A securities class action lawsuit alleges Upstart Holdings misled investors about its Model 22 AI underwriting system. The complaint says the model overreacted to negative macroeconomic signals, suppressed loan approvals, and overstated loan defaults during the second half of 2025. Shares dropped 9.71% in one day when the issue surfaced in early November. If you follow financial headlines, this is now the Upstart story.
But lawsuits about what happened in 2025 don't change what the business does in 2026.
The market is still pricing the old risk profile - an AI lending platform whose core model broke under stress. But the operating numbers already say something different. Revenue grew 64% to $1 billion in fiscal 2025. In Q1 2026, it grew another 44% year-over-year to $308 million. Upstart's full-year 2026 guidance calls for approximately $1.4 billion total revenue, $1.3 billion revenue from fees, and $294 million Adjusted EBITDA. Management says that Adjusted EBITDA is expected to be weighted toward the second half of 2026, as origination volume grows and operating expenses leverage over a larger revenue base.
That is the inflection. Revenue is compounding from a real base, not from a recovery. And the path to positive cash flow is the first time Upstart has had a credible one.
Let's be clear about what went wrong. The lawsuit centers on Model 22, Upstart's AI underwriting engine. During mid-2025, when macroeconomic indicators deteriorated, the model allegedly overcorrected: it reduced loan approvals and raised borrower interest rates beyond what warranted. The conversion rate - the percentage of visitors who get a funded loan - dropped from 23.9% in Q2 2025 to 20.6% in Q3 2025, a 330-basis-point decline that directly cut originations. The complaint alleges Upstart knew about this sensitivity and didn't tell investors.
Here's the thing the complaint doesn't address. By Q4 2025, the conversion rate had stabilized at 19.4%, and Upstart originated 455,788 loans that quarter - up 86% year-over-year. The full-year total was 1,497,149 loans originated, up 115% YoY. The model calibration problem that supposedly broke the business in Q3 didn't break it by Q4. Originations accelerated anyway.
The earnings miss is the real tape pain right now. Q1 2026 EPS came in at $0.30, missing analysts' estimate of $0.38, and net loss widened to $6.6 million, compared to ($2.4) million in Q1 2025. The stock sold off 15.8% on the report. Investors focused on the miss, not the $308 million in revenue that beat estimates. That's the classic pattern: the tape reacts to the quarterly disappointment while the annual bridge stays intact. Upstart still guides to $1.4 billion in revenue for 2026 - up 40% from the $1 billion it just completed.
The cash-flow picture is where the story gets complicated. Upstart burned -$148.1M in free cash flow during fiscal year 2025, and total debt rose to $1.90 billion at year-end 2025. That is not a clean balance sheet. But the $294 million adjusted EBITDA guidance for 2026 - if it materializes - would be a material improvement over the losses the company posted through most of its public life. The gap between that EBITDA target and the $1.9 billion debt service obligation is the question. Investors need to see whether operating cash flow narrows the burn materially, and whether the company can begin paying down principal without starving growth. I can't build a target off of projected EBITDA alone when FCF was negative and debt is rising. The financial bridge to the stock's upside requires positive free cash flow, and that proof point is still two quarters out.
What about the lawsuit itself? Securities class actions are expensive distractions, but they rarely break operating companies. The alleged period - May 14, 2025 to November 4, 2025 - is already behind. The company isn't changing its business model over what plaintiffs say it should have disclosed six months ago. Settlement risk exists, but there's no evidence the litigation will impair origination growth, bank partnerships, or the Model 22 system itself. Most of these cases resolve for amounts that matter to lawyers more than to earnings power.
This isn't about dismissing the legal risk. It's about separating tape pain from business pain. The operating setup is improving: revenue is compounding, originations are accelerating, conversion rates have stabilized, and management has put a real EBITDA number on the table for the first time. The cash-flow question remains open. Upstart needs to prove it can convert that EBITDA into free cash flow, not just accounting profit.
If you're holding the stock, the thesis to watch is whether Q2 and Q3 2026 show the EBITDA ramp management described - weighted toward the second half, as origination growth and cost leverage kick in. If those quarters deliver, the market's focus on the lawsuit and the Q1 EPS miss looks increasingly stale. If the EBITDA guidance proves optimistic and the cash burn doesn't narrow, the whole growth story needs rethinking regardless of litigation.
The setup is clear enough to say what proves it wrong: if Q2 2026 shows no meaningful progress toward the $294 million adjusted EBITDA target and free cash flow remains deeply negative, the thesis breaks. Cut without ego. The inflection won't happen on the back half alone, and holding through that disappointment costs more than the exit fee.
But right now, the market is still pricing a company whose model broke. The numbers are starting to look like a company whose model fixed itself and kept growing.