Gigafactory Financial Modeling

The purpose of this thread is to form a working group to model economics and financial options for the Gigafactories. While there are other threads that discuss the Gigafactory more generally. I would like to focus this thread on getting the assumptions right and mathematically modeling the implications. So critical and constructive contributions from a variety of perspectives are most welcome.

Some of the advantages of financial modeling include:

• Understanding the long-term capital needs of Tesla to support various growth rates,
• Understanding cost structures and sensitivity to assumption,
• Assessing financing options and deals with off-take buyers and suppliers,
• Determining the present value of a tax break or other incentives, and
• Valuing a gigafactory as a whole.

To get the ball rolling, I want to share an off-take worksheet I've been playing with. It turns out that most of the essential financial issues are exposed when you try to answer the question, how might Tesla structure a five-year off-take aggreement to sell 1 GWh per year at a fixed price? My spreadsheet attempts to answer just that sort of question, and you are invited to play with it yourself. In short, I find that a contract allowing the off-take buyer to buy up to 1 GWh per year at the price of $125 per kWh require an upfront payment of about $95 million. Moreover, this pricing is compatible with doubling Gigafactory assets every 24 months without additional capital. All this, of course, depends on making quite a number of assumptions which I would like to discuss in detail and crowd source better starting points, but before getting to all that, let's consider what such off-take aggreements could mean for Tesla.

The first thing to note is that the Gigafactory investment in plant and equiment is about $100 million per GWh annual capacity. So this off-take deal would essentially finance 95% of the capacity needed with its upfront payment. However, the agreement is essentially a five-year lease on that capacity. So for five years, Tesla can benefit from financing that does not require issuing more bonds or potentially diluting shareholder value, lock in years worth of demand at a competive price ($125/kWh), and recapture the productive capacity at the end of the agreement. So effectly, Tesla can finance, build out, and operate profitably a portion of capacity years before it is needed in other Tesla products.

Additionally, one can think about Tesla Motors itself as the off-take buyer and consider how it might finance the $95 upfront at a low interest rate. For example, a five-year amortizing loan at 8% interest would add add about $23 per kWh to the $125 base price. Thus, simple financing can provide a path to a kWh price of only $147, that still allows Tesla to double it Gigafactory assets every 24 months.

In future posts, I will review the various assumption I am making and invite us all to challenge and refine those assumptions, but for now take a look at the worksheet. It's all in there.

Doubling our way to EV dominance and the target rate of return on the Gigafactory

Elon Musk has said that he thinks in 10 to 15 years about half of all cars sold will be electric vehicles. No doubt Musk anticipates that other automakers will play a big role in that transformation. But what if Tesla had to take use there singlehandedly. Could Tesla produce enough battery packs by say 2030? And how much initial capital would it require. Let's back into this and see how we might get there.

Let's suppose that by 2030 half of the car market is 50 million vehicles and the average EV requires 64 kWh. Yes, these numbers are chosen for ease of arithmetic, but follow along with your own numbers if you like. So in 2030, Tesla needs to produce 3200 GWh of battery packs (50M * 64kWh). Now if one 50 GWh gigafactory costs $5B, apparently 64 such factories would cost $3.2 trillion. Raising that much capital with zero return will simply not happen. On the other hand if the return is just 41% per year, then by the magic of doubling every two years, then Tesla already has enough capital on hand to pull this off. That is, beginning with 50 GWh capacity in 2018, double it to 100 in 2020, again to 200 in 2022, 400 in 2024, 800 in 2026, 1600 in 2028, and finally 3200 in 2030. Easy, right? We can leave asside the notion that energy density may well double over ten years, whence by 2030 we get 128 kWh per car for the same capacity. Note also that Tesla has over $2B on hand for the Sparks Gigafactory, plus about $1B committed from Panasonic. So if it is possible to yeild 41% annually on a Gigafactory, then we already have enough to get to 50 GWh capacity by 2018.

Now, I am not at all claiming that Tesla can pull off this rate of return, necessary for doubling every 24 months, but simply pointing out that, if it could, then no further capital raises would be need to power half the new cars in 2030. Alternatives include growing slower or raising twice as much capital every two years. For example, if the annual return on the Gigafactory is just 26%, then doubling would happen every 36 months, and Tesla could get to 3200 GWh capacity by 2036 without raising more capital, which is not too shabby. If every every new factory requied say $1B in fresh capital to supplement return on prior Gigafactories, then this requirement doubles every two years: $1B in 2018, $2B 2020, $4B 2022, and so on. If the ruturn is strong, then investors may be willing to go along with this but it would be quite a strain. Raising capital is great for getting this path started, but it is no substitute for obtain a solid rate of return which can self-finance continued expansion or at least justify ever increasing capital raises.

So in my off-take worksheet I make the assumption that the return on the plant and equipment is 35% (= ln(2)/2) on a continuously compounded basis, which is equivalent to a 41% yield (exp(1.35) -1) or doubling every 24 months. The point is not that I necessarily believe Tesla can achieve this, but that I want to understand how Tesla must price the output of the Gigafactory to be consistent with the objective of doubling capacity every 2 years without recourse to raising more capital. Only once I understand the implied price can I consider whether sufficient demand could possibly be cultivated. So for me actual demand is a secondary concern. Strategy and the return needed to execute strategy is the first concern.

In my worksheet, I assume that one GWh of capacity is 2% of the plant and equipment, that is worth $20 and $80 million respectively. Further I suppose that the plant has a useful life of 20 years and equipment 10 years. So depreciation is $1 and $8 million per year. This factors into the cost of goods, but does not tell use what price we should place on such capacity. Here's where we need to use our target rate of return R. The discounted life over N years of life is (1 - exp(-RN))/R. We need only multiply the unit price by the dicounted life to get the present value of the capacity under discounting at rate R. So we take the present value of the asset and divide by the discounted life to get the target price that is consistent with out doubling objective. Thus, we price the unit of plant at about $7 million per year and unit of equipment at $29 million per year. Combined this is about $36 target revenue on $9 of depreciation. This is a gross margin of 75%, but it is what we need if we are going to double the gigafactory every 2 years. Next we will build up other components of our price per kWh and see how gross margin works out for the full product.

Why the return on the Gigafactory must exceed the return on auto factories

I'm a bit surprised that no one has challenged my assumption that Tesla should target a 35% rate of return on the Gigafactory. The thought of doubling gigafactory capacity every two years is pretty fantastic. However, there is another way to look at such a high growth rate.

Consider that Elon Musk has said that he would like to see only about two thirds of Gigafactory capacity used for Tesla vehicles. The other third would go to EV battery packs for other automakers, stationary energy storage and perhaps other applications. This assumes an abundance of capacity, 50% more than would be needed for auto sales alone. But how can Tesla justify enormous investment in so much surplus capacity. I see three basic rationales: 1) Long-term, these other segments may prove to big opportunities so it is worthwhile to do a little developmental work now. 2) Short-term over capacity is desirable to sustain growth several years out, so short-term off-take deals put assets to productive use before Tesla needs them. 3) The current and foreseeable rate of return on nonTesla auto segments is high enough to warrant building excess capacity. The first two rationales may be adequate for a small amount of investment attending to short-term timing issues and long-term opportunities, but if the auto segment is Tesla's real growth opportunity, it makes no sense to dilute this growth (and share value) subsidizing the battery packs going into other automaker's cars. (The opportunity cost is a kind of subsidization if Tesla is not fully compensated for this cost.) In my view the only rationale that justifies a massive 50% overinvestment in the Gigafactory is if the current and expected rate of return is superior to the rate of return Tesla can obtain in its auto segment.

Let's see why this is so. Suppose Tesla is able to grow its auto segment much faster than its battery business. Suppose the rate of return on its auto factories is 50%, but for its Gigafactories is only 25%. Where should it invest its next $1B, in the auto factory or gigafactory? Clearly, it should invest as much as it can in the auto factory and only as little in the gigafactory to cover the needs of the auto segment. And of course, when it does make this minimal investment in the gigafactory the return on that investment is actually higher than 25%. It's actually a 50% return owing to the fact that the 50% return on autos requires this minimal investment. So we see that if the return on the Gigactory is less than the return on the auto factory, we minimize the investment in the gigafactory which leaves very little for othe segments. Under this scenario Tesla remains almost exclussively an automaker as it is presently.

Consider the opposite situation where the gigafactories rate of return is say 50% and for auto factories is 25%. Now we are in a situation where our next billion dollar investment goes into building more gigafactory capacity. It is not necessary to grow the auto capacity to capture battery opportunities. It is under this sort of scenario, Tesla becomes much more of a battery pack maker than an automaker. Its able to double its battery business in about 17 months, but will not double its auto business with 33 months. Over the span of a decade or too if this differential does not level out, the battery business will dwarf the auto business.

So we see that we must assume that rate of return on the Gigafactory is at least as great as the rate of return on auto factory capacity. To do otherwise when pricing gigafactory off-take deals or selling stantionary storage systems leads to misallocating capital for in inferior rate of return. Basically, batteries would be priced too low, and Tesla would grow well below its full potential. We can see this right now in Tesla's battery constrained situation. It did renew its supplier contract with Toyota because it could not fetch a price high enough to pay for the full cars it would otherwise be able to sell. The margin it could make on a Model S was the full margin Toyota would have had to pay Tesla for the opportunity cost. Now when it comes to making investments in capacity a similar situation arises. Will the return on investment in surplus gigafactory capacity exceed the return on an investment in auto capacity? If not, then you either demand a higher price for spare batteries or you pass on a supplier deal.

It seems to me that Tesla has the potential to double is auto business every two years for at least the next ten years. This is just 70k cars in 2016, 140k in 2018, 280k in 2020, and so on to 1.1M in 2024. (Of course, some adjustment should be made for the fact that the Model 3 will lower the ASP.) Consequently, we should price battery packs under the assumption that that the Gigafactory can double every two year, a rate of return of abot 35%. If it turns out that demand for spare battery packs is not adequate under this assumption, then it would be better not to build out surplus gigafactory capacity but to focus growth on the auto segment.

Krugerrand said:

I got lost at: The purpose of this thread... :tongue:

Click to expand...

Sorry. Is it that dense? A common occupational hazard for a mathematician.

evme said:

I don't have ability to view a spreadsheet right now, but was wondering if you have seen this?

http://s3.documentcloud.org/documents/1291115/full-tesla-summary-report-analysis-letters.pdf

I know a lot of people missed it. And it provides valuable info.

Click to expand...

Thank you. This is very helpful. I had been intending to ask you where you got the pro forma you posted awhile back.

I'm not so sure the ramp up will be as aggressive as this pro forma suggests, but eventually I want to analyze this pro forma to see what sort of cash flow and capital requirements it could imply. But going into this I'll admit my bias is to think that a more gradual ramp up will coordinate better with the auto business and require less capital, specifically no sort of capital raise that potentially dilutes shares.

Thanks again for posting this.

Labor cost and the experience curve

Tonight I want to talk about labor costs. Curiously, Musk stated on Fox Business News that he sees the Sparks Gigafactory employing easily over 6000, but that could grow to over 10,000 eventually. This comment has sort of thrown my thinking for a loop. Let me first set out the assumptions I made in my worksheet. Then I'll try to make some sense of these more recent remarks.

First off, I've seen several different estimates of the average hourly wage. I think Tesla goes with $25, and this economic assesment linked above uses $27.35. Regardless when one considers a fully loaded cost, including benefits, training, taxes, etc., it seems it quite easily could be $40 hour or more. So I am presently assuming $40/hour as the fully loaded labor cost. If anyone has insight into better numbers for grossing up hourly wage, I'd love to get something a little more credible.

I would also point out that while Nevada will offer certain tax credits and abatements on wages, I am intentionally not factoring this into this present analysis. My aim is to get an analysis without NV tax breaks first, and then later do an analysis with tax breaks to see what the real impact is. If anyone wants to take on working out the analysis with tax incentives, please be my guest.

Second, we need an estimate of hours labor per kWh. At full 50 GWh capacity, the Gigafactory is assumed to employee 6500 full-time. This works out to 130 employees per kWh. At 2000 hour per year, we get to 0.26 hours per kWh. At $40 per hour, we get a labor cost of $10.40 per kWh.

Next, my general assumption is tha Tesla want to target a 30% gross margin. I would argue that this need not apply to the return on capital investments (as previously discussed) or on some components such as materials (as I will discuss in another post), but when it comes to labor I think the general rule should apply. Labor carries substantial liabilities, overhead and mangerial responsibilities. So it stands to reason that a company like Tesla would want a very high gross margin on labor inputs. So now grossing up $10.40 by 30%, leads me to price labor ar $14.86 per kWh.

Now here is where Musk's recent comments about 10,000 employees confuses me. If the total output remains the same, then increasing employees implies a decline in productivity. I cannot believe Musk would allow productivity to decline over time. Just the opposite, productivity should increase as workers, managers and engineers gain experience working a gigafactory. So if productivity goes up AND the number of employees increase, it must be that total output will increase. One option is to add more equipment if building space allows. The other option is to build out more space. Could it be that a second Gigafactory might be built right next to the first one, perhaps doubling output to 100 GWh? 10,000 employees producing 100 GWh per year, implies 100 per GWh, or 0.20 hours per kWh, a 30% gain in productivity. If average compensation remains the same, then labor cost goes fro $10.40 to $8.00 per kWh. Of course, some of the productivity gains should go to the employees too, so perhaps a 10% average raise is in order. $8.84 would still be a 15% decline in labor costs per unit.

My current worksheet does not consider experience gains such as this that accumulate over time and will lead eventually to $100 per kWh, but it is good to look ahead and try to envision a more dynamic model to illuminate a path to lower costs. One such analytic tool is the experience curve. (Look it up in Wikipedia and elsewhere.) The basic idea is that as cumulative production doubles, the cost per unit declines by a regular amount, typically about 15%, but depending on the area of application it could be more or less. I expect that Tesla is full of ambitious, fast learners, so a decline of 15% in labor cost per doubling seems quite modest. One can look at how quickly the gross margin has improved in Model S production to get a feel for this. So now let's suppose that $10.40 applies to the average cost of the first 100GWh of output. This is say, a four year ramp up. In the next two years, cumulative output will go from 100 to 200. Thus under an 85% (100 - 15) experience curve, this puts average labor cost for the second 100 (2019-20) at $8.84. (Yes, I cooked the numbers to make this all work out.) Using a 30% mark up this amounts to reducing the price by $2.23 on the cost of labor alone.

The beauty of the experience curve is that it gives you a way to extrapolates cost reductions well into the future. We need only know how long it will take for cumulative production to double and assume rate of reduction. So how long will it take for cumulative output to double. Well, if capacity is doubling every 24 months, then cumulative production will doube every 24 months as well. Thus, the cost and by extension the price may decline by 15% every two years. This lends more insight into why it is beneficial to double production so quickly. It will drive down cost quickly too. So my estimate of PRICE is about $165 in 2016ish. Assuming an 85% experience curve, that leads to a price of $140 in 2018, $119 in 2020, $101 in 2022, and $86 in 2024. So like Musk, I too would be disappointed if Tesla did not reach $100 PRICE within 10 years. Eight years would be on par. Note also if Tesla were to double production only every 36 months, we might not see $100 before 2025. The chances of truly disrupting both the auto industry and utilies is very much enhanced by doubling quickly, getting to $100 years before the competion.

RobStark said:

• The Tesla CEO also expects the company to start generating strong "free cash flow beginning in Q3 of 2015" and could pay for the construction of its planned gigafactory without additional borrowing

http://seekingalpha.com/news/1988595-musk-expects-fully-autonomous-vehicle-in-five-to-six-years

Click to expand...

This is excellent. Not only does it support my conjecture that no further capital raises are needed for the Gigafactory, but it also expresses a general way of managing growth. The idea is to generate strong enough FCF to power future expansion. 2015Q3 will be one of the heaviest negative cash flow quarters for the Gigafactory. It will be building out the building while installing equipment; moreover, there will be practically no production happening yet. So it is all outflow and no inflow. At yhis point of max spending, the positive cash flow from auto sales will cover all this. Deliveries look to grow by at least 40% per year. So if it is adequate at 2015Q3, it should only grow stronger going forward.

I would like to work out a cashflow model for this. Maybe I'll get a little time for it this weekend. If anyone wants to take a tab at it or already has a quarterly cash flow model for at least the auto business, please post. I'd love to collaborate.

These are exciting times. I'll definitely want to be long as Tesla goes into positive FCF. It looks like we could see this bird clear the runway in about a year.

Materials, Other Expenses and the Magic of the 30% Gross Margin

The final to categories in my pricing model are materials and a catch-all "other" costs. Here's where I probably need the most input from others.

First for materials, I have heard cost per kWh figures in the $80 to $90 range. So I went with $85. If anyone has a more accurate break down, please let me know. This area of expense is a bit ambiguous. It's not clear how far down the supply chain the Gigafactory will begin. The initial plan was that upto 15GWh of cells would flow into the Gigafactory. So that's already pretty far along in the supply chain. But the ambition is to move much closer to raw materials and even recycled materials.

My thinking on materials is that you don't want to build much of your gross margin on them. So I'm grossing them up by only 15%. Materials is the largest component of cost, but materials themselves are not really where a manufacturer adds value. Grossing up labor and capital investments makes more sense, because this is how value is added to the precursor materials. Some margin is justified, nevertheless, on grounds that their is suppier risk and price risk to manage. If an off-take partner wanted to write a contract such that the buyer would arrange for and pay for all the input material, then there would be nothing to gross up, as the buyer has assumed all the supplier and supply cost risk. On the other hand, if an off-take agreement has the producer assume this risk, then the producer should be compensated for it. The producer may need to enter into purchase aggreements with suppliers or financially hedge the risk of commodities. For example, if the price of cobalt goes up, Tesla may not be in a position to pass this price increase on to buyers or consumers. So this risk must be managed and justifies some margin.

Also thinking long-term, material costs may be component that is least likely to see experience curve price reductions. To be sure, Tesla may add substantial scale to so precursor material markets, but natural resource constraints may simply elevate prices faster than doubling a market can lead to cost efficiencies. This is also why it would be helpful to break this category down into discrete materials. Some markets may be so highly scaled already that no experience gains are foreseeable, while others may be ripe for innovation, at least in terms of how the material moves most efficiently from source into Gigafactory-specific production. Yet other components may be at substantial risk of becoming a bottleneck or driving up costs.

My "other" category simply fudges my ignorance of all the costs. I suspect that materials, equipment and labor will contribute about 90% of the cost. Energy could be a substantial cost, however, the Gigafactory is designed to be a net zero consumer of energy. Most of the cost of the solar, wind and geothermal energy should be covered in either the cost of the building or equipment. So for the most part energy is a capital expense, not an operating expense. This will be a good thing to keep in mind when people question whether a high margin for battery packs is sustainable; the energy component most surely is. Moreover, unique to this business, the Gigafactory may have substantial energy storage resources. Just doing a quality control check of running products through a charge/discharge cycle lrovides MWs of dispatchable energy. I'll be curious to learn if the Gigafactory can sell power back into the grid at peak rates. If so, they may even make net income on energy. Transportation is another big category, but it looks like the plants is well conceived and well sited to minimize transportation costs. Maintenance, taxes and insurance remain as costs. Is there anything I'm overlooking? Altogether I am setting other costs at $10 per kWh, which makes it as large as fully loaded labor. Moreover, I'm pricing it with a 30% margin. I suspect that experience curve effects could have a moderate long-term impact on this cost component.

So all in, I am getting a "no contract" total price of $164 per kWh. Cost is at $116, so this price comes in at a 31% gross margin. Of course, I could have simply totaled up the cost and grossed this up by 30% to get pretty much the same result. However, what is gained here is the understanting the the total price is consistent with the objective of doubling productive capacity every two years. This is really huge. We have also used a lower margin on materials to yeild to concerns arould commodity pricing. Yes, in the long run margins will squeeze lower as more of the nonmaterial costs are squeezed down to nothing. But we need to be very clear about the necessity of pricing the cost of new capacity into every product of the Gigafactory. While some may argue that margins will be thin on batteries, this thinness may come at the expense of being able to scale capacity. If the EV industry is to dominate the automarket by 2030, it must price about $36 into every kWh for the continued expansion of capacity. Financing can finese this, but not make it go away. So it is a least heartening to see that if Tesla simply targets a 30% gross margin on battery packs, it will generate sufficient cash flow to self-fund the expansion of all future Gigafactories.

I am still working on a long-term cash flow model for the Gigafactory. I would appreciate input and feedback before releasing it. From what I see already, I find Musk's comments about self-funding to be quite credible. There is a path to EV market dominance by 2030.