Upstart's Biggest Risk Isn't Its AI Models. It's Who Funds the Loans.

Upstart shares are down 26% year to date and about 33% over the past year, despite first-quarter loan originations rising 61% and revenue increasing 44% year over year. The company announced plans to apply for a national bank charter, which could let it fund loans with deposits and reduce reliance on third-party lenders, with a launch targeted for early 2027 if approved. However, expenses climbed 45% in the quarter and Upstart posted a $7 million net loss, underscoring that profitability remains elusive.

Analysis

The bank-charter move is less a growth catalyst than a balance-sheet de-risking exercise. If approved, it should compress Upstart’s funding-friction discount, because the market currently prices a meaningful probability that partner pullbacks or tighter credit conditions will choke originations at the worst possible time. That said, a charter also changes the business from a fee-led platform into a more capital-intensive lender, which can improve revenue durability but usually lowers valuation multiples until investors see consistent credit performance through a full cycle.

The more important second-order effect is that a charter gives management an escape hatch from wholesale-funding dependence just as the credit cycle gets less forgiving. That could support originations in a downturn, but it also shifts the earnings debate from “can they grow?” to “can they underwrite and fund themselves without hidden duration or credit losses?” In other words, the market may initially reward the strategic optionality, then quickly reprice the name if deposit acquisition costs, net interest margin, or charge-offs look even modestly worse than expected.

Consensus still seems too focused on top-line acceleration and too relaxed about expense intensity. The recent operating leverage picture is fragile: the company is spending like a high-growth software business while the model is becoming more bank-like, and that combination usually forces a valuation reset unless margins expand fast enough to justify it. The biggest near-term risk is not a collapse in originations; it is a slow bleed where growth remains good but incremental dollars produce mediocre returns on capital, keeping the stock range-bound for months.

Contrarianly, the setup may be better for a tactical long than a structural one. The stock could continue to grind higher into regulatory milestones if investors view the charter as an AI/fintech credibility upgrade, but the cleaner trade is to fade strength once the approval process becomes visible and the market starts modeling actual capital requirements. The right question is not whether Upstart can originate more loans; it is whether the charter creates enough economics to offset the drag from funding, compliance, and credit reserves.

Key Decisions for Investors

• Trade tactically long UPST into charter-related headlines, but only as a 4-8 week event trade; upside is driven by multiple expansion on lower funding-risk perception, while downside is sharp if approval timing slips or management commentary implies heavier capital needs.
• Use UPST call spreads rather than stock for a defined-risk expression; prefer 3-6 month maturities to capture the regulatory narrative without taking full balance-sheet risk. The appeal is convexity if the market rerates the name on strategic optionality.
• Fade any post-rally strength in UPST with a medium-dated put spread once the charter story is fully priced; the bear case is not growth collapse but margin compression and expense creep causing the market to refocus on profitability.
• Pair long a more profitable fintech/credit platform against short UPST on valuation discipline if the sector bids up on AI enthusiasm; the goal is to isolate model-risk and funding-risk rather than take broad fintech beta.
• Avoid initiating a core long until there is evidence that operating expenses are decelerating faster than originations growth; without that, the stock remains a trading vehicle, not a long-duration compounder.