorb fixed overheads. On energy gross margins, these remained negative for a second quarter. This is driven by Solar Roof-related ramp costs and winter seasonality in the lease PPA business.
We continue to manage through a multi-quarter backlog on Powerwall. We're working as fast as we can to increase production, and this will aid in profitability of this business as those volumes increase. On operating expenses, these increased for Q1, which is driven by our investments in technology and growth. In particular for R&D, this includes the structural battery pack and 4680 cells, investments in the new S and X, and our neural net and silicon investments. On the SG&A side, we're setting up infrastructure and support for both China and EMEA in anticipation of volume to come there. As I said before, our plans show that we remain on track for sustained industry-leading operating margins. Double-clicking on cash flows, we continue to generate positive free cash flows, and this was despite the significant working capital headwinds from S and X.
Additionally, we are making progress reducing various forms of debt. We also invested $1.5 billion in Bitcoin during the quarter, then trimmed our position by 10%, which contributed to a small gain in our Q1 financials. Taking a step back, we've generated $8 billion in operating cash flows and $4 billion in free cash flows over the past four quarters. As we look forward, our plans remain unchanged for long-term growth of 50% annually, and we believe we're on track to exceed that this year as we guided to last quarter. Global demand remains meaningfully higher than production levels, and so we're driving as fast as we can to increase our production rates.
As we think about Q2 and Q3, these quarters should largely be driven by execution on S and X, as we've discussed, continued ramp of Model Y in Shanghai, and the associated cost reductions of these programs. We expect profitability and cash generation to evolve over the course of the year in line with those improvements. As we get towards the end of the year, our story will pivot towards the launch and ramp of our newest factories in Austin and Berlin. There's certainly no shortage of exciting things for us to work on and look forward to. Thank you, and we'll open it up for questions.
Thank you very much. We'll first take retail questions from, say, a website. The first question is: how is Dojo coming along? Could Dojo unlock an AWS-like business line for Tesla over the next few years?
Yeah, I'll jump in here. With respect to Dojo.
Sorry, Zach. My apologies. I was on mute.
Oh, go ahead, Elon.
Yeah, just basically saying that although right now a lot of people think of Tesla as a car company or perhaps an energy company, I think long term, people will think of Tesla as much as an AI robotics company as we are a car company or an energy company. I think we are developing one of the strongest hardware and software AI teams in the world. Certainly, we appear to be able to do things with Full Self-Driving that others cannot. If you look at the evolution of what technologies we developed, we developed them in order to solve the problem of self-driving. We couldn't find a powerful enough neural net inference computer, so we designed and built our own.
The software out there was really quite primitive for this task, and so we built a team from scratch and have been developing what we think is probably the most advanced real-world AI in the world. It sort of makes sense that this is kind of what needs to happen because the road system is designed for a neural net computer, our brain. Our brain is a neural net computer. The entire road system is designed for vision with a neural net computer, because it's designed for eyes and a brain.
If you have a system which has very good eyes, it can see in all directions at once, it can see three focal points ahead or forward, but it never gets tired, it's never sort of texting, it has redundancy, and its reaction time is superhuman, then it seems pretty obvious that such a system would achieve an extremely high level of safety, far in excess of the average person. That's what we're doing. Dojo is kind of the training part of that. Because we have over 1 million cars and perhaps next year we'll have 2 million cars in active use, providing vast amounts of video training data that then needs to be digested by a very powerful training system. Currently we use Tesla training software.
We developed a lot of training software, a lot of labeling software to be able to do surround video labeling, which is quite tricky. This means all eight cameras simultaneously at 36 frames a second per camera, labeling video over time. There wasn't any tool that existed for this, we developed our own labeling tool. Taking it a step further, obviously the holy grail is auto labeling. Now we're getting quite good at auto labeling, where the trainers train the training system, the system auto labels the data, the human labelers just need to look at the labeling to confirm that it is correct and perhaps make edits. Every time an edit is made, that further trains the system. It's kind of like a flywheel that's just sort of spinning up.
Really the only way to do this is with vast amounts of video data. We need to train this efficiently. Dojo is a supercomputer optimized for neural net training. We think Dojo will be probably an order of magnitude more efficient on, say, I'm not sure what the exact right metric is, but say, per frame of video, we think it'll be an order of magnitude more cost efficient in hardware and in energy usage per frame of video compared to a GPU-based solution or compared to the next best solution that we're aware of. Possibly that could be used by others. It does seem as though over time, just as an observation, I think basically it's the fact that neural net-based computing or AI-based computing is more and more of the compute stack.
Conventional computing, quote, perhaps heuristics-based computing, is still going to be important, still going to be very important, but neural net will become a bigger and bigger portion of compute. Anyway, that was a long story, but I think, yeah, probably others will want to use it too, and we'll make it available.
Thank you very much. Let's go to the second question from retail investors. The recent price changes on Solar Roof have been discouraging for customers and investors. Could Tesla share more about Solar Roof challenges and if the outlook has changed at all, i.e., 1,000 roofs per week?
Yeah. First of all, I should say the demand for the Solar Roof remains strong. Despite raising the price, the demand is still significantly in excess of our ability to meet the demand to install the Solar Roofs. Production has gone fine, but we are choked at the installation point. We did find that we basically made some significant mistakes in assessing the difficulty of certain roofs, but the complexity of roofs varies dramatically. Some roofs are to be literally two or three times easier than other roofs. You just can't have a one-size-fits-all situation. If a roof has a lot of protuberances, or if the core structure of the roof is rotted out or is not strong enough to hold the Solar Roof, then the cost can be double, sometimes three times what our initial quotes were.
In those cases, what we've obviously opted to do is to refund customers their deposit. What we cannot do is go and just lose a massive amount of money. We just got to provide a refund of the deposit. What is I think most important about the Solar Roof situation, which I tweeted about this past week, is that we're shifting the whole sort of solar situation, the solar Powerwall, basically solar plus battery situation to there's only one product basically, or there's only one configuration. We will not sell a house solar without a Powerwall. That solar could either be solar retrofit, with conventional panels put on a roof, or it can be the Tesla Solar Glass Roof. In all cases, it will have the Powerwall 2. Technically, it's actually a Powerwall 2 Plus, if you will. The Plus refers to a higher peak power capability.
Basically all Powerwalls made since roughly November of last year have a lot more peak power capability than the specification on the website. They have about twice the power capability, roughly. Depends on how you count power, but about twice the peak power and about arguably twice the steady state power of the specification on the website. The energy is the same, but the power is roughly double. All installations will have the Powerwall and the difficulty of the installation will be much less. It'll be much easier because the power from the solar glass roof or the solar panels will only ever go directly into the Powerwall. The Powerwall will only ever go between the utility and the main power panel of the house. Which means you never need to touch the main circuit breakers of the house.
You never need to touch the house circuit breakers. Effectively, almost every house therefore looks the same electrically, instead of being a unique work of art and requiring exceptional ability to rewire the main panel. This is extremely important for scalability. It's the only way to do it, really. This also means that every solar Powerwall installation that the house or apartment, whatever the case may be will be its own utility. Even if all the lights go out in the neighborhood you will still have power. That gives people energy security. We can also, in working with the utilities use the Powerwalls to stabilize the overall grid. Let's say that there's like there was in Texas, there was a peak power demand.
That peak power demand, because the grid lacked the ability to buffer the power they had to shut down power. There's no power storage. No good point of power storage. With a whole bunch of Powerwalls at houses we can actually buffer the power. If the grid needs more power, we can actually then with the consent, obviously, of the homeowner and in partnership with the utility we can then actually release power onto the grid to take care of peak power demands. Effectively, the Powerwalls can operate as a giant distributed utility. This is profound. I'm not sure how many people will actually understand this, but this is extremely profound and necessary because we are headed towards a world where as we were just talking about earlier, where people are moving towards electric vehicles.
This will mean that the power needs at homes and businesses will increase significantly. There will need to be a bunch more electricity coming somewhere. In fact, if you go to full renewable electricity, we need about three times as much electricity as we currently have. These are rough numbers, but you roughly need twice as much electricity if all transport goes electric, you need three times as much electricity if all heating goes electric. Basically, this is a prosperous future, I think both for Tesla and for the utilities. In fact, if this is not done, the utilities will fail to serve their customers. They won't be able to do it. They won't be able to react fast enough.
We're going to see more and more of what we see in California and Texas of people seeing brownouts and blackouts and the utilities not being able to respond because there's a massive change going on with the transition to electric transport, and we're seeing more extreme weather events. This is a recipe for disaster. It is very important to have solar and batteries at the local level at the house. In addition, it is important to have large battery storage at the utility level. That solar and wind, which are the main forms of renewable electricity, that electricity can be stored because sometimes the wind doesn't blow. Sometimes it blows a lot. Sometimes it blows too much, and sometimes it doesn't blow enough. If you have a battery, you can store the energy and provide the energy to the grid as needed.
The same goes for solar, because obviously the sun does not shine at night, and sometimes it is very cloudy. By having battery storage paired with solar and wind, this is the long-term solution to a sustainable energy future. As I said, this really needs to occur both at the local level and at the utility level. If it doesn't occur at the local level, what will actually be required is a massive increase in power lines, in power plants. They have to put long-distance and local power lines all over the place. They'll have to increase the size of the substations. It's a nightmare. This must occur. There must be solar plus battery. It's the only way.
Thank you very much. The next retail question is, Master of Coin, can you tell us anything about Tesla's future plans in digital currency space, or when any such major developments might be revealed?
Thanks, Martin. As I noted in our opening remarks, and we’ve announced previously, Tesla did invest $1.5 billion into Bitcoin in Q1, we subsequently sold a 10% stake in that. We also allow customers to make vehicle deposits and final vehicle purchases using Bitcoin. Where our Bitcoin story began, maybe just to share a little of the context here, Elon and I were looking for a place to store cash that wasn’t being immediately used, try to get some level of return on this, but also preserve liquidity. Particularly as we look forward to the launch of Austin and Berlin and uncertainty that’s happening with semiconductors and port capacity, being able to access our cash very quickly is super important to us right now.
There aren't many traditional opportunities to do this, or at least that we found and in talking to others that we could get good feedback on, particularly with yields being so low and without taking on additional risk or sacrificing liquidity. Bitcoin seemed, at the time, and so far has proven to be a good decision, a good place to place some of our cash that's not immediately being used for daily operations or maybe not needed till the end of the year, and be able to get some return on that. I think one of the key points that I want to make about our experiences in the digital currency space is that there's a lot of reasons to be optimistic here. We're certainly watching it very closely at Tesla, watching how the market develops, listening to what our customers are saying.
Thinking about it from a corporate treasury perspective, we've been quite pleased with how much liquidity there is in the Bitcoin market. Our ability to build our first position happened very quickly. When we did the sale later in March, we also were able to execute on that very quickly. As we think about global liquidity for the business and risk management, being able to get cash in and out of the markets is something that I think is exceptionally important for us. We do believe long-term in the value of Bitcoin, so it is our intent to hold what we have long-term and continue to accumulate Bitcoin from transactions from our customers as they purchase vehicles. Specifically with respect to things we may do, there are things that we're constantly discussing.
We're not planning to make any announcements here, and we're watching this space closely. When we're ready to make an announcement on this front, if there's one to come, we'll certainly let you all know.
Thank you. The fourth question from retail investors is: Does Tesla have any proactive plans to tackle mainstream media's imminent, massive, and deceptive clickbait headline campaigns on safety of Autopilot or FSD? Perhaps specialty PR job of some sort?
Well, I'll take this one, guys. From the safety side, I continue to say, as Elon Musk has, safety is.
Yeah.
Driving point and all. Go ahead, Elon.
I think, please go ahead. I think it's perhaps worth just going through the facts. Specifically, there was an article regarding a tragedy where there was a high-speed accident in a Tesla. There was really just extremely deceptive media practices where it was claimed to be Autopilot, where this is completely false. Those journalists should be ashamed of themselves. Please go ahead, Lars.
Yeah. Thanks, Elon. I was just saying, we're committed to safety in all our designs, and that's number one in what we do here. Regarding the crash in Houston specifically, we work directly with the local authorities, NTSB, and NHTSA, wherever applicable and whenever they reach out to us for help directly on the engineering level and whatever else we can support. In that vein, we did a study with them over the past week to understand what happened in that particular crash. What we've learned from that effort was that Autosteer did not and could not engage on the road condition that, as it was designed, our adaptive cruise control only engaged when a driver was buckled and above five miles per hour. It only accelerated to 30 mph over the distance before the car crashed.
Adaptive cruise control disengaged the car slowly to complete to a stop when the driver's seatbelt was unbuckled. Through further investigation of the vehicle and the accident remains, we inspected the car with NTSB and NHTSA and the local police and were able to find that the steering wheel was indeed deformed, leading to the likelihood that someone was in the driver's seat at the time of the crash, and all seatbelts post-crash were found to be unbuckled. We were unable to recover the data from the SD card at the time of impact. The local authorities are working on doing that. We await their report. As I said, we continue to hold safety in a higher regard and look to improve our products in the future, through this kind of data and other information from the field.
Okay. Thank you very much. Let's go to the next question from institutional investors. The first question is, proponents of alternative grid storage technologies claim that lithium ion is unsuited for long-term storage at scale due to vampire drain. Could 4680 cells address this limitation? Is the limitation even relevant for charging the energy equation?
Yeah. Just let me jump in on the vampire drain.
Yeah.
That's definitely not the issue. A good lithium-ion cell self discharges less than 0.001% of its energy per day.
A myth.
Yeah.
As mythical as vampires.
Yeah. I think the challenge with seasonal storage is your value proposition drops from hundreds of useful full cycles per year to less than maybe 10, or maybe even less than five cycles per year. It's just a different type of technology altogether that would make sense given that it's more than an order of magnitude different use case.
Yeah. We've got a long way to go before we're dealing with seasonal technology issues. Certainly a way to deal with seasonal technology would be to have wind and solar on the side of more southerly latitudes, but then across a variety of longitudes. Essentially, let's say in the U.S., for example, I'm not sure if you'll understand this, that you could actually power the entire United States with just roughly a 100 mi by 100 mi grid of solar. Sometimes people don't quite understand, how much solar is needed to power the United States? Almost nothing of the required to power the United States, and it's true of almost any country in the world. The solar incidence is a GW per sq km. This is insane.
In fact, if you took the clear area, just the area, say for nuclear power plants, the area that is considered not usable because a nuclear power plant is there, in most cases, if you just put solar there, it would generate more power than the nuclear power plant. This is because they typically have pretty wide, clear areas. If you have, say, 25% efficient solar panels, and then those are 80% efficient in how they're laid out, you're going to do about 200 MW per sq km. 5 sq km is 1 GW , which might be a typical sort of power plant. It's really not much area at all, and a lot of places can have wind and solar in place. Anyways, it's entirely possible to power all of Earth with a small percentage of Earth's area.
To transmit that power through high voltage DC lines, no new technology. You don't need room temperature superconductors. This is another myth. Room temperature superconductors are almost irrelevant in my opinion. Almost irrelevant. Low cost, long distance power lines using copper or aluminum are very important. Heating is I squared R. That's current squared times resistance. As you increase voltage, you can drop the current dramatically and drop the heating dramatically to the point where it is of minor relevance. Maybe you lose 5-7% with a high voltage DC power line, something like that. I want it clear, no new physics is necessary, no new materials are necessary. We just need to scale this thing up. The technology exists today to solve renewable energy.
As soon as people say, "Well, why don't we do it?" That's because the energy basis of the Earth is gigantic. Super, mega, insanely gigantic. You can't just go and do a zillion terawatts overnight. You've got to build the production capacity for the cells, for the battery cells, for the solar cells. You've got to put that into vehicles, you've got to put that into stationary storage packs, you've got to put that into solar panels and solar glass roofs, and you've got to deploy all this stuff. It is certainly the case that we can accelerate this, and we should try to accelerate it.
The right thing to do, I think from an economic standpoint, I think most, any economist would agree, is to have a carbon tax, just as we have a tax on cigarettes and alcohol, which we think are more likely to be bad than good. We tend to tax fruit and vegetables less. The same should be true. We should tax energy that we think is probably bad and support energy we think is probably good, just like cigarettes and alcohol versus fruits and vegetables. It's just common sense. I guess on the plus side, I'm not suggesting anyone be complacent, but sustainable energy, renewable energy will be solved. It is being solved, but it matters how fast we solve it. If we solve it faster, that's better for the world.
Thank you very much.
There's no question in my mind whatsoever that the energy storage problem can be solved with lithium-ion batteries. Zero. I want to be clear. Zero. I think the bias will tend to be towards iron-based lithium-ion cells. When people say lithium-ion, people think lithium must be a big constituent of the cells. It's more like 1%- 2% of the cell is lithium. The main part of the cell is the cathode. The main mass and cost in the cell is the cathode. For high energy cells, like for example, what we use in most Tesla have nickel-based lithium-ion cells, which have higher energy density, longer range than iron-based cells. However, stationary storage, the energy density is not as important, because it's just sitting on the ground. I think the vast majority of stationary storage will be iron-based lithium-ion cells with an iron phosphate cathode, technically.
I think the phosphate part is unnecessary. It's really just iron or nickel. Unnecessary in the terminology. Just think of it as iron or nickel. There's an insane amount of iron in the world, more iron than we could possibly use. There's also more lithium than we could possibly use. Basically, there is no shortage of anything whatsoever in iron phosphate lithium-ion cells.
Thank you very much. Let's go to the next question from Institutional Investors, which is: You've suggested that between a 5x- 10x improvement is achievable in the automotive production versus the first Model 3 line on the first principles physics analysis. Where does Berlin sit relative to that limit?
Oh, I think we're still quite far away from it. The thing to bear in mind with production is for those who have never done production, they just don't understand how insanely hard production is. I want to be very emphatic here. Prototypes are trivial. They're child's play. Production is hard. It is very hard. You say production at large scale with higher liability and low cost, insanely difficult. What Tesla achieved on the automotive side was not to create an electric car. The truly profound thing on the car side is that Tesla was the first American car company to achieve volume production of a car in 100 years and not go bankrupt. Basically, myself and many others at Tesla had to basically have several aneurysms to get this done. It was so hard. You have no idea. Anyway.
The thing about making a large, complex manufactured object is, let's say you have first order approximation 10,000 unique items. If even one of those items is slow, that sets your rate. Just one. Doesn't matter how trivial. We've had production stopped because of carpet in the trunk. We've had production stopped because of a USB cable. At one point for Model S, we literally raided every electronics store in the Bay Area. For a few days there, nobody could buy a USB cable in the Bay Area because we went and bought them all to put them in the car. Literally. There's hundreds of stories like that. Solving those constraints and a logistics problem that makes World War II look trivial. I'm not kidding. The scale is insane. We're talking millions of cars, massive global supply chain, 50 countries, dozens of regulatory regimes. It's insane. Yeah.
Thank you. The last question from institutional investor is, Master Plan Part 2 talks about an urban transport vehicle that is smaller than traditional bus with greater aerial density achieved by removing the central aisle. Do you have any updates to share on this goal?
Not at this time.
Okay. Thank you very much. Let's move to analyst Q&A.
Thank you. First question is from Pierre Ferragu with New Street Research. Your line is open.
Hi, guys. Thanks a lot for taking my question. I'd love to get actually an update on what you presented on the Battery Day. In the last six, seven months, I was wondering how much progress you've made on that front. First, in terms of process development, how are things coming together on your pilot line? Are you getting to the kind of production throughput you were aiming for? Second, actually on your production ramp. I was wondering in which sites you're ramping production capacity for the 4680 cell, and where you stand on ramping up that capacity as well. I'll have a quick follow-up on energy as well, if that's possible.
Well, Drew can hopefully add to this, but we have a small sort of pilot plant, which is still big by normal standards, expecting to have a 10 GW per year capability in Fremont, California. We've made quite a few cells. We're not quite yet at the point where we think the cells are reliable enough to be shipped in cars. We're getting close to that point. We've already ordered most of the equipment for battery production in Berlin, and then much of it for Austin as well. We're really down to the nitty-gritty elements. Overall, I think we're still sort of quite optimistic about achieving volume production of the 4680 next year. Drew, what do you think?
Yeah. Thanks, Elon.
Okay.
Thanks, Elon. Just one thing I would add is, there's been a lot of questions about yields. Actually, I noticed people asking about that and the yield progress has been really strong every day, and we were really still in commissioning phase of most of the tools, to the point where we're confident that the yield trajectory aligns with our internal cost projections. We did talk about yield also at Battery Day, which is one of the reasons why it's useful to check in on that. It takes a while, as Elon just mentioned, to go from prototype to production, and it's not just parts, it's processes, it's equipment. As we've matured the process equipment, we've gotten to where we need to be on the yield side.
Yeah. Basically, this is just a guess, because we don't know for sure, but it appears as though we're about 12, probably not more than 18 months away from volume production of the 4680. Now, at the same time, we are actually trying to have our cell supply partners ramp up their supply as much as possible. This is not something that is to the exclusion of suppliers. It is in conjunction with suppliers.
Yeah.
We want to be super clear about that. This is not about replacing suppliers, it is about supplementing suppliers. We have a very strong partnership with CATL, with Panasonic, and LG. Our request to our strategic partners for cell supply is please supply us with as much as you possibly can. Provided that the price is affordable, we will buy everything that they can make. All of it.
Yeah. Specific to that, we're on track to more than double the supplier capacity over the next 18 months.
Yeah. Exactly. We do expect from suppliers willing to proceed double the cell output next year versus this year.
Yep.
Okay. I had a quick follow-up on, maybe Zach for you, on your energy business. I understand when I get to gross margin with Solar Roof ramp. I was wondering, what do gross margin look like there when you look at the storage business, and what's your ambition in terms of gross margin in that business as, I guess it's going to grow in the mix in coming years. It's important for long-term modeling.
Yeah. We're seeing a lot of.
We're aiming for comparable margins in storage as in vehicle. It is important to bear in mind that vehicle is more mature than the storage. We already are at good margins with the Powerwall, but some additional work is needed for the Megapack to achieve good margins. Yeah. Drew, what do you think?
Thank you.
Yeah. Sorry, just jumping in, Elon. Absolutely agree. Yeah. Powerwall is mature. We've been producing Powerwall 2 for three years now, and we're at good margins there, but Megapack has more room to go to achieve our targets.
We have a clear runway for improving the cost per MWh of the Megapack.
Absolutely. Yes, we do.
Thank you. Let's go to the next question, please.
From Rod Lache with Wolfe Research. Please go ahead.
Hi, everybody. I was hoping maybe just first you could talk a little bit about how you're thinking about the rollout of Version 9 of FSD and the transition to the subscription model. It sounds like some of this is about to roll out next month. I'm not sure if that's the subscription model, but maybe you could just spend a little time talking about how impactful you expect that to be.
Sorry, go ahead, guys.
Yeah, we're working on getting FSD subscription out. There's a couple of internal technical dependencies, but from a business model perspective, that's aligned, and we're hoping to roll that out soon. The key thing that I say here, there's a lot of potential for recurring revenue based on an FSD subscription. If you look at the size of our fleet and you look at the number of customers who did not purchase FSD up front or are on a lease and maybe want to experiment with FSD, this is a great option for them. One of the things we'll need to keep an eye on is a potential transition from cash purchases of FSD, who may move over to FSD subscription.
There could be a period of time in which cash reduces in the near term, and then as the portfolio of subscription customers builds up, then that becomes a pretty strong business for us over time. We're hoping to get this launched pretty soon and see what the response is to it.
Okay, great. I was hoping, Zach, maybe you could just talk a little bit about OpEx? There was a noticeable increase, even excluding SBC. Obviously a lot going on this quarter, can you maybe just talk a little bit about how we should be thinking about that going forward?
Sure. On the R&D side, what we're seeing, as I mentioned in my opening remarks, is a convergence of a series of programs that are happening. Our R&D OpEx spend correlates to where we are in the product life cycle on different programs. We're at the tail end of investments in what we call internally Palladium, which is the new Model S and Model X. We expect that to decrease over time, but it was high in Q1 for a lot of the reasons that Elon had mentioned. We're also getting very heavy into 4680 development that Drew and team are working on and the associated structural battery pack that goes along with that. These are new technologies, not only new to Tesla, but new to the industry. We're investing heavily there on an R&D side to work out those kinks.
Spend in those areas should continue over time as we continue to work through the development cycle of those. I also mentioned, Elon talked a bit about Dojo and the potential there. From neural net investments and custom silicon investments, these continue to be areas that we spend on and make investments in. On the SG&A side, the business is pivoting very quickly to be global, China is ramping quite quickly. We're trying to make sure that we are staying ahead of the volume so that we have the right sales capacity, store capacity there, local investments in IT and others to manage the growth, such that as the growth comes, the execution challenges are smaller than maybe in similar periods of growth that we've seen in the past. We're making investments there ahead of the growth.
Overall, as we look at OpEx as a percentage of revenue over the course of the year, we do expect to see a substantial drop from 2020 to 2021 as the volumes in the latter part of the year pick up.
Thank you. Let's go to the next question, please.
Thank you. From Dan Levy with Credit Suisse, your line is open.
Hi. Good evening. Thanks. Two questions. One is on COGS. I think we've gotten from Battery Day a pretty good feel about the potential for COGS reduction related to powertrain, but I'd like to get a sense of the path to reducing COGS ex powertrain, as you'd still need a meaningful reduction on that front to make the math work on a $25,000 vehicle. What levers do you have to reducing your cost ex powertrain? Is it just more scale, better supplier pricing, or is it just based on ongoing cost reductions?
Sure. I think all of the above.
Yeah, on the vehicle side, there's plenty of opportunity as well. Obviously, building a car like a Model S is quite complex and has various moving parts. Model 3 and Model Y were steps of improvement in that. When you look at some of the other advancements that we're including in the Model Y factories into Austin and Berlin, we've reduced the body part count by as much as 60%, and the part costs money. We continue to find optimizations there as well as we get economies of scale when we start to talk about the volumes we're considering worldwide with four factories building the same vehicle. Both of those things on the vehicle side will improve our COGS as well, and the powertrain continues to be integrated into that.
Great. Just related, as we see Berlin and Austin ramp, I'd like to just get a sense on the comparison of Fremont versus the new capacity. Obviously, Fremont's non-optimized because you bought the old NUMMI facility, you had to retrofit that to your needs. Maybe you can give us a sense of how your new capacity is going to differ versus Fremont. What are the areas that you have efficiencies that you previously didn't have? How much does that add up to improved COGS over time to help you achieve that $25,000 vehicle?
Yeah, I don't think we want to talk too much about future product development. Earnings calls are not the right place for make major product announcements. Yeah, we'll get there, but we'll provide later.
All right. Thank you very much. Unfortunately, this is all the time we have for today. Thank you very much for dialing in and for listening, and we'll speak to you again in about three months. Thank you.
Thanks, everyone.
This concludes today's conference call. Thank you for participating. You may now disconnect.