$CVNA: Hero to Zero?

Rational extrapolation is a bit hard these days, but let's try.

It’s been a bit under a month since I last wrote about Carvana and boy what a bit under a month it’s been.

Below I’d like to cover the key points of Carvana and how I’m thinking about them. You can skip around as you wish. In short, I believe that Carvana provides an exceptionally asymmetric bet and will not go bankrupt barring extreme macroeconomic degradation. I can of course be wrong, so take what I say with a grain of salt. I will lay out the underlying logic of my views, but leave out overly specific evidence as I’m not writing a book. I have researched all of these points diligently and believe them to be true, so if you disagree, let me know!

Also just for full disclosure, I’ve about doubled my share count post earnings and will continue to do so.

0. Opportunity

Briefly before I touch on all the bad, I just want to highlight the opportunity here. Let’s assume CVNA lives and can do $1k of EBITDA per car and sells 800k cars in 5 years. You get about $800m of EBITDA, call it a 20x EV/EBITDA multiple (KMX actual roughly and growing) , $16b EV. We’re currently at about $1.5b, so a clean 40% IRR if it works out to an EBITDA materially lower than KMX with a sub 15% unit CAGR. Not too shabby. Make it 1.4m units and 20x multiple and you start getting some silly math.

Edit: For a touch more color see here. I could have done better on the explainer, my apologies.

1. Units

Carvana units went from 117k in Q2, to 103k in Q3, and likely about ~92k in Q4. To be absolutely clear, while units underperformed sell side expectation in Q3, sell side is absolutely inept. Units were known going into the report, so that did not drive the stock reaction.

We’re currently looking at something along the lines of -20% YoY unit growth in Q4. Obviously, units going down sequentially 3 quarters in a row for a cash losing business is a bit far away from ideal. So here’s why I’m not super concerned.

First off, in recent weeks units have been relatively flat. We’re at about 6 weeks in a row now of 7300-7400 units per week. While this isn’t our ideal scenario by any means, its about the volume they were doing in Q1 and Q2 of 2021 where they were doing ~$4,000 of cash SG&A per unit. It is currently at $5,300 ex-ADESA, so meaningful room to cut back closer to what it was previously. If they were able to get back to these prior levels of SG&A, CVNA’s Q3 cash burn would drop $130m to $300m.

On the topic of deleveraging, I really don’t think this is a huge concern. Finance and Other GPU is rather scale agnostic. Wholesale GPU actually goes up with lower units all else equal since the denominator isn’t 1:1 related. Metal margin GPU suffers from some slight reconditioning fixed cost headwinds but the biggest factors here by far are labor, parts, and car acquisition. With SG&A, advertising is almost fully variable. I’d imagine >50% of Comp and Benefits is variable at least. Logistics is largely variable. Market occupancy is largely fixed but small. Other SG&A is a bit of an open question, but contains largely variable portions such as warranty and transaction expenses. All in all, I doubt 90k volumes vs 120k volumes for a quarter is more than a few hundred bucks difference. While that would be nice to have, cost cutting is the biggest factor.

For what it’s worth, management commentary has been pretty consistent about hitting EBITDA neutral at the least on current units or worse. My sense is the goal is to do this about Q2 or Q3 of 2023. Cost cuttings are going need time to flow through (more on that later) but I do legitimately think this is possible.

Market share of course is another question, as -20% YoY comps makes one wonder if CVNA inherently has a broken value proposition. As some comparison, KMX is down about 15-20% YoY for units as well. The overall market is down somewhere between 10-20%. So CVNA’s struggles are not unique to them necessarily, but still, this is a company that needs growth. So what’s with the compression?

At it’s core, I believe it to be a combination of two factors. The first is rather simple, CVNA is focusing more on profitable sales. So fringe markets like the PNW, California (>10% of volumes), Florida (>5% of volumes) and the Midwest have seen higher delivery fees, lower inventory availability, longer delivery times, etc. You will also likely start seeing more focus on pushing financing, as cash buyers and external financing aren’t particularly profitable either. CVNA can practice price discrimination here with delivery fees or other fees that will lower volumes, but be positive to cash burn. These losses are natural and I’m not super concerned, probably equates to about -10 to -20% of units. Management has stated that they are actually gaining share in the middle of the country closer to IRC’s and losing share in coastal markets, which seems to support this view.

The second factor is a bit more interesting. Ernie has called out recently that CVNA charges higher rates than the market. This is actually pretty true. Just taking a look at securitizations you see CVNA writing loans about 120bps higher than KMX on the prime side. When including subprime I’d imagine its about 150-200bps higher. If we assume say a 15% WA APR for CVNA and a 13% WA APR for KMX on a $25,000 car, with a 72 month loan term, then CVNA is going to be about 10% more expensive, at $800/m vs $720/m for KMX. If we assumed an 11% or so WA APR and a $20,000 car, you get a monthly payment of $520! Affordability is a massive headwind across the industry, and specifically for CVNA as they raised rates faster than others.

What is exceptionally interesting about rates is that CVNA ties their rates to the 2 year. Ernie Garcia has called out this dynamic on a couple of occasions and it’s interesting to me that it’s played out as true. ABS securitizations are effectively a “cost of funds” (plus floorplan fees) and are priced at spreads to the 2 year, not fed funds. This has lead to other dealers having exceptionally tight spreads while CVNA’s remained healthy (finance GPU was roughly flat in Q3 despite heavy hikes). Recently spreads between the 2yr and fed funds have tightened, and will likely invert actually going into 2023 if current rate markets are to be believed. This should equate to a nice tailwind to CVNA demand or GPU as other players are forced to take price.

In addition to the above, we are currently on pace for the worst used car market since 2009 in terms of volumes. While alternative transportation is improving all the time and cars are lasting longer than ever, the tightness in supply vs demand is likely to cause a rebound at some point. We’re currently at this weird stalemate where monthly payments are up something like 80% in a few years but people need cars, at some point these transactions need to happen. Sequential worsening in overall market volumes seems far less likely to me than sequential improvements.

So all in all what do we have with units? They’re low, but relatively stable at about 7,350 per week. This is a level they’ve been EBITDA positive at in the past so I do believe they can get back there again. Additionally, there should be tailwinds going into 2023 from a variety of factors including competitors raising price, interest rates stabilizing or lowering, car prices coming down, and overall market demand returning. At worst we might have flat units and EBITDA neutral, but I’d bet on solid unit growth going forward. I could be wrong, but if so I’d imagine its because they prioritize more margin which is liquidity accretive, or because the entire western economy blew up, in which case oh well. Given the price of the stock, I’ll take my odds quite comfortably here.

2. GPU

GPU is something I don’t think is particularly variable but I’ll touch on it briefly.

As briefly mentioned above, finance and other GPU is pretty variable and unit dependent. Other GPU ex-finance has been trending slightly upwards over time, so you might get an extra few 10’s of dollars in 2023, but nothing huge. Finance GPU I’d imagine looks a touch down sequentially in Q4 as the 10yr has popped another ~70bps and risk premiums have presumably increased with minimal capability to pass through costs given weak unit demand. I’d guess this probably drops $100 in aggregate in Q4 and troughs in Q4 or Q1 2023 as rates stabilize.

Wholesale GPU I’d imagine is also ugly in Q4, but perhaps equal to Q3 due to a lower denominator in retail units. Hard to say exactly how additional ADESA integration is offset by increased depreciation rates, but maybe a marginal headwind.

Retail GPU is where leverage can start to come into play. While the majority of recon costs are variable, there are some fixed costs. Unit production is going to be pretty low right now as inventory continues to trend down, so you’re going to see some weakness there. Depreciation is a touch higher than normal as well. These factors are offset by lower inbound cost due to ADESA, lower recon costs due to efficiencies, and buybox tightening in response to poor macro. All in all perhaps $100-200 of headwind.

One additional note here. I frequently see it quoted that falling car prices will make Carvana insolvent. This is something that make sense from the surface, but falls apart with scrutiny. As it so happens, just about everyone in the industry knows what depreciation is. For this reason, car acquisition prices tend to come down ahead of retail price increases. Go ahead and graph Manheim index versus CARG/KMX/CVNA retail price and you’ll see it pretty cleanly. Additionally, Manheim index includes things like luxury vehicles and such that aren’t going to be a clean comp to CVNA. I’m not saying depreciation can’t be a slight headwind, but it’s not going to break the bank. Also you have to consider that lower prices would improve volumes, so some short term GPU pain would give you SG&A leverage and higher demand!

Overall, Maybe we see $300 of headwind to GPU in 4Q. I’d be totally unsurprised if we come in better than this, so not a huge deal to see another $25-30m of burn. A lot of these negative factors subsequently become tailwinds going into 2023 as rate stabilize, depreciation normalizes, volumes pick up, ADESA ramps, logistics improves, fees rollout, etc. I’m pretty confident you can get to a normalized cash GPU >$4500 per unit by 3Q or 4Q of 2023, which would reduce current quarterly burn by about $60m per quarter, more than offsetting any $30m headwind in Q4. Not a huge deal to my thesis right now but I’ll pay attention to momentum to see if I’m directionally wrong for 2023 or more color on specific targets. Pretty happy with the odds this goes up instead of down in 2023.

3. SG&A

SG&A is a huge driver of liquidity, perhaps the biggest variable driver outside of CapEx which has cleanly come down. Total cash SG&A came down by about $75m QoQ in Q3 despite a 15% drop in units. This was primarily driven by the RIF from May having fully realized gains, not having RIF related costs, as well as advertising, Other, and Comp and Benefits improvements. Cash SG&A ex-ADESA actually came down almost $100m.

Really the core of the markets issue with the Q3 report was that it expected faster deceleration in Cash SG&A. While $100m or so is good, if you go line-by-line in the shareholders letter you’ll see there really wasn’t anything drastic being done. Realizing benefits from your May RIF and adding some efficiency doesn’t get you to EBITDA neutral. It’s rather plain I think that management simply wasn’t cutting, so why?

I believe management was cutting as if you go through it from their perspective, sales this year were pretty solid for the most part until about June/July. July was pretty weak, but if you’re the CEO are you going to start cutting due to 1.5 week months? They subsequently saw demand tick back up in August as well! So I kind of understand it, this is a management team famous for flying close to the sun and they did it again expecting unit growth in Q3/Q4. It didn’t happen and thus they looked a bit silly to not cut.

Fast forward to now and we see another RIF on 11/18 that’s probably saving $20-30m per quarter at least. We see hours cuts for variable roles reported over the past couple of weeks. We see declining job postings for the past few months that have accelerated over recent weeks. And just an FYI, behind the scenes attrition is pretty high in variable roles, so you can get solid savings without any RIF in many areas and it won’t be immediately obvious. All in all I wouldn’t be surprised if Comp and Benefits expenses are down 20-30%+ from a few months ago as they shed excess. I believe they can carve out another 20-30% on top of this as well to get more inline with early 2021 spending, as headcounts are still far above early 2021 levels. These cuts are going to take awhile to show up, as Q4 will see only partial realization and any GPU improvements may not flow through until Q2.

In my mind, layoffs are not something you want to be doing, hour cuts before the holidays are not something you want to be doing, etc. Management clearly has started taking things to the chopping block, and recent commentary hammering home points on chasing EBITDA is a clear shift from prior messaging. Management does seem to get it now that they need to cut out costs and not wait around for unit growth.

So at the end of the day, what’s the difference between management cutting now and 2 months ago? The street seems to think its a 50% haircut to the share price, but really the liquidity profile is no different. It’s a difference of perhaps $100m bucks? That isn’t moving the needle, the magnitude of cuts is. So I get it, you get some ugly short term stuff going on as the street doubts managements ability and capability to cost cut. You get concerns this isn’t being taken seriously. It’s all fair on the surface, but when you zoom out its just noise. The LT intrinsic value of the business and the probability of cuts being successful has only marginally changed after the Q3 report. If anything, saving $100m in cash SG&A BEFORE any cutting speaks volumes to how much fat there really is.

My sense is the scope of the cuts is quite large and there is room still to touch up. Dropping cash SG&A per unit back down to at least $4,500 is $70m per quarter of casn burn improvement. This would still be higher than Q1/Q2 of 2021 even while including ADESA. I’d imagine they can do even better here if they put their minds to it, but assuming $70m in SG&A improvements and $60m in GPU improvements, you get an AEBITDA savings of about $130m per quarter or $520m per year. Toss in CapEx dropping from $90m to $25m per quarter as guided, and your $430m of Q3 liquidity burn turns into something more like $235m of liquidity burn per quarter, quite a bit better, and I honestly believe my assumptions are conservative. If you start penciling out going past 2021 efficiency due to more cost discipline, EBITDA breakeven seems more than doable at current unit counts. Just have to wait and see!

4. Liquidity

Obviously the output of the above is going to be liquidity. To provide a brief summary:

1. Cash and Equivalents: $2.3b. This includes a revolver which they are currently not utilizing to save on interest expense. No cash is not $400m. Yes the revolver is perfectly stable. No there is no covenant risk.

2. Inventory: Inventory is still probably higher than where it needs to be and has been coming down sequentially. They saw a nice bump from this in Q3 to cash and I’d imagine they continue to see some extra liquidity flow through in Q4, probably a couple hundred million.

3. Unpledged Other Assets: Finance receivables and the like. Not really something you WANT to pledge, but it’s doable, $130m.

4. Unpledged RE: This is marked at about $2b, but lets assume they get it done for 50c on the dollar. This implies probably a 10-15% caprate, and assuming land values aren’t falling off a cliff provides a very real margin of safety to the land buyer. We even saw a vending machine sell for a 7.5% cap recently (IRC’s and ADESA would perform even better) so who knows, maybe discounting 50% is too much even.

Total, about $3.5b with RE at 50c on the dollar. Now obviously, you don’t want interest expense and tapping many of these come with interest expense. The floor plan is going to be meaningfully cheaper to tap than the RE, so no Mr. Sellside, they will not be monetizing the RE immediately that would be stupid. Why in the world would I do a big hassle for a 10-15% cap rate when I can finance my inventory for 6%? Of course they will need to manage market perception of their liquidity such to get that RE transaction done, but I don’t expect it soon.

In addition to the above, they will likely have about $700m of restricted cash requirements assuming max floorplan utilization. This haircuts our liquidity to $2.8b with RE, and $1.8b without. Now I encourage readers to do their own exercise, but under what I believe to be draconian assumptions this company doesn’t run out of liquidity in 2023. That simply just isn’t going to happen. The question is if they can right size EBITDA fast enough such that they can get to FCF positive before needing to monetize the RE or obtain other financing? If we assume that happens in Q2 or Q3 of next year, they would be left with something along the line of $700m to $900m of liquidity ex-RE and be burning less than $150m per quarter. At such a point they cleanly get through 2024 without touching real estate and hopefully hit FCF breakeven along the way. If we assume EBITDA breakeven is a 2024 equation, you probably need to start tapping real estate in 2Q24. This to me is a pretty long runway to figure it out and drastically increases the odds unit growth comes back.

There is also a lot of flexibility here. CVNA is a great partner to ALLY (more on that next) and if they can hit EBITDA breakeven perhaps restricted cash can come down in a more stable macro environment. Ernie Garcia II famously sold billions in stock and could step in to save the company from APO with RE purchases, equity raises, etc. A stock price back at $50 creates a $7.5b market cap or so where 10% dilution can nab you a year to break even and still provide stellar returns. CVNA could do a PIK restructuring of the bonds which would also be dilutive but again attractive. Dilute me out 50% and I’m still getting a 20% IRR at least, so just live guys.

All in all, a common misconception with situations like this is perception of flexibility. Carvana is not by any means a share losing uneconomic business. If they can show the market that, then just about every problem gets fixed by hitting neutral EBITDA and just scaling/raising into your interest expense. If CVNA lives it’s still the business that was posting 100% growth vs market, not a Sears. I really just can’t see anything breaking until 2024 at the earliest. This gives them about a year to get back to levels of AEBITDA/unit they’ve already achieved! Hardly a herculean task despite what the market seems to think.

5. ALLY

A natural follow on to the liquidity question is ALLY. Simply put, if ALLY pulled the MPSA or IFSA tomorrow (not how that works but lets pretend) CVNA could conceivably be immediately insolvent. I really really really do not think this is a meaningful risk.

On the IFSA side this is a floor plan backed by actual cars. Funding is determined by wholesale value of the car which CVNA turns around and sells retail, so there is a pretty good margin of safety in there already. Additionally, ALLY has steep restricted cash requirements on CVNA, requiring ~$700m in restricted cash at $2.2b of utilization. If we assume a $3,000 spread wholesale to retail ($22k to $25k) then at max utilization you’d have $2.5b of inventory retail value as well as $700m in cash. For ALLY to be remotely close to underwater here would require used vehicle prices to fall >40% over night. ALLY also has a similar floor plan agreement with VRM, who is frankly in a much worse spot than CVNA. The asset backed nature of this agreement as well as restricted cash make it rather low risk for ALLY. ALLY also renewed this agreement about 2 months ago, so I doubt there has been a drastic underwriting shift since (if anything CVNA is in a better liquidity position following cuts.)

Speaking of renewals, the MPSA agreement is likely going to reach capacity in Q1 of 2023. For those unfamiliar, this agreement is what allows CVNA to sell loans to ALLY. If this agreement were to not be renewed, CVNA would likely need to find another avenue to sell loans, which could bankrupt the company or come with a significant headwind to GPU. On the bright side, I really don’t see why this would happen for a couple of reasons.

1. If you look at securitization filings CVNA loans tend to outperform all of their peers in regards to delinquencies and loss rates.

2. CVNA is being paid about a 7-8% GoS for these loans (similar to market rate)

3. ALLY doesn’t have to underwrite these loans

This relationship is great for ALLY! They’re getting more performant than average paper with almost no overhead at market rates! Of course it is possible that the auto ABS market freezes or ALLY doesn’t want auto paper any more, but this would be functionally unprecedented even in 2008/2009. History isn’t always a great guide, but steering defaults into the realm of worse than 2009 I’d imagine is a pretty low chance. ALLY honestly is just about the least worrying thing with CVNA.

6. Fragility

Really my largest concern with Carvana is fragility. Intellectually it’s pretty obvious to me they can do what I pencil out. They have also done it before! The question really is can EG3 manage the business in such a way that nothing breaks while cutting? CVNA is a very interdependent business. When you start breaking IRC’s or logistics or market hubs operations, there are flow through effects. You have to be careful you aren’t messing anything up too bad while rapidly dropping costs. This is absolutely a tough exercise and an unknowable risk.

In the past management has executed well, you don’t sell 450k cars a year being clowns, but it remains to be seen if they have the chops to weather the storm. If the ship veers too far off course and macro really tightens, there is a very real possibility we see CVNA owned by APO and not shareholders. If you hate money but enjoy intellectual stimulation and would like to follow the story, CVNA is a great way to raise your cortisol levels for the next 6 months. At this point everything makes sense in the models, it’s just strapping in for the ride and wishing Ernie the best of luck.

Conclusion:

All in all, I’ve gone through the high level of how I think about the above while trying to avoid diving too deep into details. Rest assured I have gone through securitization filings, I have spoken to current and former employees. I have poured through their regulatory filings. I have seen the expert networks, I have checked the alternative data, I have come at this thing from just about every angle I can think off and I really do believe it works.

I’d encourage all readers to make a liquidity model and see that 2023 looks fine. I’d encourage all readers to go through securitization filings to verify loan performance. I’d encourage all readers to be exceptionally self-critical about what can go wrong here. I’d encourage all readers to absolutely internalize the fact this thing can go to 0.

Some would say don’t buy things that can zero. Some would say buying a levered car retailer into a likely recession is silly. Said people are likely right on many occasions. That being said, you don’t get differentiated returns from following the herd all the time. At the end of the day investing is you against Mr. Market. There will come a time you have to nut up and put your conviction to the test. This happens to be one of those times, and thus the rewards to be reaped if correct are vast. Maybe I’m an idiot, maybe I’m not, we will just have to see.

Anyways, for those looking to follow Carvana I’d recommend the following:

1. Data Tracking: Jason Hartman, Alternative Alpha, YipitData, M-Science, or this data set for unit volumes/ASP/etc data tracking. Jason and Alternative Alpha will be cheaper and Twitter based, Yipit and M-Science more all encompassing but quite pricey.

2. Securitizations: Finsight for ABS transaction information. SPGI for pre-sale reports of ABS transactions. Edgar database for monthly securitization performance, and InPractise for a good crash course on CVNA ABS.

3. Glassdoor/LinkedIn/Indeed/Reddit/Twitter/Facebook for employee information

4. Public property records to track RE transactions and to do your own underwriting of RE value.

5. A brain to consider the above in a critical manner.

6. A bible or other such tool to get yourself through after you exhaust the above.

Thank you dear readers and hopefully I can be of help. I always appreciate your sharing of my writing, as the higher the audience the more likely I hear a good view on why I’m wrong! I have had a multitude of productive discussions and if you DM me I will try my best to get back to you. If you’d like to buy me a burrito every now and then, feel free to subscribe. All you get is my love, but that’s priceless.