One way to frame Carvana (CVNA) is that of a loan originator that sells cars. Over 50% of CVNA’s gross profit is from finance and other income, of which the majority is driven by the accounting gains from selling the loans originated.

Source: In Practise, CVNA 10-K
Source: In Practise, CVNA 10-K
Source: In Practise, CVNA 10-K
Source: In Practise, CVNA 10-K

We’ve been studying CVNA’s finance model to understand the potential durability of finance GPU and specifically the gain on loan sales. Over the last few weeks, we've published various pieces of content on CVNA covering different aspects of the business:

The CVNA story has moved incredibly fast; the equity has declined ~90% from ATH’s and the company has a completely different capital structure post-ADESA acquisition.

This Weekly Analysis aims to explore CVNA's financing model and provide a balanced view of the potential advantages and future challenges originating profitable loans going forward.

Vertically Integrated vs Indirect Finance Models

Dealers, lenders, and investors all work together to originate loans so consumers can purchase cars. In the indirect finance model, the most common in the industry, dealers outsource the origination of the loan to external lenders like Santander, Exeter, or Westlake who take on the risk of underwriting the loan. Dealers earn most of their margin from selling the vehicle and ancillaries and either pay or receive a fee from the lender to make the economics work for both parties.

The lender typically earns the largest profit in the value chain because they are taking the most risk underwriting the loan. Investors earn a risk-adjusted return determined by traditional credit drivers (rates, spreads, duration, etc).

CVNA's model is different in that it is vertically integrated; it acts as the auto dealer that buys and sells the vehicle and the lender that originates the loan. These loans are then securitised and sold to ABS investors.

Source: In Practise
Source: In Practise

KMX is a hybrid model: it originates prime loans internally but outsources Tier 2 and 3 loans to traditional lenders. This is why comparing Carmax and CVNA’s Finance GPU isn’t apples-to-apples; CVNA is doing more of the work for lower quality but more profitable loans and therefore should earn a higher Finance GPU than KMX. Also, CVNA finances ~80% of retail sales whereas KMX is closer to 40%.

CVNA vs CACC

Comparing CVNA to Credit Acceptance (CACC), the US subprime auto lender, helped us understand CVNA’s loan origination business. CACC’s business model is based on reducing adverse selection in the indirect finance model; put simply, dealers outsource loans to CACC which shares the lending profits with the dealer over time.

This incentivises dealers to sell cars that customers can afford. Typically in the indirect model, dealers may receive a fee for effectively generating leads for lenders and even markup the loan to the customer, but they have no interest in the underlying loan performance.

This can distort incentives and encourage dealers to increase the price of the vehicle to inflate their metal margin. A higher ASP increases the loan-to-value which increases the monthly payment and puts the customer at a greater risk of default. CACC aims to reduce this adverse selection by 'holding back' some of the loan profit to share with the dealer.

CVNA goes one step further and eliminates all adverse selection by being the dealer and the lender. It owns the inventory and the relationship of the customer which has multiple benefits that drive loan performance. Firstly, CVNA owns and understands the value of the collateral and the customer’s affordability of the loan. The IRC inspection process also helps improve the collateral value which improves recovery rates. Finally, because CVNA is the dealer and the lender, it eliminates the fees exchanged to make the economics work for both parties.

CVNA’s origination advantage also seems to be visible in the underlying loan performance. For example, from 2016-20, CVNA had around ~300bps lower net charge-offs versus Exeter for a similar pool of loans. (Also, note the decline in the charge offs over the last 5 years. More on this later).

Source: In Practise, CVNA ABS Securitisations
Source: In Practise, CVNA ABS Securitisations

Because CVNA is the dealer and the lender, it doesn’t mark up the loan which increases the cost to the customer. CVNA also requires larger down payments which reduces the term, increases the equity for the consumer, and reduces the default rate. This leads to ~10% lower LTV’s which improves the affordability of the loan for customers.

This is what the Chief Risk Officer at US Auto Sales, a deep subprime buy-here-pay-here dealer said about CVNA:

When you think about performance, you need to consider the additional premium that a customer is paying from a monthly payment perspective because that's an APR added on top of your risk-based pricing…the difference between indirect to direct would be that the same customer pays a slight premium on the loan. This translates into a higher risk for a customer from a net charge-off perspective or performance because they are paying a premium for the same car. - Chief Risk Officer at US Auto Sales

For a given non-prime FICO score, we estimate CVNA has ~100bps lower annualized losses per loan than Exeter, Westlake, and other subprime lenders. This higher loan quality is what bulls argue the ABS market is paying a premium for.

Gain on loan sale

CVNA originates and sells the loans into the ABS market; mainly prime loans are sold to Ally in a forward flow agreement and non-prime loans are sold to the ABS market. CVNA retains a 5% residual of all loan tranches on its balance sheet and books a ‘gain-on-loan sale’ after selling the remaining 95% to investors.

One of the bear cases for CVNA has been that the gain-on-loan sales are artificially inflated because the loans are sold to related parties. Mark Walter, who runs Guggenheim and owns ~9m CVNA common shares, is supposedly a link to the undisclosed majority buyer of CVNA’s securitisations. This could be true. However, it seems unlikely to us that Ally and the whole ABS market were also artificially bidding up CVNA loans.

Another bear case is that the cumulative origination cash proceeds is less than the originations of receivables on the cash flow statement. Thus, the loan economics don't work. However, this is likely because CVNA has been growing so quickly; the 2021 vintage of origination receivables is on a larger number of units than the cash proceeds received in 2021. As growth slows, we could see cash proceeds outweigh origination receivables.

Source: CVNA 10-K
Source: CVNA 10-K

Origination Challenges Today

Carvana may have an advantage originating loans but the business must endure for the FCF to flow through to shareholders. The company recently raised $3.25bn in 10.25% 2030 notes and $1bn in equity to finance ADESA just as both the used car and capital markets turn over.

Access to capital markets is crucial for CVNA because it needs to recycle capital by selling existing loans to fee up cash to originate new loans and sell more cars. This quote from our recent Investor Dialogue highlights the biggest risk to CVNA:

If there's something I fear is that the securitization markets dry up. That’s the scenario where your Carvana is worth zero if the equity gets wiped out because their Ally financial facility covers what would be a third of their volume. They can't support the fixed costs of the business just flowing everything to the MPSA, so they need the securitization markets. They just have to cycle the capital in this business, but if the securitization markets for auto loans dries up, and nobody can cycle capital, it's not just going to be Carvana. We know Carvana's loans are generally performing better than their peer group. If they dry up for Carvana, what happens to the auto market in general? - IP Investor Dialogue

In Q1, CVNA reported a $412 decline in Finance GPU over Q1 21. Management believes it was the speed of the increase in rates that drove a $600 GPU decline.

Looking sequentially, the rapid rise in benchmark interest rates and widening credit spreads had a significant impact on the spread between funding costs and origination interest rates in Q1 versus Q4. This increase in spread had a more than $600 impact on Other GPU in Q1. We expect this spread to move toward more normalized levels over the coming quarters. - CVNA CFO, Q1 2022

Higher ASP’s also drove a $150 increase in GPU. Management believes the $450 net decrease in finance GPU was a one-off impact driven by the rapid increase in swap rates that temporarily compressed CVNA spreads.

As we originate the loans our customers use to buy parts from us and then sell them later, interest rate increases between initially showing our customers their financing terms and ultimately selling those loans lead to a reduction in the value of the loans we sell, which had the impact of reducing Other GPU. These factors combined to lead to a clear step back in our financial results. While this isn't what we are shooting for, it is straightforward to understand, and it suggests straightforward solutions. - CVNA CEO, Q1 22

Although there may well be one-time impacts on spreads from Q1, the table below shows the compression in excess spreads from CVNA’s prime and non-prime securitisations over the last year.

For prime loans, the excess spread has declined from 635bps in March 2021 to 350bps May 2022. This has led to a 30% decrease in the gain-on-loan because the coupon the ABS market is demanding has increased from 0.78% to 4.29%. CVNA hasn’t securitised any non-prime loans in May but the subprime customer is currently facing challenges.

Source: In Practise, CVNA, Exeter. The GOS would need to be multiplied by the % of vehicles CVNA finance and adj for Ally MPSA GOS to reconcile with CVNA reported numbers
Source: In Practise, CVNA, Exeter. The GOS would need to be multiplied by the % of vehicles CVNA finance and adj for Ally MPSA GOS to reconcile with CVNA reported numbers

The movement in the prime credit spreads highlights the potential risk for CVNA's finance GPU.

Source: In Practise, CVNA
Source: In Practise, CVNA

The average loan amount and excess spread drive the total profit per loan. Both of these are moving in the wrong direction for CVNA.

Investors are demanding higher rates of return and better collaterisation terms which will put pressure on CVNA to increase the loan APR to maintain profitability. If management do pass-on higher cost of funds to customers, the growth in unit sales will decline further which will deleverage CVNA even further.

Because CVNA is a dealer-lender for both prime and subprime customers, it can better manage the mix between customer types to stablise overall finance GPU: if subprime customers are hit hardest, they can flex pricing to attract more prime customers. Management may already be doing this given Q1 unit sales to customers with 700+ FICO increased 50% yoy.

Although management have levers to pull to manage the loan spread, it can’t stop mean reverting used car prices.

If we see used car prices normalise to $19,000, after adjusting for somewhat normalised credit rates, we still see a healthy premium per loan for CVNA in both prime and subprime. However, it’s around 50% of the premiums per loan earned over the last 18 months.

Source: In Practise
Source: In Practise

CVNA is now a stressed credit with over $6bn in debt in front of shareholders. If we take Apollo's position in the 10.25% bonds, it can give us a sense of how the credit market values CVNA. The 10.25% bonds issued 2 weeks ago now yield ~13% and the ABS notes are also trading at a discount. In the event of bankruptcy, at today's yield, Apollo will be made-whole at ~130% on the $3.25bn invested. This equates to $4.25bn plus the existing $3bn in debt = ~$7bn in claims in the event of bankruptcy.

If the recovery on the bonds is ~75%, this implies Apollo thinks it's worth at the very minimum ~$5.25bn to breakeven on their capital. It's likely worth multiples of this otherwise Apollo wouldn't be involved. If Apollo think it's worth $15bn+, the implied exit equity value with $7bn of debt is $8bn. As of today, the current market cap is $5bn. The equity seems to be priced for bankruptcy today.

We estimate CVNA has ~$2bn in pro-forma cash but with the ADESA unencumbered assets liquidity is closer to $4.5bn. CVNA burned ~$800m in FCF in Q1 22 alone.

We believe there are clear advantages to CVNA's model that deliver a superior customer experience, but the management team have a huge executional challenge to reduce burn and prove to the capital markets that the economics work. We didn't discuss the operational changes and improvements possible with ADESA but check out our Investor Dialogue which goes through the liquidity and potential operational risks in detail. Let us know if we've missed anything important about CVNA's finance business.