I really think a lot more smart people should be getting into manufacturing, it's kinda fun. It's super exciting I actually really love manufacturing, more people should get into it. It sort of got a bad name for a while but it's really interesting.
The biggest thing is designing the car for manufacturing. Generally when designing an object you can focus on aesthetics or focus on functionality. You can make something that's beautiful but weak on functionality, or you can make something functional but not very good looking. Having both aesthetics and functionality is super hard especially for a SUV. That's why there's not a lot of really attractive SUVs.
We just want people to know that whenever Tesla comes out with a product it is something that we think is amazing. It drives me crazy that other carmakers will come out with these cars that are not good, like why would they even come out with that? I don't know it just blows my mind, you can take a body panel and stamp it with this shape or that shape, and yet they choose to do the bad shape but it costs the same either way. There are some things that cost a little more in terms of the quality of materials and getting things really to fit accurately, so there are few things that cost more but a lot of it doesn't. You know, you can make an ugly expensive car or you can make a good looking expensive car.. and the same goes for affordable good looking cars or an ugly affordable car. I think the cost differences are really relatively small, I don't know I think maybe large car companies are just trapped in their own history.
One thing I really don't like about the car industry is that they will do these like show cars, and that show car never comes to reality. That just kills me, you shouldn't show somebody something really cool and then not do it. It's like, look at this delicious cake, and by the way it's made of plastic. Thus far, I think we've done a good job on design and technology of our products. I said we will build the world’s best car, we did that. At SpaceX, I said we will build the world’s best rocket, we did that.
I do use a phrase with our engineering and design team that “aspirationally, we're in pursuit of the platonic ideal of the perfect car” who knows what that looks like actually. But you want to try to make every element of the car as flawless as possible. There'll always be some degree of imperfection, but try to minimize that and create a car that is just delightful in every way. I think if you do that the rest kind of takes care of itself.
At SpaceX I hired a bunch of people from the auto industry to run manufacturing. People in the space industry have a really difficult time manufacturing things. They're pretty good at designing them in the first place but they don't actually know how to make them in volume. In taking the manufacturing techniques that were developed in the automotive industry and applying them to rocketry has been really helpful, in fact, our head of production at SpaceX came from the car industry. He used to run production of the Mini for BMW. The car industry is really good at making complicated objects at a low cost and what really goes into high-volume manufacturing of something that has to be extremely reliable. It's actually quite incredible somebody can buy a decent car for $20,000. It's nutty how much stuff is in a car. In automotive 400 engines per year is nothing. There are a lot of techniques which the car industry has developed to be able to do high volume production, but also be very reliable and consistent in doing so. I'm very confident that with the Merlin 1D design we'll be able to build 400 engines per year or frankly even 600 or 700 engines per year if we need to, and then the same with the cores. We are making a significant investment in tooling and production process efficiency, honing our software systems within the company that manage the procurement, assembly, and launch, trying to automate as much as possible.
I think anything - if you can 3D print something with sufficiently good material properties then that's the easiest way to do it. Certainly, in the volumes of a rocket company, it's harder to make it work for a car company. With printing you can print something that you can't make by any other means. So it actually ends up being lighter and cheaper than if we had built it by traditional methods. We actually print the SuperDraco engines, they’re printed out of titanium and inconel, and that actually allows us to reduce the cost of those engines quite a bit. In particular because we can print integral cooling channels. The biggest limitation on 3D printing right now is the size envelope, there’s a limit on how big we can print something. We're able to print the turbo pump components and much of the injector, not the whole thing but many of the critical parts we can print. That actually helps us in speeding up the development. So instead of waiting for castings to be developed, which can take several months, and then if the casting is wrong you've got to iterate in the casting, and each iteration can take several months. With printing those iterations can be reduced to a matter of weeks or months, so that actually helps with the speed of development as well. I can give an illustrative example in the air frame. That may be helpful. The normal way that a rocket air frame is constructed is machined isogrid. That's where you take high strength aluminum alloy plate and you machine integral stiffeners into the plate. This is probably going to go slightly technical but imagine you have a plate of metal and you're just cutting triangles out of it, that’s normally how rockets are made. Most of a rocket is propellant tanks, these things have to be sealed to maintain pressure and everything, and they have to be quite stiff. The approach that we took is to build it up, to start with skin sections and friction stir weld stiffeners into the skin sections. This is a big improvement because if you machine away the material you're left with maybe 5% of the original material, so a 20 to 1 roughly wastage of material, plus a lot of machining time. It's very expensive. If you can roll sheet and stir weld the stiffeners in then your material wastage can be 5%. That's the inverse essentially, instead of having a 20 to 1 ratio you have got 1.1 ratio. Instead of having 95% wastage, it's 5% wastage, it’s a huge improvement. That's one example, but there are many such things.
That cross-fertilization of knowledge from the rocket and space industry to auto back and forth I think has really been quite valuable. It's certainly been very valuable for me in thinking about how do we make mass-optimized vehicles, because in space mass optimization is extremely important. Me running both Tesla and SpaceX has been helpful because I see how both industries work and I can take things from one to another, so it's been good and of course the companies aren’t competing in anyway, so it's been quite helpful, actually. I'll give you an example, the Model S is an all aluminum body and chassis and we employ some advanced joining techniques and advanced casting techniques and so forth, that's like a fairly obvious thing if you're coming from the aerospace industry. You don't see a lot of aircraft made out of steel but cars are almost all made out of steel. In order to make the Model S light enough to make the non-battery-pack portion of the mass light - you can get more range. The two big factors for range are - what is your aerodynamic drag and how much does the car weigh. Those are overwhelmingly the big factors for range and so making the car weigh less was really important, and not just for range also for handling and acceleration. That was like an obvious thing to make a car out of aluminum, the Model S is actually the only car made out of aluminum in North America. In fact, there are very few cars made out of aluminum worldwide. The only other passenger vehicles that are made out of aluminum are made in Europe.
For SpaceX and Tesla our goal was not, initially, to do huge amounts of internal manufacturing. I generally think that there's been a bit too much outsourcing in general. Both outsourcing out of California and outsourcing out of the United States. Businesses sometimes tend to be a little sort of faddy. For a very long time there was a very strong outsourcing fad. I don't think people really looked at the fundamentals in a lot of cases when they outsourced.
We actually tried to do as little manufacturing as possible at first, but we found we had to insource more and more over time. We didn't start out insourcing 70 to 80% of our hardware. Initially we thought we will do as little as possible, but then over time we in sourced more and more out of necessity. It's not from the standpoint of we really believe in insourcing or outsourcing, it's just given - if there's a great supplier, then we'd love to use a great supplier, then if there's not then we need to do it ourselves. For rocketry, there's also ITAR limitations. Which is that rockets are considered advanced weapons technologies, so we can't just outsource it to some other country.
In the case of SpaceX unfortunately the supply chain in the rocket business tends to be very shallow. Often there's only one supplier and it's a very expensive supplier and they are really not designed for reusability. You're screwed if you don't make it yourself basically. That's led to SpaceX making sort of 70 to 80% of the rocket being built in house, literally from raw materials.
Another way to look at it is, in the way that to the degree that you inherit the legacy of the components, you inherit in the legacy of the cost structure and limitations. We found many times we'd sign a deal for supply of a component, and then that supplier would find a reason to triple the price. Basically as soon as they thought we didn't have any way out they would start with the conclusion which is 'triple the price' and insert reasoning. That's happened several times, and then we've in sourced the part to best price but often with a lot of grief. In the case of Tesla the automotive supply chain is much better. It's much more competitive, there are many suppliers for any given component, and so maybe 40-50% of the Tesla Model S is in sourced.
I also believe in having a tight feedback loop between engineering and production. If production is far away from engineering you lose that feedback loop. Someone who designed the car in a particular way doesn't realize that it is very difficult to manufacture the particular way that is designed, but if the factory floor is 50 feet away from their desk then they can just see it and it's obvious and they can have a dialogue with people on the floor. Likewise a lot of people in the manufacturing team have great ideas about how to improve the car, but if they are far away they can't communicate that to the engineers who designed it. I think it's something that's often neglected but having that strong feedback loop between engineering and production is really helpful for making the car better and finding efficiencies and lowering the cost. Particularly when the technology is evolving rapidly, it's important to have a very tight iteration loop between engineering and production, so as soon as you design something you can bring it to production right away. The engineers can go on the floor and see the mistakes that they've made, the production people can talk to engineers and say, 'here are some good ideas,' and so you can evolve the product and get to a better design solution faster. I think this is an important thing that's often overlooked.
We did literally a series of weekly iterations with the design team on Model S. Every Friday afternoon I would meet with the design team and we go over every nuance of the car. Every bumper, every curve, every tiny little piece of the car. What's right, what's wrong, and that has to be filtered against the engineering needs and the ergonomic needs, and the regulatory requirements. There’s a lot of constraints, you can't make a car just any old shape you want. It has to meet all the regulatory requirements, crash safe and all that stuff. It just requires a lot of iterative activity and caring about every millimeter of the car. That's what results in a good product.
There's a lot of things about Ford that I think are really interesting. Ford was just the kind of guy that when something was in the way he would just find a way around it, he just got it done. He is often associated obviously with the moving production line, that was a big innovation. He was also big on vertical integration, which I actually think is good. People have started to think that vertical integration is bad but I think Ford was right, you do need to be vertically integrated, not to a silly degree but you do need to be vertically integrated. Ford at least in the beginning of his career, I think he got a little too high on his own supply in his later career or Ford would have remained the largest car company in the world, but in the beginning he was actually really focused on what the customer wanted and what the customer needed. Sometimes the customer doesn't actually know what they need, but he really figured out that if we can make a car really affordable, reliable, and something that a farmer could really depend on their livelihood for, that's going to really make a difference in peoples lives, and he really got focused on that. Now overtime he should have decided that sometimes people want a color that is not black, and so you should provide that to them, but at least in the beginning of his career he had a tremendous insight to what would really make a difference as a product.
When you try to make something there is a big leap between the first prototype and manufacturing it in large quantity with good quality. It's really hard to make that leap. When you have new technology it takes time to make it lower-cost and mass market. This is true for anything. If you had like a soap factory, let me tell you, your first bar of soap would be like millions of dollars, but then you get to volume production, and then it's like $2. It’s true for any manufacturing situation, think of the earlier days of cell phones, or laptops, or any new technology it starts off expensive. Remember the giant phone that the guy on Wall Street was walking around with? that was cutting edge technology.
For some reason people decided they were going to do engineering here and do the manufacturing in the other side of the world. I think that ends up often being inefficient. The vertical integration is pretty important because one way to think about manufacturing efficiency is how long a journey did that molecule take from when it was mined. If it was mined in one part of the world, then went halfway across the world to get processed, then back halfway across the world to get processed another way, and eventually there's several trips around the world before it ends up in a finished product, that's obviously fundamentally going to be expensive. You just can't send things on world trips and expect it to be cheap or affordable, it's just not going to happen. So it makes sense ultimately for rail cars of raw material to come in one side, and for finished vehicles to exit the other side.
With Tesla today it is sort of split up, we've got the factory in Fremont California and the battery factory in Nevada, but I think for Gigafactory 2 and beyond I think we are just going to integrate that into one big facility. It's a three-step process - raw material - bunch of stuff happens - out comes the car. We actually have been steadily acquiring the buildings around us in California, so we're sort of growing like the Borg.
The Model 3 efficiency as a whole really is a quantum change in productivity, like really, really, crazy. We've just got to scale up production, and production is a hard thing. It's real hard particularly when it's new technology and cutting edge technology it's really hard to scale up production, because you've got to design the machine that makes the machine, not just the machine itself. That's where we have most of our engineering team working on.
The thing that happens once you start making almost all major subsystems internally, your supplier count actually grows dramatically. You have far more suppliers, not far fewer, but they're at the component level not at the major subsystem level. I really want to remind people that a car consists of several thousand unique items. We can only go as fast as the slowest item. What we were trying to do in advance of the Model 3 production was increase the scope of Tesla's internal capabilities, so that we're internally capable of making almost anything. Kind of like reserve troops. You don't know exactly where they'll be needed, but it's a good idea to have them, so that we can minimize the degree which a single supplier can stop the entire production line.
There's a whole bunch of little issues that are kind of trivial, that are challenges when you're making a new product because there are several thousand unique parts in the car, 90% of them are fine, 5% of them are slightly problematic, 3% or 4% are problematic and 1% are extremely problematic. You can't ship a car that is 99% complete. With software you just have to get stable functionality, but with a car, you know, you can't ship it without a steering wheel, or without a back seat, or something like that. It's an integrated product with thousands of unique components, so we are somehow at the mercy of what the slowest component is from several thousand suppliers. Things move as fast as the least lucky and least competent supplier. Any natural disaster that carries a name we have had happened to our suppliers. Factory has been burned down, there's been an earthquake, there's been a tsunami, massive hail, there's been tornado, the ship sank, there was a shootout at the Mexican border, no kidding, that delayed trunk carpet at one point. The border patrol wouldn't give us the trunk carpet because it had bullet holes in it, ‘we just want our trunk carpet’ that downed the production line for example for several days. That's the biggest issue the supply chain stuff. There's always something wrong. At any given point, there's always something wrong, because there's just too many things going on. There are thousands of unique components-- and even if one of those things is missing, you can't make cars.
One fiasco was-- I kid you not-- we were missing a $3 USB cable. OK, so we could not complete cars, because it's part of the wiring harness. You can't put the interior in without this cable. We could either make a whole bunch of cars minus the interior, which means that you've got to stack them up in the yard. It can be done, but then things go out of sequence, and it's way more inefficient, you don't have a moving production line. You have to send people out to hundreds of cars that are sitting in the storage yard. This happened to be a particularly pernicious cable, it was kind of routed under the carpet in a difficult place. It was literally $3, and we basically had to send people throughout the Bay Area to go and buy USB cables, at like Fry's. You were going to have a hard time getting a USB cable at Fry's because we bought every one of them so we were able to continue production.
I don't want to belabor the anecdote, but essentially the supplier was in China. We had plan A and plan B, plan A was like the normal supply chain process, but what the supplier did was instead of sending our parts in their own package they grouped it together with a bunch of other stuff for other companies, and sent that all via some extremely slow boat from China to LA. When it got to LA the other stuff didn't pass customs, and so they wouldn't let our stuff through, because-- I don't know what they put it in, but something that customs didn't like, the paperwork wasn't in order or whatever. So it got stuck there for like a couple weeks. Then we had plan B, we called and said, look you've got to air freight some of these cables-- cause they're just little cables-- to us. We talked to their US subsidiary and ordered from the US subsidiary, who then communicated to China, but then because this was another batch of parts, so it was kind of double the order, it exceeded the credit limit that we had. It bounced off the credit limit, so they didn't ship it. I mean, it's pretty farcical. That's just like one example, but there's many things like that. You move as fast as the slowest item in the whole car.
For battery packs one of the challenges we had was cobalt actually, cobalt is only available in a few places in the world, it's quite expensive, and the biggest source is the Congo, which tends to vary in its political stability. That's why going to the Model S we changed the chemistry to require only about a quarter as much cobalt, and thus reduce the cost of the battery pack, and also increase the energy of the pack.
The thing that is not well appreciated about cars and any kind of new technology is how hard it is to do the manufacturing. With maybe 50 or 60 people we can make a prototype of practically anything in six months. To manufacture that thing we need 5000 people to spent three years, and that is considered really fast. Specifically with respect to Model X we had a lot of challenges in the production ramp. That's always the most difficult time when you're going from zero to 1,000 cars a week. It's just you've got to pull this huge baggage train of suppliers along with you, and you've got to solve a lot of issues internally, so that production ramp is a lot of hurt. Basically, we were in production hell for the first six months. I mean, we knew this. Signed up for it. Not blaming hell, because we bought the ticket.