Andrew Boone Overpaid or Clueless?
Maybe Both. Read it You Decide
Andrew Boone recently came out with a $20 target on UPST. Everyone is entitled to their opinion. I found Andrew’s assessment reckless, lazy, and most of all uninformed. When making such a statement you would hope for accuracy. When I say lazy and uninformed, last stated price target of 20$ was 70$ back on May 10th 2022. As we all know UPST went down below 26$ 2024 when we went to buy. We rode it and got out between 67$ and 82$. Andrew did NOTHING! He never changed his 2022 target. We are back in it. The current environment is risk off investing (UPST is a risk on asset. We think risk on investing happens again once we see clear signs of lower interest rates from the Fed. You decide! Read On! Andrew, you should have stayed on the sidelines!
A Point-by-Point Destruction of Lazy, Uninformed Analysis on Upstart (UPST)
February 15, 2026 | Our Trade Desk
What follows is a point-by-point refutation of a recent bearish take on Upstart Holdings (UPST). Every claim made by the analyst in question was either factually wrong, demonstrably lazy, or both. We verified each point against Upstart’s actual Q4 2025 earnings report, the full earnings call transcript, SEC filings, and publicly available data. The score? The analyst went 0-for-5. Not a single claim survived contact with the actual numbers.
This wasn’t a close call. This wasn’t a matter of interpretation. This was someone who either did not read the earnings release, did not listen to the call, or did not understand what they were reading. Any of those three explanations is disqualifying. Let’s walk through it.
Claim #1: “Rising Credit Risk & Delinquencies”
THEY SAID: Cited a worsening credit environment, specifically noting “spooked” credit markets
due to rising delinquencies in the auto loan sector. This creates a higher risk profile for
Upstart’s AI-driven lending platform.
Let’s start with the obvious: this person looked at a headline about general auto loan delinquencies and concluded it applied to Upstart. That is the analytical equivalent of reading that restaurants are struggling and concluding that Chick-fil-A is going bankrupt. It reveals a complete failure to distinguish between industry-wide credit trends and company-specific credit performance—which is, you know, the entire thesis of Upstart’s existence.
Here’s what Upstart actually reported—the information freely available to anyone with the intellectual curiosity to, say, read the press release:
|
Metric |
What Upstart Actually Reported |
| Credit Performance | Every single vintage exceeded Treasuries by at least 270 bps |
| Auto Origination Growth | 5x year-over-year (grew 344% through early 2026) |
| Individual Vintage Floor | Every single vintage exceeded Treasuries by at least 270 bps |
| Auto Origination Growth | 5x year-over-year (grew 344% through early 2026) |
| Q4 Auto Funding by 3rd Parties | 92% funded via third parties — partners are not running FROM auto,
they’re running TO it |
| Model Improvement (24/25) | Default rates 0.8% LOWER while simultaneously INCREASING approval
rates |
| Roll Rates (delinquency→default) | 15% year-over-year IMPROVEMENT through personalized collections |
| Automation Rate | 91% of funded loans fully automated — up from 69% in Q4 2019 |
So to recap: Upstart’s credit performance is beating U.S. Treasuries by 608 basis points across 12 consecutive vintages. Their AI models are producing lower defaults at higher approval rates. Their auto lending partners—the people actually writing the checks—are so “spooked” that 92% of Q4 auto originations were funded by third parties, with 13 additional partners signed for 2026. Capital partners don’t sign multi-year commitments to asset classes they’re “spooked” by. That’s not how capital markets work. That’s not how anything works.
But sure, “rising credit risk.” Maybe next time try reading past the Bloomberg headline before writing an investment thesis. The information was right there. On the front page of the press release. In large font.
| VERDICT: The analyst confused industry-average auto delinquencies with Upstart-specific credit performance. Upstart’s entire value proposition is that its AI outperforms industry averages. That is literally what the company does. Citing industry-average data to argue against a company whose business model is outperforming industry averages is the kind of circular reasoning that should disqualify you from public commentary. |
Claim #2: “Worsening Macro Environment / Weak Labor Market”
THEY SAID: A weakening labor market heading into 2026 was identified as a significant threat to Upstart’s borrower base.
This one is almost endearing in its laziness. Yes, macro conditions have deteriorated. No one disputes that. The UMI (Upstart Macro Index) sat at roughly 1.4 for most of 2025. For reference, that means a statistically identical borrower is 43% more likely to default today than they were in 2021, when the UMI was 0.8. So the macro backdrop is objectively harder.
And yet—here’s where our analyst apparently stopped reading—Upstart’s 2025 results surpassed 2021’s peak performance. Let me say that again for the people in the back: in a macro environment that is 43% more difficult for borrowers, Upstart delivered better credit performance than it did at the peak of the easiest lending environment in a generation. That is the entire point of AI-driven underwriting. The model adapts. It re-prices risk. It tightens where it needs to tighten and expands where the data supports expansion.
CTO Paul Gu specifically addressed this on the call. Model iterations 24 and 25 produced 0.8% lower default rates than previous models while simultaneously increasing approval rates. The model is getting better faster than macro is getting worse. That’s not a vulnerability—that’s a competitive moat widening in real time.
And the borrower base itself? Management explicitly noted a deliberate shift toward prime borrowers and larger loan sizes. Higher credit quality, higher lifetime value, more stable repayment. Their small-dollar relief loans showed “exceptional credit performance with over 100% sequential growth. The product-market fit across personal, auto, and home segments was described as “strong”—not by some cheerleader on Twitter, but by the people running the business, backed by the numbers they just reported.
Here’s what the analyst might have known if they’d bothered: Upstart’s guidance for 2026 assumes a constant UMI—meaning they’re not projecting improvement. They’re guiding to $1.4 billion in revenue and a 35% CAGR through 2028 while assuming macro stays exactly this difficult. That’s called conservative guidance. You’d know that if you read the transcript. Which, clearly, you did not.
| VERDICT: Saying “worsening macro hurts Upstart” while ignoring that Upstart just outperformed its own best year in a 43% harder environment is like warning someone their umbrella won’t work in the rain while they ’re standing in a hurricane, bone dry. The model adapts. That’s the product. You’re supposed to understand the product before writing about it. |
Claim #3: “Lack of Revenue Clarity / Won’t Hold Loans on Balance Sheet”
THEY SAID: Noted Upstart’s reluctance to hold loans on its own balance sheet, combined with a decreased willingness from capital market participants to fund originations, limits the company’s growth potential.
I genuinely cannot tell if this person is confused or contrarian for sport. The entire investment community has been demanding Upstart reduce its balance sheet exposure for two years. That was the number one concern from analysts, institutional investors, and everyone else with a functioning Bloomberg terminal. The complaint was: “You’re a marketplace lender acting like a balance sheet lender, fix it.”
So Upstart fixed it. They reduced loans on their balance sheet by 20% quarter-over-quarter in Q4, from $1.2 billion to $985 million. And now this analyst’s critique is that they… reduced their balance sheet? That’s the bear case? “They did the thing everyone asked them to do and they succeeded at it”? That’s not analysis. That’s sour grapes from someone who missed the move.
As for “decreased willingness from capital market participants”—this is not just wrong, it is the photographic negative of reality. Here are the facts:
| Capital Markets Metric | Q4 2025 Actual |
| Auto/Home loans funded by 3rd parties | 70% in Q4 (up from near-zero during incubation) |
| Auto-specific 3rd party funding | 92% of all Q4 auto originations |
| Active funding partners (Auto/Home) | 11 different partners in Q4 |
| NEW partners signed for 2026 | 13 additional partners |
| HELOC partner progress | Majority of Q4 production taken by newly signed partners |
| Balance sheet loan reduction | -20% quarter-over-quarter ($1.2B → $985M) |
| Management characterization | “Achieved liftoff” in capital markets for secured products |
| Partner type expansion | Banks, credit unions, AND institutional capital — including larger
banks |
CFO Sanjay Datta addressed this directly. He said the company is “very deliberate” about private credit partners, that these partners conduct deep diligence, and that co-invested loans are “performing better and better.” The contribution margin decline to 53%? Also addressed on the call: it was intentional. They’re accepting lower take rates on secured products to capture a massively larger total addressable market in auto ($760B) and home equity ($1.4T).
You’d know all of this if you read the earnings call transcript. It’s publicly available. It’s free. You can find it by typing “Upstart Q4 2025 earnings call transcript” into any search engine. I just did it in four seconds. Apparently that was four seconds more than our analyst was willing to invest.
| VERDICT: The analyst claimed capital markets are unwilling to fund Upstart. In reality, 24 capital partners are actively funding or signed to fund in 2026, auto funding hit 92% third-party, the balance sheet shrank 20% by choice, and management described the capital markets position as “liftoff.” The claim is not partially wrong or directionally wrong. It is 180 degrees wrong. It is the opposite of what is happening. |
Claim #4: “Capital Market Dependence / Limited Visibility”
THEY SAID: “Limited visibility” into when capital markets would fully reopen or stabilize for Upstart, making the current risk/reward profile appear unbalanced. This is the same claim as #3 wearing a slightly different hat, which suggests our analyst had exactly one idea and decided to present it twice for volume. But let’s address the “limited visibility” assertion specifically, because Upstart just gave the market more visibility than virtually any fintech company in public markets:
| Transparency Enhancement | Detail |
| Monthly origination reporting | NOW LIVE at upstart.com/volume — real-time monthly data |
| Multi-year guidance | 2025–2028: 35% CAGR, terminal EBITDA margin ~25% |
| 2026 revenue guidance | $1.4 billion total, $1.3 billion in fees |
| 2026 EBITDA margin guidance | 21% (implying ~$294M in adjusted EBITDA) |
| Secured product revenue target | >$100M in associated 2026 fee revenue |
| Disclosure philosophy | “Offer the world a courtside seat” — Girouard, earnings call |
Read that table again. Upstart is now publishing monthly origination data for anyone to monitor in real time. They issued multi-year guidance through 2028. They gave specific product-level revenue targets. They told you exactly what EBITDA margin to expect. And our analyst’s takeaway is “limited visibility.”
How much visibility do you need? Would you like Dave Girouard to come to your house with a whiteboard? Should Paul Gu text you his daily origination numbers? Would it help if Sanjay Datta projected the P&L in skywriting above your office?
This company just went from issuing quarterly guidance to issuing annual + multi-year + monthly origination data precisely to give investors more visibility. The 2026 guidance of $1.4 billion came in 10.1% above analyst consensus estimates. That’s not limited visibility—that’s a company sandbagging less and showing more. But you have to actually look at it.
| VERDICT: Upstart just introduced the most comprehensive disclosure framework in the AI lending space—monthly data, multi-year targets, product-level guidance—and beat its own full-year guidance on revenue, EBITDA, and margin. Calling this “limited visibility” is like standing in a lighthouse and complaining about the dark. |
Claim #5: “Post-Earnings Miss”
THEY SAID: The adjustment followed a series of earnings reports where “quantifiable blame”
was placed on misses.
This is the crown jewel of incompetence in this entire analysis, and it deserves the full autopsy.
Let’s start with what Upstart actually delivered in Q4 2025:
| Metric | Result | vs. Consensus |
| Revenue | $296.1 million | BEAT ($289M est.) — +2.5% |
| Fee Revenue | $265 million | BEAT — +33% YoY |
| GAAP EPS (diluted) |
$0.17 |
BEAT ($0.15 GAAP est.) — +12.7% |
| Adjusted EPS | $0.46 | Essentially in line ($0.47 adj. est.) — missed
by ONE PENNY |
| Adjusted EBITDA | $63.7 million | BEAT ($63M est.) — +64% YoY |
| Net Income | $19 million | BEAT (vs. -$2.8M year ago) |
| Originations | 456,000 loans / $3.2B | BEAT — +86% / +52% YoY |
| Full-Year Revenue | $1.04 billion | BEAT prior guidance of $1.0B |
| Full-Year EBITDA | $230 million | BEAT — 22% margin vs 18% guided, vs 2%
prior year |
| 2026 Revenue Guide | $1.4 billion | BEAT consensus by 10.1% |
| Balance Sheet Loans | $985M (down from $1.2B) | BEAT expectations — -20% QoQ |
| Automation Rate | 91% fully automated | Continued improvement |
Now here’s where the “miss” narrative came from, and it’s a masterclass in how financial media fails retail investors:
Some headline-generating services compared GAAP diluted EPS of $0.17 against a consensus of $0.46 that was built on adjusted earnings. That is an apples-to-oranges comparison. It is factually inaccurate. It is the kind of mistake that a first-year accounting student would catch.
Here’s what happened: when a company turns GAAP-profitable, accounting rules require the inclusion of all convertible note shares in the diluted share count. This inflated Upstart’s diluted denominator to 112.2 million shares. On the basic share count of 97.6 million, EPS was $0.19. On an adjusted basis (the metric that actually measures recurring cash generation), EPS was $0.46 versus a $0.47 consensus—a miss of one cent. One. Single. Penny. While beating on literally every other line item in the report.
Meanwhile, Investing.com simultaneously reported that GAAP EPS of $0.17 beat the GAAP consensus of $0.15 by 12.7%. So which is it? Did they miss or beat? The answer depends entirely on whether you’re comparing apples to apples—and the analyst who cited this “miss” clearly wasn’t.
StockStory’s analysis, which actually did the comparison correctly, headlined:“Upstart Surprises With Q4 CY2025 Sales. ” Revenue beat. GAAP EPS beat. Adjusted EBITDA beat. Forward guidance crushed consensus by 10%. Free cash flow swung from -$135 million to +$105 million sequentially. Operating margin went from -2.2% to +6.4%. The company’s own full-year EBITDA target of 18% was exceeded at 22%.
But sure. “They missed.”
Let me put this in terms even our analyst might understand: if you walked into a performance
review where you beat every revenue target, generated 20x more EBITDA than last
year, automated 91% of your entire operation, reduced your risk exposure by 20%,
signed 13 new partners, guided 10% above Street expectations for next year, and
your boss docked you for being one penny short on one adjusted metric—you wouldn’t call that a “miss.” You’d call HR.
|
VERDICT: Upstart did not miss earnings. They beat revenue. Beat GAAP EPS. Beat EBITDA. Beat origination growth. Beat balance sheet reduction targets. Beat their own full-year profitability guide. The only “miss” was one penny on adjusted EPS caused by convertible note dilution math—while simultaneously guiding 2026 revenue 10% above consensus. Anyone citing this as an “earnings miss” either didn’t read the release, doesn’t understand the accounting, or is being intentionally misleading. Their confusion is your opportunity. |
Final Scorecard: Analyst 0, Reality 5
| Claim | Analyst Said | Reality | Grade |
| Credit Risk | Spooked credit markets, rising delinquencies | Exceptional credit, +608bps over
Treasuries, 12 consecutive vintages |
F |
| Macro Threat | Weak labor market
threatens borrowers |
Outperformed 2021 peak in 43%
harder environment |
F |
| Balance
Sheet/Funding |
Capital markets
unwilling to fund |
92% auto funded by 3rd parties, 24
partners, -20% BS |
F |
| Limited
Visibility |
No clarity on future
revenue |
Monthly reporting, multi-year guide,
$1.4B 2026 rev (+10% vs est.) |
F |
| Earnings
Miss |
Quantifiable misses | Beat revenue, GAAP EPS, EBITDA,
FCF, originations, guidance |
F |
Five claims. Five failures. Zero research. This is what happens when someone writes a thesis based on vibes instead of filings. Every single data point cited by the analyst was either outdated, misapplied, or flatly contradicted by information Upstart published in its earnings release and discussed openly on the conference call.
Look—being bearish on a stock is fine. Thoughtful short theses are valuable. But a bear case has to start with the actual numbers and work backward to a conclusion. What this analyst did was start with a conclusion (“UPST is broken”) and then grab whatever headlines supported it without checking whether the headlines matched the underlying data. That’s not analysis. That’s confirmation bias with a byline.
Upstart just reported: $1.04 billion in revenue (+64% YoY), $230 million in adjusted EBITDA (22% margin, 20x YoY improvement), $11 billion in originations (+86%), 91% full automation, GAAP profitability, shrinking balance sheet, expanding partner base, and guided to 35% CAGR through 2028. The stock is being handed to you at a discount because AI-generated financial headlines compared the wrong EPS numbers.
After the hours it took me to refute Andrew Boone’s laziness, I have established a new ongoing award to every analyst, journalist or personality who gets lazy and spews rubbish.
Andrew is the first winner of our newly established Dunce award.
He brought a butter knife to a gunfight—and the gun was loaded with data he could have found in 15 minutes if he’d bothered.