Tesla Production Plans
The recent Tesla (NASDAQ:TSLA) conference call that was supposed to be focused on Q1 results but instead became focused on Tesla’s future. CEO Elon Musk started the conference call with an announcement.
– Tesla intends to be the best manufacturer in the world
– Earlier goals of producing 500,000 cars by 2020 are being pulled forward to 2017
– Information released by the company does not give a full picture of how expectations will be met.
These are bold projections with material effects on financial results. Investors are right to carefully consider the risks and probabilities that Tesla can achieve these ambitious goals.
What are the company goals?
Like much Tesla investor information, the company goals have been given somewhat piecemeal with much room for interpretation. This information confusion has created a cottage industry on SA dissecting the goals and strategies of the company. So any analysis of the company must begin with a restatement of what I perceive the company production goals are:
Tesla will achieve 500,000 car/year run rate by the end of 2017
The Model S/ Model X assembly line has a capacity of 2,000 cars/ week (about 100,000/ year with holidays but no shutdowns)
Tesla will make as many as cars as possible at the Fremont factory (no plans for other assembly plants)
The Fremont factory may be used to achieve these goals, or plants in Asia and/or Europe may be necessary. (Yes, somewhat contradicts above statement)
The Fremont factory is the largest auto factory footprint in the world, second only to a tulip warehouse in Holland.
The Fremont factory once made “almost” 500,000 vehicles in one year.
Model 3 Product Program Status
In order to estimate when production of the Model 3 might begin we should look at the status of the product program. Neither Tesla nor Elon Musk use industry standard language in describing where the Model 3 product program stands.
In March a drivable Model 3 was shown to a select few at the product unveiling. The fact that there was restricted access to the car tells me that this car was what industry calls a mule. My take is that this is analogous to what Tesla calls an alpha prototype. After major elements of the design are complete a model shop will build a small number of cars with parts that may be available from current production models or hand crafted parts from prototype tooling. The idea is to get a feel for how the car will go together, how systems are “packaged”, as well as how it looks. The mule can also act as a test bed for new chassis, suspension, or drive train designs. It is a proof of concept vehicle and is not built off production tooling so the finished product will almost surely be different in at least small ways. Sometimes there are significant design changes before production of the prototype. Some parts may have been made with 3D printers or hand laid composites, but these parts are not adequate for production. That is one reason why all manufacturers try to hide their mules and why car magazines make a sport of hanging around proving grounds trying to get spy photos of mules.
From this prototype stage, that Tesla wants to map to the software “alpha” stage, all design drawings must be updated with any changes that come from testing, cosmetic or ergonomic changes. Manufacturing will weigh in on prototype design. This is where Design for Manufacturing (DFM) comes in (and other places). Tesla says that they will use DFM on Model 3 to avoid the start-up issues of the Model X. DFM is a 25 year old idea that has some software to aid implementation but it is also a cultural change that can take years to fully implement. Since this was just discussed at the last conference call as something that is yet to happen it would appear that the Model 3 design is not complete.
Design changes are generally made after DFM evaluation so that the car will build cheaper or with higher quality. This is an interactive process. With a firm design costing of parts can be requested from suppliers and samples requested. Part of Elon’s 2017 deadline seems to be this request to suppliers to supply parts off production tooling. When all those parts are in-house a “real” car can be made off production tooling and design changes made as necessary. Cost estimates will start to firm up at this stage. So at this stage Tesla is unlikely to have more than an educated guess at factory cost for the Model 3.
Tooling is where a large bulk of the product program cost is and vendors will begin incurring tooling costs to fulfill Tesla requests. Other manufacturers sometimes pay for this tooling so that ownership is clear in the case of vendor bankruptcy. In the past there have been statements that Tesla puts this cost on the vendors, such as with the Panasonic relationship. While that saves Tesla capital it does not give complete control over the tooling and that seems to have been a problem in the past. Large manufacturers have an army of engineers who travel the world to inspect these production tools and perform statistical analysis of the parts to see that the tooling can produce the rated number of parts consistently. It does not seem that Tesla has the manpower to do this step rigorously or the Model X problems would have been detected earlier. With the updated production plan dates Tesla will be forced to rely on vendors word that the tooling is ready and capable.
Some of the longest lead-time tooling will be the sheet metal dies. Many SA comments refer to the fact that press lines are in place so that part of the program is done. Press lines last fifty years, it is the product specific dies that you change in and out of the presses that makes the parts.
Last year Tesla bought a small tool die shop in Michigan and that should help with expertise and emergency modifications but it is unlikely that shop can produce a whole set of dies for a product program like the Model 3. Today, most large die sets are made in Korea. For example a single fender will need a blanking die, or possibly a laser cut program, and four stamping dies to progressively form the fender. Each body panel die can weigh 20,000 pounds or more. So we are talking about forming a 20,000-pound block of metal into a sculpted fender shape with a Class A finish. Suffice to say there is a lot of engineering and skilled trade time involved and it is hard to speed up. Here is the part that has me confused. A complete set of dies for a product program can take a year to design, test, and ship. That means that Tesla needs to be designing those dies right now but the Model 3 design is likely not complete. Starting die design without complete dimensional data is a crazy endeavor. Perhaps Tesla has found a new way to build cars since last year. Anything is possible but as an investor you have to equate this to a biotech penny stock where there is a rumor that the company has discovered something great but can’t announce it yet.
Taking all of the above into consideration, if the Model 3 were a “normal” product program of 3-4 years, there would be at least another 2 years until start of regular production (SORP). SORP is when the first car can be sold. Assembly lines often make hundreds of cars that cannot be licensed before SORP. Musk is pulling the program forward 6-12 months from a normal schedule. If he achieves this it will probably be a record for the auto industry.
The BMW plant in Spartanburg, South Carolina claims to have the fastest automotive start-up in history at 23 months. This is a good plant to compare. It is slightly smaller than Fremont and will be expanding next year to 450,000 annual capacity. That production has been achieved over twenty years with $7 billion investment. Note this seems to be assembly operations only.
Fremont Assembly Capability
Tesla has made the claim that the Fremont plant, under management of NUMMI, once produced “almost” 500,000 cars. Elon Musk has also referenced a Wikipedia article, recently started, that lists the Fremont factory as having the second biggest building footprint in the world, behind a tulip warehouse in Holland. Lets look at these claims and match them with the production goals announced in the conference call.
Wikipedia claims that the Fremont factory has the second largest footprint of any building in the world. In a televised interview Elon Musk even made mention of this. This really is not true. Even restricting the comparisons to auto plants, Fremont’s sister General Motors Assembly Division (GMAD) plant in Lordstown, Ohio is larger. General Motors claims a plant size of 6.2 M sq ft while the Fremont plant was listed for sale at 5.6 M FT SQ. The entries in Wikipedia are anonymous of course, but the article is relatively new and has been updated and reorganized regularly. As far as auto factories go, the Fremont site is perhaps just above average. The Ford Rouge Plant is certainly more impressive although it has multiple buildings.
Tesla has further touted the Fremont plant as having produced “nearly” 500,000 cars when operating under the previous NUMMI joint venture. In fact the plant never came close to 500,000 cars. The best year was closer to 420,000 total production. There used to be a good summary of NUMMI production data on Wikipedia but it was edited out of existence last February in a flurry of 49 edits in three days. Fortunately we do not have to rely on Wikipedia. Toyota, the NUMMI operations manager for the joint venture, has an excellent corporate history on their website. In 1994 a full size truck assembly line was added to the small car assembly line. At that time rated capacities were listed as a 150,000car/year truck assembly line and a 200,000 car/year mixed small car line (hitting 250,000/yr in a few peak years). It is likely the truck line capacity was increased after that in order to make the record year production of 420,000. In the world of auto assembly plants actual production rarely if ever reaches the rated capacity. At least one document in a lawsuit claimed that the record production year was less than 400,000.
The whole of the Tesla auto assembly operations are currently taking place on a single mixed assembly line (Model S and Model X) at the Fremont plant, the old NUMMI truck line. The capacity of this line has been stated by Elon Musk as being 2,000 jobs / week. In one interview Mr. Musk declared that 2,000 jobs/ week would be the rated capacity of the line for some time to come. In the latest 10Q report though the company says that investment in manufacturing capability will continue on all fronts (paraphrased) so it is unclear if the S/X assembly line capacity will be increased. The Model S and Model X were designed to follow the BMW 5 and X5 models in size and market appeal. The 2015 MY worldwide production for the BMW Model X5 and X6 were less than 200,000.
This brings us to the Tesla claim that the company will achieve a run rate of 500,000 cars/ year by end of CY2017. Elon Musk admits this is a stretch goal. Since it looks like the Model S/X capacity will be at 100,000 / year for now, the Model 3 capacity will have to be 400,000/ year. The reality of whether Tesla can produce this many cars and how much capital will be required is a key question for investors. Lets look at some back of the napkin numbers to do a sanity check.
A typical high volume assembly plant will produce about 200,000 cars per year. Many assembly lines run at other capacities but there are many reasons why 200,000 is a nice output level and is generally not exceeded by much. The BMW Spartanburg plant has invested $6B for two 200,000 jobs per year ((jpy)) lines.
Higher volume production is often difficult to sell at times without discounts and logistics can become a real problem. On the other hand, lower volume assembly lines, like the Models M S, tend to have nearly as much capital investment as a 200,000 car plant. Another operational issue at higher output levels is that of matching production to orders. Popular conception of assembly lines is that there is an angry manager in a room with a dial to infinitely adjust output, like in the movie Gung Ho. In fact large assembly plants can only vary the line speed within fairly narrow ranges. This is why you often see manufacturers removing a second shift (assembly plants only run two shifts for operational reasons) in order to match output with orders. So if you have a very high capacity plant in a down market the plant may have to be idled for extended periods then run for a few weeks. It just makes production scheduling very difficult. This is not just an inconvenience but drives costs up since there are large fixed factory costs and inventory losses associated with shutdowns. In the US unemployment benefits (FUTA SUTA) and other worker benefits have made wages nearly a fixed cost whether running or not.
Consider that the highest output single assembly line in the US is the above-mentioned Lordstown Assembly at about 400,000 jpy. Consider that both Fremont and Lordstown are similar age and size and both were built to a GMAD template by General Motors. In order to convert an assembly line like that at Fremont from a 200,000 car/ year to 400,000 jpy capacity many facility changes would need to be made. You need to produce 25% more vehicles in the same space. There are further logistics complications by producing three models on two assembly lines (Fremont) versus one model on one assembly line (Lordstown).
One of the biggest problems is the paint shop would be too small. Paint must be dried at a certain temperature for a fixed amount of time, so the length of the paint shop would have to be extended. In fact the length would have to be increased at least 50%, double if you want 400,000 real output and not “maximum rated” capacity. A maximum capacity throughput for a “standard” plant would be 60 car/hour X 7.5 production hours/ shift X 2 shifts/ day X (260 production days – 9 holidays – 14 days shutdown/year) = 213,300 cars/ year. The world being as imperfect as it is such a plant would average 200,000 cars / year.
Assembly line staffing for a dual line, 400,000/ year facility would be 4,500 line workers, if industry norms are met. Running overtime to meet these objectives quickly increases labor costs to the point where there is no profit. That kind of overtime is only done to take market share at all costs or to win a “most” prize.
Recently an analyst from a bank was given a wide aisle tour and reported on a fantastic new press line he was shown at Fremont. This press line is supposed to be key to Tesla reaching the 500,000 annual production goal. Randy Carlson also reported on this on as a key to production increases. First, the old NUMMI made 400,000+ cars with the existing press lines. It seems likely that they could do that and more.
Second, the stamping plant is decoupled from the body shop/paint/assembly line. Due to the decoupling the press plant does not have to run the same schedule as the assembly line. No one seems to be able to confirm what this magic press is, but it seems almost certain that Tesla finally finished refurbishing and installing one of the two Schuler presses they received from the General Motors bankruptcy. Claims that they will have 1,000 times the speed of the old lines are deceiving. The refurbished Schuler transfer presses are better suited to smaller parts, or “rail” type parts, and can probably make multiple parts per stroke. The strokes per minute for the Schuler press or the existing NUMMI tandem presses both fall in the 15 to 20 strokes per minute (spm) range. The Schuler presses can be set up to make a number of small parts per stroke. One good use for such a press would be to stamp the copper disk laminations that make up the motor rotor or perhaps battery module containers. If there were 100 laminations per motor (just a guess) you would need 50,000,000 lamination stampings per year, or a million each week. The new press might be helpful there but not strategic. The stamping and body shop welding and marriage stations have been called the “dumb” part of auto manufacture. The Body-In-White (BIW) construction is key to throughput and overall build quality and Tesla clearly struggles in this arena. It is process here that counts, not the size of your machine.
I did not want to get so into the weeds on describing the plant capacity but there has been so much misinformation on regarding what it takes for an auto assembly plant to make 500,000 cars in a year. One thing that comes up over and over is the assumption that because an electric motor is simpler than a petrol or diesel engine that the assembly will go much more quickly. In fact, the engine and transmission are not assembled at the car assembly. The engine, transmission, brakes and brake lines, and exhaust are lifted into place as one subassembly and fastened in two or three work-stations (about three minutes). Battery pack install is probably done using the battery pack assembly machine that was introduced by Musk as the secret to battery swapping. It is telling that the machine had a rate of about 80/ hour or about 280k/ year.
The entire assembly process takes about 24 production hours more or less, or three shifts to make the full circuit. If the plant is running a typical 200,000 cars/ year then there can be 800-900 assembly stations. Each station must take up about 30 feet of line space. The paint shop takes up a good portion of this production line. There are two primary paint ovens at Fremont, one for the Model S/X line and one for the planned Model 3 line. There are ancillary paint operations and drying ovens such as for ELPO primer and fascia (bumper) painting. The paint shop is on the roof, or second story if you will, of the Fremont plant. If you are looking at the plant from the road the paint shop is the part of the factory with all the stacks. In order to dry paint at a much higher production rate the drying oven conveyors must be extended or the paint formulation must change. Tesla has not explained their strategy and most manufacturers do not talk about the paint shop. In any event Tesla will be starting a brand new process and that always takes time to fine tune.
Tesla reported working on both paint shops several years ago when upgrading the S/X line. There appears to be more work going on based on news reports and the Eisenmann (paint shop builder) web site. Paint shops can cost north of $100M and lead times are long. It is good that Tesla is working on this now, but it seems that without expanding the paint shop capacity it will be very difficult to reach 400,000 cars on the Model 3 line. At least one of the auto paint shop primer conveyors lists a maximum speed of 70 bodies/ hour.
Summary
While nothing is impossible, it does seem highly unlikely that the Fremont plant will ever produce more than the 400,000 annual rate that Toyota managed. Engineers are famous for saying “give me enough time and money, and I can deliver anything”. Well, Mr. Musk seems to have taken the time variable away so it is a good guess that much more money will be spent to get Fremont producing by 2017, likely $2B at Fremont alone. It is a real long shot that the extra capital will get the kind of time reduction that Tesla has committed to.
Until there is more clarity on the Model 3 product program it seems unlikely that Tesla stock price will trade outside of it’s rather large range of the past year.
Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from ). I have no business relationship with any company whose stock is mentioned in this article.
Creative Commons (CC) article source: http://seekingalpha.com/article/3977783-look-teslas-model-3-product-program