The Factory Is The Product
Tesla Master Plan Part 3
Do you ever wonder why Elon Musk continually repeats mantras like “prototypes are easy; production is hard,” “manufacturing is underrated,” and “the factory is the product”? It’s because too many people still haven’t really absorbed the message, and this appears to be the root of most of the disappointment felt by many fans and investors following last night’s Tesla Investor Day.
Many wanted specifics on the anticipated FSD Hardware 4 computer, or refreshes of the S3XY models, or details consumers will care about for Cybertruck, and most of all, they wanted to get a glimpse of the vehicles going into the Gen 3 platform. Instead, we got grandiose, big-picture ambitions for electrifying all of human civilization with solar, wind and batteries, followed by a bunch of nerds droning on for three hours about optimizations they’ve made, cost trends, supply chain, manufacturing changes, enterprise software, and other boring technical details. What a letdown, right?
Wrong!
I was extremely pleased with the presentation last night, although it didn't cover everything I expected. Notably, the promised incorporation of SpaceX and Boring company to Master Plan Part 3 never came last night, which makes me wonder if more is coming later. We were told to look forward to the Master Plan white paper, the next-gen vehicle reveal event, and the next Impact Report, so Investor Day was only the first in a series of info dumps. Also, on Tuesday the president of Mexico in his announcement of Tesla’s intention to build a Gigafactory in Monterrey said to expect more information next week, though it’s not clear about what. Probably water conservation.
The big story for the car business last night is that they explained how they’re going to cut $15k or more from the cost to produce each car while slashing billions of dollars from the CapEx investments needed for each incremental 1M cars/year of production capacity. This translates to accomplishing the mission and earning trillions of dollars of net profit over just the next couple decades. They showed us how they'll make Gen 3, not the cabin design and sheet metal and shape. That stuff is relatively easy and everyone in the industry knows how to do it.
The Machine that Builds The Machine
Scaling to Extreme Size
On Battery Day in 2020, the Tesla team first announced its ambition to scale battery deployment to TWh scale and produce 20 million vehicles per year by 2030. It sounded crazy to most people and still to this day most people don’t believe that Tesla will grow to this scale. I have not seen any institutional analysts who are modeling for Tesla hitting this goal. The Battery Day presentation also included claims that the new technology would unlock the ability to sell a $25k mass-market car and that “Manufacturing will be Tesla’s long-term competitive advantage.”
What Tesla showed us yesterday wasn’t what as much as how. Goals without a realistic plan are worthless. Yesterday, we got a summary of Tesla’s plan to solve the manufacturing, supply chain and affordability challenges that stand between them and their growth target, and by extension the speed of transition to a civilization powered by sustainable energy. Tesla also showed, for the first time, actual data on cost trends for several of the major vehicle subsystems.
In fact, Tesla showed us last night that they are the only company in the entire automotive industry that can produce compelling, affordable EVs for the masses, the kind of cars that will make everyone want to voluntarily abandon internal combustion engines. It’s not hyperbole to say the ideas shown to us yesterday will, if successfully executed, revolutionize the way cars are built and give Tesla an overwhelming advantage on cost, scalability, worker safety, quality, and serviceability. Competitors still have no answer for the 2012 Model S, let alone Model Y, let alone the Gen 3 platform.
8 Wastes
Toyota originated a useful eight wastes framework that basically is the first principles of efficient industrial engineering. Eliminating these wastes improves safety, quality, throughput, morale and cost.
Overproduction
Inventory
Waiting
Motion
Transportation
Poor quality yield (rework & scrap)
Overprocessing
Underutilized human potential
I recommend rewatching the manufacturing sections of the Investor Day presentation with this waste checklist in hand and considering all the ways Tesla is deleting waste throughout the value chain.
Unboxed Assembly
This slide and the build sequence animation revealed Tesla’s excellent new “unboxed” process for manufacturing the Gen 3 platform of cars. In 2020, I was impressed with the Model Y Supertub (see image below). Last year I was impressed when Munro Live showed us the Model Y structural battery pack assembly with the seat and everything built up atop it. Now Tesla has revealed they’re taking that concept to a whole new level by completing almost all of the general assembly work before mating the major structural elements. This will revolutionize access, ergonomics and safety.
The body-in-white structure traditionally has been an annoying obstruction to general assembly because it’s basically a big enclosed cage. Humans have to lean and contort into awkward postures to get work done, and robots have to slowly and carefully move in and out of the openings. No more of that with this new design. This accessibility will make automation easier and cheaper as well.
Plus, as Tesla noted, this will improve the density of value-added operations in a given amount of space and time by about 30%, even as it also deletes traditional non-value-added operations like putting the doors on for painting only to remove them and put back on again in final assembly.
In manufacturing, every single action costs time and money. Every acceleration of mass, every move of a robot arm, every footstep, turning head, bending elbow, and twisting wrist. If it’s not directly doing something the beneficial to the customer then it’s waste. Tesla is designing their production around a 45-second cycle time. A few seconds here, a few seconds there, and before you know it a whole new station in the production line is needed and an extra few dollars of labor cost added to the bill for millions upon millions of cars over the life of the factory. We want to delete the need for these actions whenever possible. This reduction in motion will also reduce the energy embodied in the manufacturing of each car, because less motor power will be used on all this motion, and less energy for factory HVAC, lighting and other energy burned just running the building itself.
Tesla also mentioned "tested subassemblies” which is key to unlocking the full possibility of unboxed assembly. Fixing quality problems is much easier before they’re covered up with other hardware. In Final Assembly when all those modules are mated together, everything needs to connect and work properly on the first try or there will be lots of rework to perform. Tesla has adopted a software-style approach with continuous integration testing. Test early, test often. This, along with Factory Mode which Tesla formally revealed last night, also serves as the basis for Tesla’s digital self-management approach, in which employees receive almost gamified feedback from the vehicle itself on how well the build is going and if anything is wrong. Former Tesla employee and Agile consultant Joe Justice has discussed this topic extensively on Youtube in the last couple of years, and Tesla’s comments on Investor Day lend credence to what he’s been saying. Tesla is basically using both the car’s display and other lineside displays the computer is communicating with as real-time visual quality control and status indicators, the ultimate embodiment of the traditional andon and poka-yoke concepts pioneered originally by Toyota. Joe Justice has also discussed the benefit of Tesla’s software-style continuous integration testing in the factory with respect to reducing risk associated with introducing design changes, which is a crucial factor in how Tesla introduces several times more design improvements per year compared to any industry competitors. Failures are detected quickly and efficiently by a barrage of automated tests and this infrastructure allows for cheaper experimentation with new ideas.
Parallelization is yet another benefit of this unboxed assembly approach. The pieces are assembled separately, each in its own feederline, and then they come together near the end of final assembly. In doing so Tesla can also squeeze in 43% more people to simultaneously operate on each vehicle in a given amount of space and time. I spent five years working in a Boeing jumbo jet factory and witnessed many efforts to move work off of the planes into feederlines. The reason? Crowding of the interior of the plane and the doors to get in and out. This causes bottlenecking that forces work to be done in a series of steps over time, thus increasing the overall flow time. The same problem applies, on a smaller scale, to cars. In general, it’s better to find a way to assemble small pieces separately into increasingly larger combinations than to build one large monolithic assembly all in one go.
Many of us, including me, had been hoping to see Tesla announce a complete full-body casting as the next stage of development. However, this unboxed assembly plan is superior. A full-body casting would really be just a better way to complete the old assembly flow. Instead, Tesla will be casting the front and rear underbodies, where casting makes sense, but will stick with traditional stamped parts for the rest of the structure in order to allow for the benefit of unboxed assembly.
My big question about this new assembly sequence is exactly why no one else has done this before. Is it feasible only with an EV architecture? Are gigacastings or structural battery packs required to make it practical? Has every other company failed to change because of internal politics, corporate inertia, and fear of taking risks? I really don’t know. What I do know is that everyone else will have serious trouble matching Tesla’s cost to produce unless they copy this method.
The bottom line is this:
“Faster, less CapEx, more output per unit dollar.”
Lars Moravy, VP of Vehicle Engineering
Electrical
Tesla divulged a lot of juicy information in terms of their electrical and computer engineering advancements, and most of the audience appears to have underestimated the significance of it all.
Drive Unit Simulation and New Motor
First, Tesla says they developed advanced electromagnetic physics simulation software that they say is faster and more accurate than any commercially available options. They’re using this software to design better motors that take less investment, time, money and space to build. In other words, they literally have a better model of reality, and this enables them to squeeze more efficiency out of their motors in a way no one else can match without making their comparably accurate software.
Tesla also showed how they leveraged their prowess at power electronics design to make new inverters that use 75% less silicon carbide and one custom controller instead of four off-the-shelf controllers. Pierre Ferragu of New Street Research estimates that this change alone will save about $350 of cost, or about 1% of the Tesla’s current $35k average cost to produce.
In all, the Gen 3 drive unit costs only about $1k to make, uses no rare-earth elements, and is compatible with any battery chemistry. Tesla claim that they’re achieving the same power at lower cost for a permanent magnet motor without using neodymium, which is astounding and I’m wondering how they did it. We’ll probably need to wait for a teardown from Munro & Associates to find out.
Wiring
Wire harnesses are problematic in general. If you haven’t worked in manufacturing you might not realize just how terrible wire harnesses are, from fabrication to assembly to troubleshooting to service. Tesla showed last night a strong trend of reducing wiring in the vehicles and a path to reduce even more in Gen 3. I won’t rehash what they said, but it’s a big deal for the low-voltage system. Tesla not only eliminated much of the wiring but also made testing easier and claims to be set up for greater automation with Gen 3. Tesla has filed numerous wiring-related patent applications in the past, many of which were specifically tailored towards enabling automation, such as this one from 2019. Last night’s event suggests that some of that tech will finally be implemented in Gen 3.
Microcontrollers
Tesla spent a lot of Investor Day covering how digital twin and digital thread principles are deeply embedded in their entire business, though they didn’t use those exact terms. Tesla’s tight feedback loop between the fleet and the engineering team has been one of their strongest advantages since the release of the Model S in 2012, and now this strategy will reach its full potential with Tesla’s Gen 3 moving to using only custom electronics designed by Tesla for Tesla. It will all be vertically integrated, optimized not just at the vehicle level but also in conjunction with the backend server infrastructure to which all the cars are connected. The feedback loop will be completely closed for all systems in the car.
This move into deeper vertical integration in electronics and software is the culmination of 20 years of deep electrical and computer engineering research and development for electric cars since the company’s founding in 2003. Now Tesla’s engineers will, for the first time ever, have total control over the brains of the vehicles, and will capitalize on all the associated opportunities for predictive maintenance, over-the-air updates, reducing cross-car wiring, and more.
David Lau described the opportunity perfectly in his segment of the presentation:
Software-controlled hardware is fundamentally about being able to sidestep what otherwise would’ve been static tradeoffs between one attribute or another in a piece of hardware. We are able, with software, to instill intelligence, context awareness and context-specific behavior into what otherwise would’ve been a piece of hardware optimized for one specific scenario.
Tesla is moving deep into true full-stack software engineering. They are integrating and customizing everything from low-level integrated circuit components to firmware to networking to high-level software, and they’re now taking this to its logical conclusion by doing it for the entire car. Like Apple is doing with Apple Silicon, Tesla is achieving world-class computer performance by designing their own hardware.
“People who are really serious about software should make their own hardware.” - Alan Kay
48 Volts
The car industry has been stuck on 12V since the mid-20th century. It was a good option at the time when lead-acid batteries were the only viable option and when there weren’t many low-voltage loads, but as Pete Bannon noted in the presentation, with 21st-century cars the demand for current has exploded to such a level that moving to 48V finally makes sense. Tesla is not the first to realize this and it’s been talked about for years, including by the Tesla Technoking himself in this 2021 discussion with Sandy Munro:
Elon Musk: 12 volts, I mean, what are we still doing with 12 volts?
Sandy Munro: 42 is where everybody should be.
Elon: I think 48.
Sandy: Or 48, yeah. But at the end of the day we need to do something. That’ll reduce the wire diameters and stuff like that, and that’s just nothing but cash in the bank.
Elon: Yeah.
…
Elon: 12 volts is very much a vestigial voltage. It’s like absurdly low. So, I mean even, like, basically powered ethernet is like around 50 volts. … That’s really what the car’s low-voltage system should be at.
Sandy: Yeah, absolutely, I agree. In fact when we tore apart the Y I was expecting to see that, because you said you were getting rid of weight and the length of wires and whatnot, and when I pulled it off the harness it looked kind of similar.
…
Elon: You really want to put power and data over the same wires and have it be higher-speed than CAN bus, so you can like basically dump data on the bus instead of having all these point-to-point wires.
Despite these clear advantages of 48V, 12V has proven to be an extremely sticky standard for low-voltage components. There’s so much inertia and path dependence with these kinds of standards.
The 48V bombshell last night was Tesla’s announcement that not only are they moving to 48V but they’re doing it for the entire Gen 3 architecture to 48V in one fell swoop, while the rest of the industry is still dragging their feet or implementing it in a piecemeal fashion. This also is a power move signaling Tesla’s leadership in the technological development of the industry. Tesla has many suppliers that will need to get on board with this change. Fifteen years ago, Tesla barely had enough credibility and clout to get suppliers to call them back; now they’re pulling the whole industry into the future. Like with the 4680 battery cell form factor, I anticipate that other companies and suppliers will follow Tesla’s move to 48V. So, this advantage will not be sustainable for too long because it will be relatively easy to copy, but Tesla is getting a big head start and will have an edge for as long as other cars still use 12V. More importantly, the reduction in wiring means that one less obstacle for scaling and for critical mineral consumption (copper) is being removed, and that benefits the mission.
The Bigger Opportunity Tesla Didn’t Talk About
The beginning of the presentation covered what is needed to transition the world to sustainable energy, with an implied assumption that there will be a one-for-one replacement of current applications. I think this is a profoundly inaccurate assumption. Tesla leaders might realize this and just didn’t say it in order to keep the scope of the event focused on just the first mission of transitioning to sustainable energy without harsh sacrifices and austerity. Enough people think that what was presented was crazy as it is.
Solar, wind and batteries are tracking to deliver the cheapest energy in history. In some places, solar power already is the cheapest energy ever. My thesis is that overall energy consumption is actually going to increase many times over, despite the greater efficiency of electric motors compared to any heat engine. Whenever supply of any fundamental factor of production increases such that it’s cheaper, better and more available, economies consume more of it. This will happen with clean energy, and it already has been happening with thermal energy sources since the beginning of the First Industrial Revolution. We consume far more joules of coal/oil/gas per capita than we ever did with less-efficient traditional biomass like wood. The long-term cumulative numbers presented last night should be viewed as a baseline scenario showing it’s technically and economically possible to get humanity off of fossil fuels by 2050, but it’s not presenting the magnitude of the true long-term opportunity of energy and the new ways we will use it.
What else we got that stood out to me:
FSD Beta collision rate finally revealed: 6x lower than average human drivers on roads in the USA. Human-supervised FSD Beta has decidedly NOT increased risk for customers or others sharing the road. One could argue that even with fewer collisions there may be an elevated risk of catastrophic failure, but seeing as there have been zero reports of deaths or injuries, I’m inclined to believe that the collisions that have happened on FSD have not been especially serious compared to the statistical norm for human drivers.
Specifics on just how cheap superchargers are
Clear guidance that the crazy OpEx cost control we’ve seen is likely to continue as business scales, implying amazing leverage is coming
Lots of miscellaneous bits of technical info that the people who insist on negativity seem to have completely missed
Demand and Economic Implications
The demand for Tesla vehicles in a few different models under $40k would be plenty for selling 20M per year, if not more. The question for me has always been whether Tesla can actually make good cars for such little cost, even despite multi-thousand-dollar battery costs that are approximately the same no matter how expensive the rest of the vehicle is. After Investor Day my confidence in that happening has majorly increased.
Quick math:
$5k per car and 20M cars per year is $100B gross profit.
This is just for the car hardware and doesn’t include extra profit from software, services, charging, accessories, insurance and any government subsidies.
$10k per car, $200B
No company has ever earned $200 billion in one year. Apple and Saudi Aramco are the current leaders and each make about $100B/yr, and both are valued at about $2T, more than triple Tesla’s current valuation.
But wait, will Tesla stop at 20 million? After last night I’m more skeptical of that than ever. The mission involves moving as fast as possible and in all my time following the company I can’t remember any time they’ve said that 20M is going to be the steady state output. They just put 20M up as a target for 2030 and have stuck with that, but what about after 2030? We have almost 2 billion ICE cars to replace. 25M/year accelerates the transition to sustainable energy more than 20M, and 30M even more so.
Global vehicle sales peaked at nearly 100 M/yr in 2017 (according to Statista) before hitting a slump more recently and declining to 80M. Nevertheless, the long-term trend is for growth. Therefore, 20M vehicles in the 2030s would represent about 20% market share, or maybe as low as 15% if industry growth is strong. Plus, it’s fairly likely that demand for new vehicles will be temporarily elevated due to people retiring obsolete gas and diesel vehicles early in order to upgrade to electric. Governments may even offer incentives for doing so, like the 2009 Cash for Clunkers program in the United States in which the federal government subsidized trade-ins of old vehicles for more fuel-efficient options. So, is 15-20% market share actually a reasonable expectation for steady state production rate for a firm that has gigantic leads on both cost and value and that’s selling into a market that’s very price sensitive? I think that’s a completely unreasonable expectation, especially in light of the fact that Tesla already has achieved 11.2% market share in California (source), one of the world’s leading markets for EV adoption, with just the S3XY models which are too pricy for the vast majority of the population and have been priced so far above the cost to produce that the resulting gross margins are 25 to 30%. If that’s what Tesla can do now, then what can Tesla do as EV demand reaches the mainstream majority of the car-buying population and only Tesla has the cost structure and blitz-scaling strategy to ride the wave harder and higher than anyone else?
At Investor Day, Tesla was very clear in saying that the price elasticity of demand for Tesla products is extreme. That is, the quantity of orders is exquisitely sensitive to price. This isn’t surprising considering that cars are usually the second-most expensive purchase people make and it typically constitutes something like 15% of the entire budget for most households. Tesla’s huge price hikes in 2021 and 2022 followed by huge cuts in recent months have shown company leaders more conclusively that the price elasticity is in fact very strong. Today, several weeks after the January price cuts, orders are still flowing in far faster than the rate of production.
This price elasticity has important implications for the strength of Tesla’s margins and total addressable market as they drop lower-priced models onto the market and gradually reduce S3XY prices over time as well. Tesla appears poised to utterly dominate the $25-45k market segment, and this price range is the affordability-value sweet spot where most of the revenue can be made. The question is, at what profit margin? Tesla showed at Investor Day how they can get there and earn roughly $5k to $10k per car.
Prototypes are fun and exciting but what we really needed to know was the plan for resources, logistics and production cost and that’s what Tesla delivered. Don’t get upset that they fed us veggies and brown rice instead of candy. As a long term investor, the core of my thesis has always been that Tesla can make more cars than anyone in history with low OpEx overhead and strong gross margins as the lowest-cost producer. This is how the auto business can hit $100 or $200 billion annual earnings in the 2030s even conservatively assuming that level 5 autonomy development will have totally failed. In conjunction with Battery Day, Investor Day eliminated most of my remaining doubt.
Summary
The themes of Investor Day for me were:
Extraordinary engineering integration
Cost control
Clear path to world domination
Top-notch engineers are still running the show at Tesla. They still evidently aren’t too worried about impressing Wall Street, even at an event supposedly for investors. This event, like most Tesla events, was probably more effective at recruiting top engineers than any other purpose.
Tesla showed plausible ways to cut cost to affordable levels and scale to extreme size. Note that competitors like GM and Toyota have tended to do the exact opposite; they announce cheap vehicle models and show what they look like but show no viable plan for selling them at positive gross margin or in any meaningful volume.
Tesla addressed concerns about supply chain management and mineral supply, and they checked every box needed for making the transition.
All-in-all, a great Investor Day.
Disclaimer: I am about 99.9% all-in on Tesla via TSLA stock and call options. These are notes for my model that I’m sharing. I fully believe what I have written and I’ve put all of my money where my mouth is, but this is not advice regarding your personal investment or financial decisions. I'm not a certified professional investment or financial advisor and even if I were one, broad buying or selling advice would be inappropriate because there are so many variables to take into consideration such as your goals, risk tolerance, financial outlook, income, age, tax situation, dependents, and so on. Make your own choices. Thanks for reading and I hope you find this analysis helpful and will comment below on any mistakes you find.
.