Originate to Sell: The Fragile Foundation of Upstart’s Revenue Surge
When the tide’s coming in, everything floats and in bull markets blind monkey’s can throw darts and hit bullseye’s. SPACs with no revenue. Meme stocks with no plan. Mortgage companies left for dead. Retail department stores with high short interest. Clothing brands with new spokesmodels. Crypto tokens with cartoon mascots. EV startups with one prototype and a $30b valuation. People putting nuclear reactors on the moon. Companies promising to reinvent an industry with nothing more than a buzzword and an investor deck. Even digital dollar bank accounts trade at stupid multiples.
When money is easy, everything gets a bid and I’ve seen this movie enough to know the ending: the tide eventually goes out. It always goes back out. And when it does, you find out which stories are fully clothed and which aren’t wearing their trunks. The smart investor’s job isn’t to admire the froth but instead to understand the risk hiding underneath it. Every position in your book should have a “what happens when the music stops” chapter. This doesn’t mean ignore the bullish vibes, it just means thinking about the downside if the narrative flips and the world doesn’t look so rosy all of a sudden.
Which brings us to Upstart.
TL;DR: Upstart just posted blowout Q2 numbers,102% YoY revenue growth (off easy comps), $2.82B in originations, and 92% fully automated loan closings, delivering exactly what management promised on the tech and automation front. More and more AI originated loans. But the engine driving that growth is an originate-to-sell model that depends on private credit funds and bank buyers with a strong appetite for high-yield consumer loans. That plumbing runs hot in a credit bull market, but it can seize fast if buyers pull back or losses spike, making the business (and the stock) highly cyclical and vulnerable in a downturn.
Upstart is having a moment right now, for real. Q2 2025 revenue came in at $257.3 million, up 102% year over year, smashing the Street’s $226 million consensus. It was beats all across the way. Platform originations hit $2.82 billion, up 154% YoY. GAAP profitability returned with $4.5 million in operating income. Heck, they funded 372,599 loans, up 159% YoY, with loan conversion climbing to 23.9% from 15.2% a year earlier. Contribution margin held at 58%, adjusted EBITDA hit $53.1 million at a 21% margin, and fee revenue reached $240.8 million. CEO Dave Girouard told investors, “We’ve now put together four consecutive quarters of accelerating growth, which is something we haven’t seen since before the rate hikes began.” These are the kind of numbers that make growth investors lean forward. And make the YOLO buyers foam at the mouth.
Management has been telegraphing their strategy for years: automation, technology, and AI as the keys to the model. Originate loans cheaper, faster, and with a better borrower experience than anyone else. And on that side of the equation they delivered. In Q1 2025, 92% of loans closed fully automated from rate request to funding, no human touched them. Girouard said, “Our automation levels are at record highs, and we believe we’re the only lender operating at this scale with this level of efficiency.” Factually true. And the trend is good going back a bit further, loan conversion has steadily moved up, hitting 19.3% in Q4 2024 versus 11.6% a year earlier, and rising again in Q2 2025. In my integrity-checker framework, that’s a clear win. They said automation would be a core edge, and they showed it in the numbers. They have turned themselves into the fastest, most tech forward, easiest place to get a debt consolidation loan.
Here’s the important thing though, Upstart doesn’t hold most of these loans. It originates them, collects a fee, and sells them to outside balance sheets hungry for yield. Blue Owl’s alternative credit strategy signed up for $2 billion of forward-flow capacity over about 18 months. Fortress agreed to more than $1 billion. Over 100 banks and credit unions now buy from the platform. “We continue to broaden and deepen our funding relationships,” CFO Sanjay Datta said, “and we’re pleased to add another multi-billion-dollar institutional buyer this quarter.” Read as: we are happy to have people willing to gorge on paper from people who are going from revolving 28% debt spread amongst 6 cards, to 1 card charging them 24%. And when risk appetite is high, this plumbing hums, but it can and will seize if buyers hesitate. Which is why loans held on Upstart’s balance sheet growing to $1.02 billion in Q2 from $815 million in Q1 spooked investors a little bit. The Company explained it (likely truthfully) as new products like Auto and HELOCs, but the world knows if Upstart can’t find outs, they’ll be eating future losses at scale. And heaven forbid, if forward-flow partners pull back, that number spikes, and a growth story becomes a capital burn down story.
Remember too that the typical Upstart borrower isn’t a payday-loan customer, but they’re also not sitting across from a JPMorgan private banker either. The weighted average FICO in a recent ABS deal of theirs was about 678. In recent quarters, 26% to 32% of borrowers have been in the 720-plus super-prime tier, but that also means that 70% plus are mid-tier credit profiles at best. Most are borrowing for debt consolidation and credit-card refinancing, a pitch Upstart makes explicitly. “The largest single use case for our loans remains consolidating high-interest credit card debt,” Girouard told analysts.
The average loan size is $8,300. The platform funds as little as $1,000 and as much as $50,000, but most land in the low-to-mid single-digit thousands. Terms are three or five years. APRs range from 6.7% to 35.99%, but recent securitizations show weighted averages around 23% to 24%. The minimum income to apply is $12,000 per year, and the model ingests employment and education alongside bureau data. The hook is speed: rate in minutes, funds in a day. The sizzle is “AI based” lending, which again is factually true. But when you extend credit, the fact that you’re the fastest doesn’t mean you’re the best.
For the customer, it’s quick sure. But for investors, the credit math on fast debt consolidation loans is less comforting. Moody’s base-case cumulative net loss projection on a recent Upstart pool was 16.5%; in a severe stress, 58%. Datta was candid: “We design our models to withstand a range of economic conditions, but in a severe recession you would expect elevated loss rates.” For context, bank credit-card charge-offs are about 4% to 5% today and peaked just above 10% in the Great Recession. These loans are not the loans you want in any economic softness.
The structure itself is almost beautiful. Leverage AI to originate as much debt as possible, knowing that you can sell it off to buyers in massive chunks, clipping the fees, and moving on to the next batch of debt consolidators. And if this sounds familiar, it’s because the business model rhymes with pre-2008 originate-to-distribute. Not in fraud, just in mechanics. Think Countrywide in the Margin Call sense: produce greater and greater quantities of loans, sell to “willing buyers at fair market prices,” and avoid being left with the exposure. Like a coldly capitalistic game of musical chairs. In good times, it’s an elegant machine. In bad times, origination stalls, buyers go on strike, and the business model grinds. “Our business is highly sensitive to funding market conditions,” Datta acknowledged, “and any material disruption would directly impact our ability to originate.” That’s the risk most people underprice, especially the growth buyers that don’t even know what they do.
Those same growth investors will pull up a chart and see 2021 and remember the good old days of massive revenue growth, but it’s worth remembering that in 2021, Upstart’s growth was supercharged by stimulus money sloshing through the system. Triple-digit loan growth was a given. The multiple went parabolic. And that’s when we got the now-famous Mark Minervini interview, asked what the Company literally did after he pounded the table bullish, he pretended his connection dropped. “What’s that? Sorry, our connection is breaking up”. So where am I at with Upstart? The setup today feels different, but the rhyme is there. The macro is bending under a strained consumer, unemployment has ticked up, inflation has eaten into household cash flow. Credit-card balances are at record highs, and delinquencies have been climbing. Maybe leveling off. But I can promise you the mid-tier borrower is not flush with cash. And they are most certainly not prepared for any whiff of economic downtimes.
On top of this, the Q2 beat was real in the numbers, but a lot of it was optics and timing. The comp was soft, $127 million revenue in Q2 2024, and originations were juiced by partner appetite who have too much dry powder and madates to put it to work. The valuation, at roughly a $6.5 billion market cap, 7.5× trailing sales, and about 5× 2026E sales, is frothy for a pro-cyclical model this tied to funding sentiment and credit cycles.
Mechanically, if credit starts to deteriorate, two things happen quickly. Origination stalls, conversion drops, marketing spend becomes less efficient, volumes contract. And buyers go on strike, forward-flow agreements dry up, ABS prints get smaller or more expensive, and loans pile onto the balance sheet. Both kill the linear growth Wall Street loves to extrapolate. This is a model that works in a credit bull market. Just like the one that Countrywide was taking advantage of in 2005-2007.
And this isn’t just an Upstart story, it’s a private credit story. When the bid is strong, yield-seeking capital will buy just about anything that clears a hurdle rate, whether it’s pristine collateral or mid-prime consumer paper. That same appetite that fuels Upstart’s originations is one of the reasons I think names like Blue Owl are vulnerable in a downturn. The private credit bid has extended far beyond consumer loans into corporate credit, real estate, and specialty finance. In good times, that bid supports asset prices across the board. In bad times, it evaporates faster than public market liquidity, and with it, a lot of the mark-to-model valuations that have been built on the assumption that the bid is permanent.
I’m personally short UPST right now and have been for a week or so. I can admire the execution, because they continue to deliver on what they promised, and still fade the stock. Risks to being short are many: credit stays good, the government kicks in a big fiscal stimulus program, short rates come down a lot, private credit receives more mandates for paper, and so on.
But remember, Upstart is an AI-enabled loan factory. The AI part makes the process slick. The loan part makes it cyclical. In a benign macro, everything works. In a stressed macro, everything stops. And when that happens, it doesn’t just stop for Upstart, it stops for the private credit funds that have been bidding up every asset class in sight, from consumer paper to corporate credit. That bid has looked unstoppable in this cycle, but I’ve lived through enough turns to know that even the strongest bids vanish when the tide goes out.
Add on top of this the fact that capital rushed into this type of Company in droves over the past 2 weeks as investors actively sought out the worst balance sheets, with lots of short interest, that were severely beaten down. This capital though is fleeting. This capital is not loyal. Like a locust seeking new fields to infest, the second the stock stops going up, they move on to the next one.
It’s your job as an investor to make sure you don’t get caught holding the bags when the tide goes out. Kudos to the UPST team for executing. Their business model actually does as much as it can to take risk off their balance sheet and put it on someone elses. But the facts remain that this one is mega-cyclical only thriving and growing when the credit tide is high.
Until next time,
Victaurs
How do you view PGY vs UPST, valuation is quite a bit lower and, while it seems maybe a bit more diversified than UPST, it's still credit beta driven - thoughts?