Same Movie, Different Asset: How Carvana and Opendoor Actually Make Money
Carvana already filmed this one. Opendoor looks like it's running the same script — buy the asset, survive the crash, then make the real money on everything around the transaction. A field guide for retail investors.
Not advice, a map. This explains how these businesses make money and why one looks like it’s copying the other. It is not a buy or sell call. Do your own work.
One sells used cars online. The other buys and sells used houses online. On the surface they’re in different universes — a car costs $30,000, a house costs $350,000.
Look closer and they’re the same machine pointed at two different assets. And the younger one, Opendoor, is running a script the older one, Carvana, already shot — including the part where it nearly dies in the second act.
flowchart TD
C["CARVANA<br/>buys & sells<br/><b>used cars</b>"] --> M{"Same machine:<br/>flip the asset, then make<br/>the real money on the<br/>financing & services<br/>wrapped around it"}
O["OPENDOOR<br/>buys & sells<br/><b>used houses</b>"] --> M
class M anchor;
Let’s take the proven one apart first, because once you see how Carvana actually makes money, Opendoor stops looking like a house-flipper and starts looking like a company copying a playbook.
Part 1 — How Carvana makes money
The story they tell you: a giant, friendly car flip
The pitch is simple. Carvana buys a used car (from a person trading in, or from an auction), reconditions it (inspect, fix, clean, photograph), lists it online with no haggling, and sells it for more than it paid — delivered to your driveway or dropped from one of those glass “car vending machines.”
flowchart TD A["Buy a used car<br/>from a person or auction"] --> B["Recondition it<br/>inspect, fix, clean, photograph"] B --> C["List online<br/>no haggling, delivered<br/>or from a vending machine"] C --> D["Sell for more than<br/>they paid: <b>the spread</b>"] class A anchor;
That gap between what they pay and what they sell for — minus the cost of fixing the car up — is the profit on the metal. Wall Street tracks it with one obsessive number: GPU, gross profit per unit. Think of GPU as the scoreboard for “how much do we make every time a car goes out the door.”
The scoreboard has three columns
Here’s the first thing most people miss. That ~$6,900 of gross profit Carvana made per car in 2024 didn’t all come from the car.
flowchart TD U["One car sold<br/><b>~$6,900 gross profit</b> (2024)"] --> R["Retail GP<br/>the markup on the car itself"] U --> W["Wholesale GP<br/>flipping cars they don't retail,<br/>at auction (that's the ADESA piece)"] U --> OT["<b>Other GP</b><br/>financing + warranties + GAP<br/><i>the high-margin part</i>"] class U anchor;
The car itself (retail GP) is the burger. It’s fine. But the money that makes Carvana interesting is in that third column — Other GP — and that’s the fries and the soda.
The real engine: selling you the loan, then selling the loan
When you buy a Carvana car, about 8 out of 10 buyers finance it — and most finance through Carvana. Carvana writes the loan, then doesn’t keep it. It sells that loan — to a big partner like Ally, or by bundling thousands of loans together and slicing them into bonds (a “securitization”) sold to pension funds and asset managers.
Think of it like a baker who sells you a whole pie, then turns around and sells the recipe and the next 10,000 pies to Wall Street — and books the profit on all of it today.
flowchart TD S["You finance the car<br/>~8 of 10 buyers do"] --> O["Carvana writes the loan"] O --> Sell["Carvana sells the loan<br/>to partners (Ally) and as<br/>bonds (securitized to investors)"] Sell --> G["<b>Gain on sale</b>, booked up front.<br/>In one recent 9-month stretch:<br/>~1/4 of all gross profit,<br/>and more than 2x net income"] class S anchor;
Read that last box again. In a recent nine-month stretch, the gains from selling loans were about a quarter of Carvana’s gross profit and more than double its net income. The car is the hook. The loan is the business. Add in extended warranties (VSC) and GAP insurance — classic high-margin “F&I” products every dealer loves — and you see it: Carvana is quietly a finance company wearing a car-retailer costume.
Why it can do this cheaply: vertical integration
Carvana owns the whole pipe. Its 2022 purchase of ADESA (a national network of auction sites, ~$2.2B) gave it physical locations it’s turning into “megasites” — one roof for reconditioning, wholesale auctions, and shipping. Owning the inspection centers, the logistics, and the auctions means every extra car costs less to process.
That’s a flywheel: more cars → fixed costs spread thinner → lower cost per car → better prices and more loans/services → more cars.
flowchart TD V1["More cars sold"] --> V2["Spread fixed costs<br/>over more units<br/>(megasites, logistics)"] V2 --> V3["Lower cost per car"] V3 --> V4["Sharper prices +<br/>more financing & services"] V4 --> V1 class V1 anchor;
Think of the infrastructure as a toll road. Building it costs a fortune. But once it’s built, every additional car driving through is almost pure margin. That’s “operating leverage,” and it’s the whole bet.
The part the script requires: a near-death experience
Carvana did not glide here. In 2022, it had borrowed heavily (including for ADESA), interest rates spiked, used-car prices fell, and the stock dropped ~99%. It became the most-shorted stock in America, priced for bankruptcy.
What saved it: in July 2023, an Apollo-led deal with bondholders swapped unsecured debt for secured debt and pushed the due dates out to 2028 — buying time. Then the cost cuts and the flywheel kicked in.
flowchart TD T1["<b>2017</b><br/>IPO, vending machines,<br/>grow at all costs"] --> T2["<b>2022</b><br/>big debt + rate spike<br/>stock -99%, most shorted,<br/>priced for bankruptcy"] T2 --> T3["<b>2023</b><br/>Apollo-led debt deal:<br/>swap unsecured for secured,<br/>maturities pushed to 2028"] T3 --> T4["<b>2024</b><br/>first full-year profit:<br/>$404M net, $1.38B adj. EBITDA"] T4 --> T5["<b>2025</b><br/>record units & GPU,<br/>~$2B+ adj. EBITDA"] class T2 danger; class T5 anchor;
The asterisks (read these before you fall in love)
- It’s a family affair. CEO Ernie Garcia III runs Carvana; his father, Ernie Garcia II, controls DriveTime (and loan-servicer Bridgecrest) and is Carvana’s largest shareholder. A lot of business flows between related parties. That’s legal and disclosed — but it means “trust the numbers” requires trusting a tangled web. Short-sellers (notably Hindenburg in 2024) have attacked exactly this; Carvana disputes them.
- The profit engine depends on selling loans. If credit markets seize up, or buyers’ loans go bad faster than assumed, those upfront “gains on sale” can shrink fast. The engine runs on Wall Street’s appetite for the bonds.
- Still carrying real debt. The 2023 deal bought time; it didn’t erase the balance sheet.
Keep all three in your head. Now watch Opendoor do the same dance, two acts behind.
Part 2 — How Opendoor makes money
The same flip, with houses
Opendoor makes you a cash offer on your house in days, buys it, does light prep (clean, small repairs), and relists it on the open market. Its profit is the spread plus a service fee, minus what it spent holding and fixing the home.
flowchart TD A["Buy a house<br/>cash offer in days"] --> B["Light prep<br/>clean, minor repairs"] B --> C["Relist & sell<br/>on the open market"] D["Spread + service fee<br/><b>minus</b> holding & repair costs"] C --> D class A anchor;
Opendoor’s version of GPU is “contribution margin” — profit per home after the costs directly tied to selling it. Its target is a 5–7% contribution margin. In Q1 2026 it was 4.4% and climbing (it had been negative in 2025). Same scoreboard idea, different name.
The “fries” they’re now building
Here’s the tell that Opendoor is chasing the Carvana model: it’s bolting services onto the transaction, exactly where Carvana makes its fat margins.
flowchart TD H["One home transaction"] --> T["Title & escrow<br/>(bought OS National, 2019)"] H --> M["Mortgage<br/>(Opendoor Home Loans)"] H --> CL["Closing & escrow tech<br/>(Doma deal, 2026)"] H --> X["Insurance, warranty,<br/>moving — on the roadmap"] class H anchor;
If Carvana proved that the money is in the financing and services around the asset, Opendoor is trying to build the housing version: mortgage, title, escrow, closing, insurance. The catch — and it’s a big one — these are still tiny for Opendoor today. The home itself, and the risk of holding it, still dominate the P&L.
Opendoor’s near-death act (sound familiar?)
flowchart TD T1["<b>2014–2021</b><br/>iBuying pioneer,<br/>scale fast"] --> T2["<b>2022–23</b><br/>rates spike, housing freezes,<br/>big inventory write-downs, layoffs"] T2 --> T3["<b>2025</b><br/>Zillow partnership,<br/>new CEO, warrant dividend"] T3 --> T4["<b>2026</b><br/>revenue shrinking on purpose,<br/>but unit economics improving"] class T2 danger; class T4 anchor;
Same second act: borrow and grow, get hit by a rate shock, nearly break, then shrink and get disciplined. Notice Opendoor’s revenue is falling ($720M in Q1 2026 vs $1.15B a year earlier) — on purpose. It’s selling down aged inventory (homes sitting >120 days dropped from 33% to 10% of its book) and prioritizing margin over volume. That’s the Carvana 2023 move.
(If you want the deep dive on Opendoor’s warrant dividend — the free “coupons” it handed shareholders in late 2025 — that’s a separate piece.)
Part 3 — The playbook, side by side
Strip away cars vs. houses and both companies are running the same seven-step script:
flowchart TD P1["1. Pick a giant, fragmented,<br/>universally-hated transaction<br/>(buying a used car / selling a house)"] --> P2["2. Put it online:<br/>simple, no-haggle, branded"] P2 --> P3["3. Eat the spread on the asset<br/><i>thin margin, scary inventory risk</i>"] P3 --> P4["4. Get hit by a rate shock<br/>survive via restructuring + discipline"] P4 --> P5["5. Bolt on the high-margin money:<br/>financing + services"] P5 --> P6["6. Vertically integrate<br/>to cut cost per unit"] P6 --> P7["7. Operating leverage at scale<br/>→ real profit"] class P1 anchor; class P7 anchor;
Where each one is on the script today:
| Step | Carvana | Opendoor |
|---|---|---|
| 1–3 (build + flip the asset) | ✅ done, at scale | ✅ done |
| 4 (survive the rate shock) | ✅ survived (2023 debt deal) | 🔄 in progress (shrinking, disciplining) |
| 5 (bolt on financing/services) | ✅ this is the profit engine | 🔄 early — title, mortgage, closing being built |
| 6 (vertical integration) | ✅ ADESA megasites | 🔄 partial (title/escrow/closing) |
| 7 (profit at scale) | ✅ profitable, record margins | ❌ not yet — still losing money |
Carvana is at the end of the movie. Opendoor is somewhere around act 4–5, and we don’t yet know if it gets a happy ending.
The scoreboard, today
| Metric | Carvana | Opendoor |
|---|---|---|
| Asset | Used cars (~$30k) | Used homes (~$350k) |
| “Profit per unit” metric | GPU ~$6,900/car (2024) | Contribution margin ~4.4%, target 5–7% |
| Latest direction | Revenue growing fast (Q3’25 +55%) | Revenue shrinking on purpose |
| Bottom line | Profitable ($404M net, 2024) | Still a net loss ($173M, Q1’26) |
| The high-margin add-on | Financing — huge, proven | Mortgage/title — small, early |
| Near-death moment | 2022 → survived 2023 | 2022 → working through it now |
Why the analogy has limits (don’t over-trust it)
A house is not a car. The differences are exactly where the Opendoor bet gets harder, not easier:
flowchart TD
Car["<b>A used car</b><br/>~$30k · depreciates predictably<br/>standardized reconditioning<br/>deep, proven loan + bond market<br/>financed ~80% of the time<br/>you sell people many over a lifetime"]
House["<b>A used house</b><br/>~$350k · huge capital tied up per unit<br/>brutal holding + interest-rate costs<br/>mortgage attach is harder & regulated<br/>bought once every several years"]
Car --- VS{"Same playbook —<br/>much heavier, riskier asset"}
VS --- House
class VS danger;
- Inventory risk is brutal at house prices. Holding a $350k home while rates jump can vaporize the spread. iBuying already killed a giant competitor — Zillow shut its home-buying arm in 2021 after big losses.
- The financing engine is harder to copy. Carvana’s loan-sale machine is a deep, well-oiled market. Mortgages are bigger, slower, heavily regulated, and Opendoor is years behind on attach rates.
- Following a playbook ≠ winning the game. Carvana proves the model can work. It does not prove Opendoor will make it work.
What a retail investor should actually watch
For Carvana — is the engine durable?
- Are loan-sale gains holding up, or is profit leaning on Wall Street’s mood?
- Debt paydown and the related-party (DriveTime/Bridgecrest) picture.
- Can volume keep growing without GPU slipping?
For Opendoor — is it climbing the script or stuck?
- Is contribution margin marching to that 5–7% target and beyond?
- Is it getting to positive adjusted EBITDA, then real net income — or just shrinking?
- Are the adjacent services (mortgage, title, closing) actually growing into a Carvana-style profit engine?
- Dilution. Watch the share count and the warrants — the asset is capital-hungry, and capital usually comes from issuing more stock. (See the warrants explainer.)
The one-sentence version: Carvana showed that the way to win the “buy-the-asset-online” game is to stop selling the asset and start selling the money around it. Opendoor is betting it can run that same play in housing — and it’s currently somewhere in the hard middle of the story.
Sources: Carvana and Opendoor SEC filings and quarterly results (Carvana FY2024 8-K; Opendoor Q1 2026 8-K), company press releases, and public reporting on Carvana’s 2023 debt restructuring, the ADESA acquisition, and Opendoor’s OS National / Doma acquisitions. Figures are approximate and as reported at the time. Educational only — not investment advice.