I am someone who believes in manufacturing. I love manufacturing and really building objects that bring value to people. For some reason that got out of fashion, I'm not sure why. I think very often people think of manufacturing as just some rote process of making copies, which actually it isn't if you think of manufacturing not as some boring process of making copies, but rather that the manufacturing system of the car itself is a very complex machine. Just as innovation applies to the design and engineering of the car you can apply and should apply engineering and innovation to the machine that builds the machine. Manufacturing is building the machine that makes the machine and more often than not what I've found is the manufacturing is harder than the original product.
The way that people look at factories is often kind of thought of as a boring thing, this like catalog engineering. This is really the wrong way to look at it you really need to look at a factory like it is a product. It is a giant machine and it deserves more innovation and more engineering skills than the product itself. That's what we have done with the Gigafactory. The most important point that I want to make is that we have realized that the true problem, the true difficulty, and where the greatest potential is, is building the machine that builds the machine. In other words building the factory and really thinking of the factory as a product, not sort of a hodgepodge of things where the machines are sort of bought from a catalog. I can't emphasize enough how important this conceptual framework is, because really almost everywhere they think of a factory as this bunch of basic machines that are stuck together to produce copies. And it's mostly pulled from catalogs like get this machine from that company, get that machine from that company. If you apply that same principle to product design it wouldn't make any sense. Like if we would make the Model S out of parts from a bunch of different cars it would be ridiculous. It would be like a bumper from a Honda, a steering wheel from this, a motor from something else, it wouldn't make any sense to make a car from bits and pieces from other cars. Actually just like we do with the car, we don't try to create a car by ordering a bunch of things out of a catalog, we design the car the way it should be and then we or the suppliers we work with, make all of those individual components. There's almost nothing in the Model S that is in any other car. I think the same approach is the right approach to take when building the machine maker, the factory. You really need to design the car as an integrated product from the ground up and that's the same approach we have taken to the Gigafactory, and it's really quite unusual for that to be the case.
You can create a demo version of a product with a small team in maybe 3 to 6 months, but to create the machine to build machines takes at least 100 to 1000 times more resources and difficulty. For example, at Tesla we can make one of a car very easily, but to make thousands of a car with high reliability and quality and where the cost is affordable, is extremely hard. I'd say, maybe 10 times harder than just making one prototype - maybe more. I actually think that the potential for improvement in the machine that makes the machine is a factor of 10 greater than the potential on the car side, I think maybe more than a factor of 10. I have really come to appreciate that when I've been on the production floor sort of all the time and seeing things, and running production personally at a detailed level. I don't even have a desk or office anymore I'm just basically standing on the production floor and occasionally meeting in a conference room.
In terms of analyzing of how the factory should work I'm a big fan of using physics as a framework. I do my favorite thing which is apply physics first principles, that is like the best tool possible. Designing the factory from physics first principles means optimizing the density of the factory. You can think of the fundamental efficiency of the factory as the density of useful stuff, like what is the percentage of the factory that is actually useful stuff versus not useful stuff. When you think of a production facility on a fundamental level for a given size factory, the output is going to be volume, times density, times velocity. People don’t realize just how much improvement potential is possible and this is thick. We’re talking high school level physics necessary to figure this out it’s not like mega-complicated. Just go to a factory and say do a volumetric density calculation, say what percentage of volume of the inside of this building is doing useful stuff versus either air or not doing useful stuff. You’ll be shocked at how tiny that percentage is, like low single digits.
Let's sort of look at our factory, what is the density of useful to non-useful volume? it’s crazy low, it’s like 2 or 3% when you look at it volume metrically not just on a plane level, it’s literally 2 or 3% when you say car to non-car volume metric ratio. That seems like there is a lot of room for improvement, why is the volumetric efficiency of a car factory usually in the mid to low single digits? that’s very low. Why shouldn’t it be at least a volumetric density of 30% or 40%? 30% seems very, very achievable. You can also think of it like the design of a modern system on a chip or a computer. If you look at it say the complexity of the board, and you see how close together the line traces are, and how focused things are on the clock speed, and data transfer from RAM to say solid state disk for an internal CPU cache, it's like wow there is crazy potential for improvement here. Nobody would design a chip that had volumetric efficiency of 2% that would look ridiculous and yet they design factories that way. We will basically design a factory like you would design an advanced computer and in fact use engineers that are used to doing that and have them work on this. It is essentially designed like a very high density multilayer integrated circuit, an advanced CPU. If you think about it that is obviously how it should be done, I think actually over time the manufacturing process may look a lot like one of those super fast chip pick-and-place machines, it is super optimized for speed and density. Think about how did we improve the capability of your phone or your laptop? it wasn't by making a really giant computer the size of a table, it was by increasing the clock speed and density, the same principles apply to manufacturing.
Then you say like velocity, the output of the factory is the velocity of products from the factory. Then what is the exit velocity of the product? how fast are things moving out the exit? what’s the mass flow of the factory? what is a reasonable expectation for the exit velocity of vehicles from the factory? At first you may think that some of these advanced car factories around the world are very good at making cars and they may make a car every 25 seconds. That sounds fast, but actually if you say well the length of the car plus some buffer space is approximately 5 meters, so it's taking 25 seconds to move 5 meters. That is 0.2 meters per second, basically you are not much faster than a tortoise at that point. So that really doesn't seem fast, a slow to medium walk would be approximately 1 meter per second and a fast walk would be 1.5 meters per second. The best car factories in the world are doing 0.2 it’s like really low, like the fastest car plants in the world the car exit velocity is basically grandma with a walker. It’s real slow point 2 meters per second that’s really, really slow, we could do way better than that. The fastest person can run 10 meters per second, faster than 10 meters per second, so why is car exit velocity only 0.2 meters per second? that’s ridiculous, seems like you should be able to have cars exit at least at walking speed, this doesn't seem so crazy. Actually our speed on the line is incredibly slow, I think we are in terms of the exit velocity of vehicles on the line, including both X and S, is probably about five centimeters per second. This is very slow, I'm confident we can get to at least one meter per second, a 20-fold increase. One meter per second just to put that into perspective is a slow walk or a medium-speed walk. A fast walk could be one and a half meters per second, and the fastest humans can run is over 10 meters per second. At 1 meter per second you can still walk faster than the production line. So with significantly less engineering effort we can make dramatic improvements to the machine that makes the machine. I think probably a lot of people will not believe us about this, but I am absolutely confident that this can be accomplished. I am really fired up about that because I think it's one of those things that so much more is possible than people realize, I think it's really going to positively surprise people there. I'm really excited about revitalizing manufacturing, because it needs love and we're gonna give it. The results can be amazing. It's going be head and shoulders above anything else, it’s better than anything I've heard anyone even announce that they will do in near future, and we will do it in the present.
You can come up with some really cool new ways to manufacture a vehicle. There are many ways to skin a cat and it's remarkable how you can achieve the same objective with a hugely varying degree of difficulty. I have found that once you sort of explain this to a first rate engineer the lightbulb goes on. They spend huge amounts of effort trying to get a fraction of a percent of improvement on the product itself but actually that same amount of effort will yield an order of magnitude greater results if you focus on building the machine that builds the machine. It's just that a lot of engineers don't realize that this is possible, they are basically operating according to these invisible walls.
You can take an analogy and say if you wanted to kill a fly, you can kill a fly with a thermonuclear weapon, with a MOAB, with a cruise missile, with a machine gun, or a flyswatter. The end result is the same but the difficulty is considerably more significant from one to the other, and the collateral damage is considerably more significant.
What really matters to accelerate to a sustainable future is being able to scale up production volume as quickly as possible. That is why Tesla engineering has transitioned to focus heavily on designing the factory itself into a product. A first principles physics analysis of automotive production suggests that somewhere between a 5 to 10 fold improvement is achievable by version 3 on a roughly 2 year iteration cycle. The first Model 3 factory machine should be thought of as version 0.5 with version 1.0 probably in 2018. Sort of an internal codename for the factory machine is the ‘Alien Dreadnought’ You look at that and it’ll seem like an alien dreadnaught, like what the hell is that? At the point in which our factory looks like an alien dreadnought we know it's probably right then you know you've won. We think with Model 3 it will be alien dreadnought version 0.5 approximately, and then it will take us about another year or so, I don't know, summer 2018 to actually get to alien dreadnought version 1, and probably a major version every two years thereafter, by version 3 it won't look like anything else. It might look like a giant chip pick-and-place machine or a super high-speed bottling or canning plant. And you really can't have people in the production line itself otherwise you'll automatically drop to people speed. There's still a lot of people at the factory but what they're doing is maintaining the machines, upgrading them, dealing with anomalies. But in the production process itself there essentially would be no people with version 1 not version 0.5. I don't want people to think, oh, Tesla's going to have a factory without people. It's going be a huge number of people, but they will be maintaining machines and upgrading the machines and dealing with anomalies, and the output per person will be extraordinarily high.
Where it’s most obvious is in the cell production, our engineering teams work very closely with Panasonic to make dramatic improvements to the cell manufacturing efficiency and we think they are probably approaching 3X the efficiency of the best plant in the world, so that’s pretty good but there’s still a lot of room for improvement. Cells are going through that thing like bullets from a machine gun, in fact the exit rate of cells will be faster than bullets from a machine gun.
At Tesla we’re putting a lot of effort into becoming the world’s best manufacturer and I really mean that. Tesla is hell-bent on becoming the best manufacturer on Earth. I’m highly confident we will be, not by a small margin but by a margin that people don’t even think is possible.