Who is Paul Gu and why is he becoming Upstart's CEO?

Paul Gu co-founded Upstart in 2012 and has spent over a decade leading the company's AI, machine learning, and data science efforts. He was announced as Upstart's next CEO beginning May 2025, succeeding co-founder Dave Girouard. Gu and Girouard have partnered on virtually every aspect of building Upstart over 14 years, and both have always believed in the power of founder-led companies for long-duration missions like transforming consumer credit.

What is Upstart's Model 25 and why does it matter?

Model 25 is Upstart's 25th major model update. It incorporates data from loans Upstart did not make — either because borrowers were declined or chose a competitor — by joining third-party data about those consumers' subsequent loan performance. This overcomes a longstanding bias in credit modeling where models can only learn from historically approved borrowers, allowing Upstart to improve accuracy on exactly the borrower segments it previously had blind spots about.

What is the Upstart Macro Index (UMI)?

The Upstart Macro Index is a proprietary system that isolates the background level of consumer credit risk after controlling for borrower-level characteristics, loan types, and mix changes. It allows Upstart to respond faster and more precisely to macroeconomic shifts than traditional lenders who rely on portfolio-level averages. Upstart estimates the UMI framework would have avoided approximately 80% of the credit disruption experienced during the 2022 macro shock.

Why did Paul Gu buy 100,000 shares of Upstart stock?

Paul Gu purchased approximately 100,000 shares of Upstart on the open market for roughly $4 million because he believes the stock price will go higher. He describes himself as a fundamentals-based value investor who has his own mental model of what the stock is worth, and at some point the gap between that model and the market price became large enough to justify increasing his already concentrated position.

What is Upstart's long-term strategy for consumer credit?

Upstart's strategy is to become the best place for consumers to go for every major category of credit. CEO Paul Gu describes this as a Costco model — broad coverage with competitive pricing across personal loans, auto loans, HELOCs, and eventually all major consumer credit categories — rather than a Whole Foods model of high-margin niche dominance. The company has guided for approximately 35% annual revenue compounding over the next three years.

How does Upstart's AI break the growth-profits-credit performance tradeoff?

Paul Gu describes a timeless law of credit: without deep technology differentiation, lending platforms are forced to trade off between growth, profits, and credit performance. AI is what allows Upstart to operate on a higher frontier — growing fast, running a profitable business, and still delivering strong credit outcomes. The proof is in 12 consecutive quarterly loan cohorts delivering over 600 basis points of excess return over treasuries.

What are Upstart's new products beyond personal loans?

Upstart has expanded beyond unsecured personal loans into auto loans, HELOCs (home equity lines of credit), and a recently announced cash line product. These new products currently operate at lower margins than the core business — some at zero or negative margins — but are expected to reach meaningfully higher terminal margins as they scale. The TAMs for auto and home lending are substantially larger than personal loans.

How fast are Upstart's AI models improving?

Upstart typically ships one to two major model updates per quarter, with five to seven major research initiatives running in parallel at any given time. The company recently crossed 100 million repayment events, a milestone that took 14 years to reach. The next 100 million are expected in less than a few years as the loan base grows and new products like auto and HELOC contribute additional data.

Who is Henry Dierkes and what is Rebellionaire?

Henry Dierkes is an analyst at Rebellionaire, a fiduciary financial advisory firm (a brand of Halter Ferguson Financial, based in Carmel, Indiana) that specializes in concentrated Tesla investors. Rebellionaire helps clients optimize concentrated stock positions through strategies like covered calls, LEAPs, and tactical position management rather than forcing traditional diversification. The firm serves clients nationwide.

EXCLUSIVE Interview With Upstarts Incoming CEO

Henry Dierkes
Feb 25
6 min read

Rebellionaire sits down with Paul Gu — Upstart co-founder, AI architect, and incoming CEO — to talk model upgrades, product expansion, insider buying, and why the best is still ahead.

The Man Behind the Models

Paul Gu doesn't do things for optics. That's one of the first things you pick up when you talk to him.

He co-founded Upstart in 2012 alongside Dave Girouard, and for the past 14 years he's been the person most responsible for the AI and machine learning that makes the whole thing work. On the February 10th earnings call, Upstart announced that Paul would step into the CEO role beginning in May — a transition he says has been in the works for a long time.

I sat down with Paul to dig into what the transition means, where the technology is headed, and why he just put nearly $4 million of his own money into the stock.

Why Founder-Led Matters Here

Paul and Dave have always been aligned on one thing: transforming credit is a multi-decade mission. It's not a vertical SaaS play that can be built and scaled in a few years. Credit is one of the oldest, most foundational industries in civilization — and bending it with AI is going to take conviction and a time horizon that most professional managers simply don't have.

That's the core argument for founder-led companies in situations like this. You need someone at the helm who isn't optimizing for the next two quarters. You need someone who's willing to be patient while the rest of the world remains skeptical.

Paul on the mission: "Since the beginning, the world has been skeptical that AI would be relevant to credit. I think we've slowly started to prove the world wrong on the first pieces of that — but there are a lot more years where that's going to continue to grow."

A Very Different Business Than 2022

If your mental model of Upstart is stuck in 2022, it's outdated. Back then, the business was essentially one product — unsecured personal loans to near-prime consumers. When the post-COVID stimulus hangover hit and personal savings rates plummeted, Upstart had to tighten approvals dramatically. Since they were a volume-based business with a single product, the impact was severe.

Today, two major things have changed.

First, Upstart built the Upstart Macro Index (UMI) — a framework designed to be the fastest and most precise system for reading changes in background consumer risk. Paul estimates the UMI would have avoided roughly 80% of the credit disruption they experienced in 2022. That's not an incremental improvement. That's a structural change to how the business absorbs macro shocks.

Second, the business is meaningfully diversified. They now have a growing HELOC product, an auto lending product, and a product lineup that's significantly less sensitive to the kind of narrow shock that hit them three years ago. It's no longer a single-product company.

14 yrs

AS CO-FOUNDER

100M+

REPAYMENT EVENTS

~80%

UMI DISRUPTION AVOIDED

CONSECUTIVE STRONG VINTAGES

Model 25: Learning From Loans They Didn't Make

This is one of the most interesting things Paul shared. Upstart's 25th major model update broke through a longstanding bias in credit modeling: the fact that models can only learn from loans they approved. If your model has blind spots about certain borrower types, it never gets the data to fix them because it never lends to those people in the first place.

Model 25 solves this by pulling in third-party data about consumers who were either declined by Upstart or chose a competitor. If those consumers went on to take loans elsewhere, Upstart can now join those data sets and learn from exactly the borrower segments it historically struggled with.

That's not just a model improvement — it's a structural unlock in how fast the models can get better. And it comes at a time when the company just crossed 100 million repayment events. Paul said the next 100 million will come in a fraction of the time — potentially less than a few years — especially as new products like auto and HELOC start contributing data.

The $4 Million Statement

In November, Paul purchased roughly 100,000 shares of Upstart on the open market for just under $4 million. When I asked him about it, his answer was refreshingly simple: he thinks the stock will be higher.

He's not doing it for the signal. He describes himself as a fundamentals-based value investor who has his own internal model of what the business is worth. At some point, the gap between that model and the market price got wide enough that adding to an already concentrated position made sense.

His math is straightforward. Upstart recently gave three-year guidance calling for roughly 35% annual compounding. If the company hits those numbers, Paul argues that almost any reasonable valuation framework — price-to-earnings, price-to-sales, price-to-book — applied to the 2028 business produces compelling returns from today's price. The specific multiple matters less than the time horizon you choose to apply it over.

"Businesses compounding at 35% don't suddenly decelerate to 5 or 10%. So if you're going to value us on price-to-book, you have to make sure you're applying it to the right year — the year where growth rates actually look like the businesses you're comparing us to."

Costco, Not Whole Foods

One of the most clarifying moments in the conversation was Paul's analogy for where he wants to take the business. He wants Upstart to be Costco — not Whole Foods.

The Whole Foods path would mean staying in personal loans, maximizing take rates, and running a high-margin niche business. That's a real opportunity, but it's a smaller one. The Costco path means going after every major category of consumer credit — home, auto, and beyond — even if some of those products carry lower margins than the core personal loan business. The TAMs are enormous, the loan sizes are larger, and the brand effect compounds: one place to go for the best credit, period.

This is already in motion. The new products currently operate at thin or even negative margins — that's just the reality of scaling immature businesses that don't yet have the right ratios of automation and support. But the terminal margins will be meaningfully higher, and the real payoff is a consumer credit ecosystem where borrowers don't have to think about whether Upstart is right for their specific need.

Capital Partners See the Trifecta

Paul framed the credit investing world through what he called the "timeless law" of the business: any lending platform without deep technology differentiation is forced to trade off between growth, profits, and credit performance. You can optimize for one or two, but you can't have all three.

AI is what breaks that trade-off for Upstart. It's what creates a higher frontier — where they can grow fast, run a profitable business, and still deliver strong credit outcomes. And the proof is in the capital relationships: 12 consecutive quarterly cohorts delivering over 600 basis points of excess return over treasuries.

If investors didn't believe in the AI differentiation, Paul noted, the only rational conclusion from Upstart's growth and profitability profile would be that the credit performance must be suffering. The fact that capital relationships continue to expand — across banks, credit unions, and institutional investors — is the market's way of saying the technology is real.

What's Next

The big banks are coming, but slowly. Paul was candid about the two historical obstacles: large institutions don't care much about personal loans (not a meaningful part of their portfolio), and adopting AI-based underwriting is a major paradigm shift that takes time for risk-averse organizations.

Both of those barriers are eroding. As Upstart's product mix shifts toward home and auto — categories that actually matter to large banks — the conversations get more interesting. And the broader AI adoption wave is making the technology less foreign to institutional decision-makers. Paul expects the first movers among larger institutions to benefit significantly.

The marketing strategy is evolving too. Upstart is moving from purely targeted direct-response campaigns to broader brand marketing. The logic is simple: if you've invested in building the objectively best product across multiple credit categories, it only makes sense in a capitalist framework if consumers actually know about it. A real mobile app is coming. A cash line product was recently announced. The business is shifting from a one-time transaction model to an ongoing relationship model.

The Bottom Line

Paul Gu is the architect of Upstart's AI and he's about to run the whole company. He's buying stock aggressively with his own money. The technology moat is deepening with every model iteration. And the business is transforming from a single-product lender into a multi-product consumer credit platform operating in what Paul calls an "essentially unlimited TAM."

The world has been skeptical of AI in credit since day one. Paul's been proving them wrong for 14 years, and he's just getting started.

Watch the full conversation above.

Concentrated in a High-Conviction Stock?

Rebellionaire helps investors optimize concentrated positions through strategic tools like covered calls, LEAPs, and tactical position management — not by telling you to sell. Sound like a good fit? Reach out and learn more.

Disclosure: Rebellionaire is a brand of Halter Ferguson Financial, a fiduciary advisory firm. This content is for informational purposes only and does not constitute personalized financial advice. Positions referenced in this interview are those of the interviewee, not recommendations. Everyone's situation is different — consult a qualified advisor before making investment decisions. Henry Dierkes is an analyst at Rebellionaire and may hold positions in securities discussed.