Near-future quarterly financial projections

[Doggydogworld]

@Doggydogworld
Added delivery costs to the COGS:

"All hands on deck" probably reduces COGS. A bureaucratic company would have employees fill out time cards with time charged to the appropriate accounts based on what they were working on for each 15 minute interval. Tesla doesn't do that. Engineers are R&D and accountants are SG&A, even when they deliver cars.

The battery pack and motors are the costliest items in the car. So, if the cells cost $125/kWh and the motor $2500 - that would be $9k + battery packaging - lets say $12k. The rest of the car can't cost $30k ! If it does - then, there are gross inefficiencies that they can fix.

I don't have a single great explanation for persistently high COGS, unless Musk is blatantly lying about cell costs.

There are tons of little things. All that belt-driven stuff you see on an ICE -- AC compressor and pumps for cooling, power steering and brakes -- must be motor-driven in an EV. Such parts are available because full hybrids also use them, but selection is more limited so prices are higher. Model 3 still carries a superfluous 12v battery as a bridge to connect to the conventional auto parts economy. Model 3 unique features like glass roof add cost. All the FSD hardware probably adds a grand or so vs. your average sedan. Wheels are tires cost more than a typical sedan. Tesla's warranty reserve is ~1k more than a normal sedan. They feed some Supercharger costs through COGS. Internet connectively, too. Delivery costs more since their factory is poorly located and Musk has sworn off rail. Model 3 volumes are pretty high, but Tesla's overall business for commodity parts is smaller than the major carmakers so they have less leverage. It costs suppliers more to finance 60 days of Tesla receivables than most others thank to Tesla's crappy credit rating.

They'll grind away at some of this over the years, but other carmakers won't stay stupid forever. The auto biz is a grind. Silicon Valley isn't really set up to grind.

 

[EVNow]

Here is an interesting article I found (from 2012). @Doggydogworld @neroden @Spacemanspliff

Around the World With 5,500 Cars

Good question would be, why are EU ships taking so long to leave. Apparently takes just a working day to unload 3,000 cars @ 400/hour. I expect loading to take roughly the same amount time.

Got some answers on twitter. @Doggydogworld @neroden @Spacemanspliff

⛴ Morten Grove on Twitter

Around 1500 for China Around +3000 for Europe So 2000-3000 average would be a good basis, yes.

It based on loading time. A ship in Q1, can't remember which, loaded for a short period of time, sailed to China and the Chinese reported there being 1500/1600 cars aboard. So loading time seem to be consistent with number of cars aboard. Why this is the case? I'm not sure.

Limit to how many can physically be loaded, strapped in at a time. 12-1500 day as told by loaders.

Simply depends on number of decks.loading at one time.I have worked a ship loading 5 decks at the same time.If you have enough drivers you can strap a car in 1 minute or less 60 per hour per deck.
​

So, they are loading only 1,200 to 1,500 per day. That explains why China ships take a day and EU ships take 2 to 3 days.

Instead of looking at # of hours a ship is docked, we should look at # of "working hours". For eg. Grand Quest left SFO after docking for 12 hours (0.48 days). But it came in at 4:50 AM - finished loading and left at 4:19 PM. Asian Glory docked for 23 hours (0.95 days), but it came in the evening at 6:45 PM and after loading the next day, left at 5:40 PM. So, those 0.48 days by Grand Quest and 0.95 days of Asian Glory probably resulted in the same ~1k cars being loaded.

Similarly, EU ships, Morning Cornelia stayed for 2.5 days vs Grand Phoenix for 3.82 days. But they each had 3 working days, since Grand Pheonix docked at 8:50 AM, but Cornelia at 5:46 AM.

tl;dr : Small changes in # of hours docked doesn't mean much in terms of # of cars loaded. But we can get decent estimates based on # of working days at the yard.

 

[EVNow]

* I actually believe per boat avg in Q2 is slightly larger more like 3,146 model 3’s per boat due to average loading times but that’s just a guess.

In Q1 we had 37 loading days (16+21). So, about 44k @ 1,200 per day.

In Q2 we have 24 loading days (1+8+15). So, about 28k @ 1,200 per day.

If we think about 10k from Q1 was left over, we get a potential total of 38k in Q2 for outside NA.

 

[Zhelko Dimic]

Got some answers on twitter. @Doggydogworld @neroden @Spacemanspliff

⛴ Morten Grove on Twitter

Around 1500 for China Around +3000 for Europe So 2000-3000 average would be a good basis, yes.

It based on loading time. A ship in Q1, can't remember which, loaded for a short period of time, sailed to China and the Chinese reported there being 1500/1600 cars aboard. So loading time seem to be consistent with number of cars aboard. Why this is the case? I'm not sure.

Limit to how many can physically be loaded, strapped in at a time. 12-1500 day as told by loaders.

Simply depends on number of decks.loading at one time.I have worked a ship loading 5 decks at the same time.If you have enough drivers you can strap a car in 1 minute or less 60 per hour per deck.
​

So, they are loading only 1,200 to 1,500 per day. That explains why China ships take a day and EU ships take 2 to 3 days.

Instead of looking at # of hours a ship is docked, we should look at # of "working hours". For eg. Grand Quest left SFO after docking for 12 hours (0.48 days). But it came in at 4:50 AM - finished loading and left at 4:19 PM. Asian Glory docked for 23 hours (0.95 days), but it came in the evening at 6:45 PM and after loading the next day, left at 5:40 PM. So, those 0.48 days by Grand Quest and 0.95 days of Asian Glory probably resulted in the same ~1k cars being loaded.

Similarly, EU ships, Morning Cornelia stayed for 2.5 days vs Grand Phoenix for 3.82 days. But they each had 3 working days, since Grand Pheonix docked at 8:50 AM, but Cornelia at 5:46 AM.

tl;dr : Small changes in # of hours docked doesn't mean much in terms of # of cars loaded. But we can get decent estimates based on # of working days at the yard.

Ships moving these cars are 'multi-tenanted', booked by different brands, going to multiple destinations, so primary determinant is capacity that was booked and available at time ship docks.

From that point, I feel that rounding by workdays loses too much granularity. But your point about non-work hours stands, if there are indeed non-work hours. I would expect that these operations are 24/7, but I am not privy to that world.

 

[EVNow]

Ships moving these cars are 'multi-tenanted', booked by different brands, going to multiple destinations, so primary determinant is capacity that was booked and available at time ship docks.

Clearly they have booked greater capacity on EU ships than China ships.

Also small differences could simply because different number of people are available on individual days.

Ofcourse if one ship loads in 4 hours and the other in 8, its likely the first one is carrying fewer cars - unless they are loading simultaneously in double the number of decks.

 

[Spacemanspliff]

"All hands on deck" probably reduces COGS. A bureaucratic company would have employees fill out time cards with time charged to the appropriate accounts based on what they were working on for each 15 minute interval. Tesla doesn't do that. Engineers are R&D and accountants are SG&A, even when they deliver cars.

Not the sales people helping with deliveries but the costs to customer, as per Musks email "We also need to address the total cost of getting a car from our factory to the customer. Last quarter, there were many expedite fees and routing inefficiencies that led to higher than expected delivery costs."

Are delivery costs at all in COGS? I know Tesla charges a 1200 delivery fee but what happens if thats exceeded.

 

[neroden]

Ships moving these cars are 'multi-tenanted', booked by different brands, going to multiple destinations, so primary determinant is capacity that was booked and available at time ship docks.

From that point, I feel that rounding by workdays loses too much granularity. But your point about non-work hours stands, if there are indeed non-work hours. I would expect that these operations are 24/7, but I am not privy to that world.

They're not. Dockworkers are typically unionized and rather insistent about their hours of service.

 

[tmoz]

On a ship to China (or Europe), loaded with 2500 cars, does anyone know the cost of the:
A) Boat ride (for all 2500)
B) loading cost (for all 2500)
C) unloading cost (for all 2500)
?
I just picked 2500 as an example number.
It sounds expensive.

 

[Doggydogworld]

Not the sales people helping with deliveries but the costs to customer, as per Musks email "We also need to address the total cost of getting a car from our factory to the customer. Last quarter, there were many expedite fees and routing inefficiencies that led to higher than expected delivery costs."

Are delivery costs at all in COGS? I know Tesla charges a 1200 delivery fee but what happens if thats exceeded.

Delivery costs are in COGS.

It's hard to quantify Musk's complaining about the cost to expedite deliveries. He's somewhat bipolar and can freak out about tiny things, depending on mood.

 

[Spacemanspliff]

Delivery costs are in COGS.

It's hard to quantify Musk's complaining about the cost to expedite deliveries. He's somewhat bipolar and can freak out about tiny things, depending on mood.

Thanks, then what does the 1200 people are charged for delivery go towards? is that included in GM?

 

[Doggydogworld]

Thanks, then what does the 1200 people are charged for delivery go towards? is that included in GM?

The 1200 is in automotive sales revenue. For leases it's amortized over the lease term and included in automotive lease revenue.

 

[brian45011]

They're not. Dockworkers are typically unionized and rather insistent about their hours of service.

Unions bargain for shift premiums.

 

[EVNow]

Yes, deliveries.

2,324 model 3 overhang in April

Notes for my China estimate
-5 ships confidently delivering in Q2 with avg 775 3’s a ship
-250ish X delivered in April (plus it seems safe to say atleast 50 S), gives me 300 S/X per month in Q2


2,324 model 3 hangover
3,875 Q2 model 3 boat deliveries
900 S/X

Took another look at the 6 ships to China.

First 2 ships stayed for 2 working days at the dock. Next 4 stayed for 1 working day. So, even if we take a low of 1k per working day (the earlier linked tweets said 1,200 to 1,500 per day), they have probably shipped 8k cars to China. Along with left over from Q1, they have over 10k cars in China. Question is, how much will they deliver.

This is for all models, not just 3.

 

[Fact Checking]

Ships moving these cars are 'multi-tenanted', booked by different brands, going to multiple destinations, so primary determinant is capacity that was booked and available at time ship docks.

This is true in general, but note that Tesla's car shipments are largely "counter-cyclical": they are using Hyundai's Glovis fleet, which primarily carries cars from Asia to North America and from the EU to America. The Glovis ships probably used to be mostly empty on the way back: Detroit exports are a fraction of their EU and Asian counterparts, so there's a lot more cars shipping from Asia to the US than on the way back.

So because Tesla is the only car factory in California, and the only US carmaker with explosive growth, they possibly have a sweet deal utilizing these mostly empty ships - and don't have many other tenants competing for space.

The location of Fremont is actually pretty fortunate in this regard.

 

[Fact Checking]

Got some answers on twitter. @Doggydogworld @neroden @Spacemanspliff

⛴ Morten Grove on Twitter

Around 1500 for China Around +3000 for Europe So 2000-3000 average would be a good basis, yes.

It based on loading time. A ship in Q1, can't remember which, loaded for a short period of time, sailed to China and the Chinese reported there being 1500/1600 cars aboard. So loading time seem to be consistent with number of cars aboard. Why this is the case? I'm not sure.

Limit to how many can physically be loaded, strapped in at a time. 12-1500 day as told by loaders.

Simply depends on number of decks.loading at one time.I have worked a ship loading 5 decks at the same time.If you have enough drivers you can strap a car in 1 minute or less 60 per hour per deck.
​

So, they are loading only 1,200 to 1,500 per day. That explains why China ships take a day and EU ships take 2 to 3 days.

Instead of looking at # of hours a ship is docked, we should look at # of "working hours". For eg. Grand Quest left SFO after docking for 12 hours (0.48 days). But it came in at 4:50 AM - finished loading and left at 4:19 PM. Asian Glory docked for 23 hours (0.95 days), but it came in the evening at 6:45 PM and after loading the next day, left at 5:40 PM. So, those 0.48 days by Grand Quest and 0.95 days of Asian Glory probably resulted in the same ~1k cars being loaded.

Similarly, EU ships, Morning Cornelia stayed for 2.5 days vs Grand Phoenix for 3.82 days. But they each had 3 working days, since Grand Pheonix docked at 8:50 AM, but Cornelia at 5:46 AM.

tl;dr : Small changes in # of hours docked doesn't mean much in terms of # of cars loaded. But we can get decent estimates based on # of working days at the yard.

Thank you, this is very convincing, and it IMO comes the closest to explaining the ship mystery.

I'd still guess the error bars are several thousand units: 1,000/day or 1,400/day makes a huge difference to the end result. We might be able to better calibrate ship loading speed based on Q2 deliveries, if there's good regional data available.

 

[Fact Checking]

It's hard to quantify Musk's complaining about the cost to expedite deliveries.

I'd guess a big factor was overtime pay in the EU. Unlike the U.S., overtime is often statutory regulated in many EU countries, and mandatory weekend work can be super expensive.

Another cost might have been express and ineffective shipping around cars within the EU to deliver as many as possible before the end of the quarter: depending on how deliveries were progressing they might have shifted inventory from one delivery center to another - because they didn't know the max processing (and customer scheduling) constraints in advance at such unprecedented volumes.

 

[Fact Checking]

That being said if Tesla had 7k inventory left, plus what docks with ships with your higher boat average, with S/X sales that could be 13-14k units to sell in China.

This might explain why Tesla China pulled so many demand levers in Q2 already, despite sending fewer units in Q2:

• The Shanghai Model 3 price unveil: higher price and later availability than expected,
• the 0% leasing offer in Beijing,
• the Autopilot/FSD promotion

I.e. they want to be certain that all inventory is moved by end of June.

There's also the free lifetime supercharging offer in all markets on inventory S/X. This a big thing in the EU where consumers are a lot more fuel cost conscious. Getting a deal on pre-Raven units and free supercharging is a good combination.

 

[Fact Checking]

I just don't get how this car is not 8k+ cheaper to build.

Where do you see the discrepancy? The German and Munro tear-downs both agreed that the SR's marginal material and labor cost is around ~$28k.

If we add warranty and service reserves, D&A, stock comp, delivery costs, SR+ costs (glass roof, more premium sound, power seats, heated seats, extra USB ports and other amenities, extra labor cost for more labor intense interior, etc.), increase parts prices by 5k/week volume pricing instead of the 10k/week volume the tear-downs assumed, we get north of $35k quickly...

Beyond penny-level cost cutting, there's IMO three obvious ways to reduce per unit direct CoGs and improve margins:

• Scale up production and organic demand in lock-step: this reduces D&A, but also creates a larger pool of customers for higher ASP units.
• Reduce battery costs. "Battery day" will tell us more I suspect.
• Improve the take rate of zero CoGs, 100% margin options, such as the FSD option.

These measures seem to be at the center of Tesla's attention currently.

 

[Fact Checking]

Pre-Raven S/X inventory is still a mystery to me. It stood at 13.7k on 3/31 (includes demos, loaners, etc.). They built another few thousand in early April before switching to Raven. That's 16k+ pre-Ravens they need to move. The 2k Euro sales in April/May are all pre-Raven, but the 4275 US sales is a mix. Call it 3k pre-Raven. China looks <1k (360 in April according to JL Warren data given in this Seeking Alpha article). That's 6k of the ~16k+, leaving 10k to sell in June or at even greater discounts in Q3.

If they blow 80-90% of those 10k pre-Ravens out in June plus deliver 80-90% of the ~10k Ravens built in Q2 (some of which are on boats), that'd be 23k S/X deliveries this quarter. It'd also boost Euro sales above my estimates. And much lower Finished Goods Inventory would boost 6/30 cash on hand. A nice scenario all around.

But where are those 10k pre-Ravens today? The inventory trackers only show about 1k. They've never shown anything close to 15k, though, even when P&D reports and balance sheet indicated such. Those cars aren't all at stores and the death cult stalkers never found huge parking lots full of S/X. Maybe most of them were in China? That doesn't really make sense to me, but if true it means my June blowout sale scenario is unlikely.

How did you arrive at the 13.7k pre-Raven S/X inventory number?

 

[Fact Checking]

Here is the rough cash flow statement. I've only done the Operating cash flow & Capex part - not the financing part. The delivery numbers used are optimistic (as shown here) at 92k, 105k, 117k for the next 3 quarters.

I've also included the excel file that has p&l and the cash flow for you to play with.

View attachment 412065

Fantastic work and thanks for sharing the model!

Regarding the cash flow statement, I have two observations:

• I believe the "Accounts payable and accrued liabilities" numbers listed for Q2/Q3/Q4'19 appear to be written-in figures of +$25m, +$50m, +$75m, right? Fortunately I believe Tesla is going to post a significant payables expansion in Q2, which is going to create a very significant upside surprise in Q2. Basically with a record quarter and revenue being between Q3-Q4'18 levels, payables will expand again to the same level as in Q3-Q4'18: instead of the +$25m you modeled I think $600m-$700m will be the real figure.
• The other item is "Accounts receivable": which too is modeled with fixed figures of -$25m, +$50m, +$75m. Here we saw a net outflow in Q3-Q4'18 - this is a mirror image of payables expansion: many deliveries pending at the end of the quarter create a lot of dollars in flight.

Fortunately accounts payable expansion is the larger force as production expands: automotive suppliers have to be paid in ~60 days, while average cash/wire/loan clearing delay is measured in days. As a result in the year of 2018 payables expansion (largely from the expansion in Model 3 production) gave a net cash position improvement of +$1,722m, while receivables grew only by -$496m. So there's a rough 1:3 relationship between the two values and production expansion improved Tesla's cash position by +$1,226m.

In Q1'19 there was a whiplash effect from this: -$99m accounts receivable outflow from the tail-loaded deliveries in the quarter, while the full weight of record Q4 production drained Tesla's Q1 cash, on 42% lower revenues.

In Q2'19 if we go with your estimate of $5,375m revenues, which is 88% of Q4'18 revenues, about 75% of the drop will be recovered.

Furthermore, because much of production capacity will be delivered in the U.S., not in Europe and China, cash should be clearing faster and there might be less of a receivables expansion.

I.e. while the Q4->Q1 drop was pretty much the worst-case scenario to drain cash, Q1->Q2 is pretty much the best case scenario. If we go by the ~75% figure then the net cash improvement could be 0.75*624=+$468m.

But the precise timing of the transactions has a big effect on how much, plus there was the curious case of slower than expected payables expansion in Q4 that I remarked on at the time in this thread: I believe Tesla intentionally paid suppliers faster than required in Q4, to help the Q1 payables effect.

In Q2'19 they can take full advantage of payables expansion and make a big statement about solvency in the 'net cash and cash equivalents on June 30' field ...

I.e. it's IMO possible for Tesla to post more than $5b in cash at the end of Q2, if they so choose. (They might again offload some of that to Q3, to help the Q3 figures though.)