Summary

• Fintech’s first wave overpromised disintermediation; the next wave is about improving bank unit economics — and Upstart is proving that better underwriting, not just better UX, drives real value.
• Traditional banks still lean heavily on FICO-based models that overlook rich, modern data sources like payroll streams and real-time cashflows — Upstart steps in with ML-native models that unlock approvals banks would otherwise miss.
• By owning the borrower interface and decisioning layer, Upstart captures first-party data that enables more precise risk stratification — helping banks expand loan volumes without taking on disproportionate risk.
• Banks benefit directly: more approved borrowers, lower loss rates, and capital relief via ABS sales — all while keeping customer relationships. Upstart’s take rate includes a rebate when banks eventually sell into ABS markets, aligning incentives.
• Innovations like APR-as-a-feature, PTM, embeddings, and Aqueduct let Upstart sharpen its models and approve more “near-miss” applicants — growing volumes and profitability for both Upstart and its partner banks.
• Part 2 will explore how this framework is translating into real loan growth across personal, auto, and HELOC categories — and how Upstart’s TAM and bank partnerships may evolve from here.

Fintech – Cycles of Hype and Renewal

For over a decade, fintech has promised to democratize finance and bring services to the unbanked and underbanked. Yet the industry’s journey has been anything but linear. Regulatory hurdles, legacy banks actively resisting change, and volatile interest rate regimes have repeatedly slowed its progress. Many once-promising startups faltered: some were acquired cheaply by incumbents, like Simple (acquired by BBVA), Level Money (acquired by Capital One), or Credit Karma’s tax unit (sold off for just $50m), while others like Clinkle and Powa Technologies collapsed outright.

Fintech has also been tarnished by a number of frauds. Payments startup Bolt faced allegations of inflated metrics and internal mismanagement, which eroded investor trust. Supply-chain finance firm Greensil collapsed after packaging risky loans as safe assets, leaving banks and investors nursing billions in losses. And Wirecard, once a flagship of European fintech, imploded in a $2 billion accounting scandal that exposed deep governance failures. These have been just more reasons for regulators to over regulate, for banks to block integrations, and for investors to stay on the sidelines.

Even fintechs that survived the initial boom often saw their ambitions clipped — Plaid’s valuation fell from $13bn to around $6bn (from the peak in April 2021 to the first clear public valuation reset in April 2025), and Chime dropped from $25bn to $11.6bn (from August 2021 to its IPO in June 2025). The fintech sector has consistently shown how quickly hype can give way to consolidation, decline, or survival mode.

The post-COVID period showcased these extremes vividly. Liquidity injections, stimulus checks, and near-zero rates created a boom for consumer-facing fintechs, fueling everything from stock-trading apps to crypto wallets. But inflation’s surge forced the Fed into one of the sharpest rate-hiking cycles in modern history, and suddenly underwriting models buckled. From late 2021 to mid-2023, public fintech names like Upstart (UPST), Opendoor (OPEN), SoFi (SOFI), Carvana (CVNA), and Affirm (AFRM) saw share prices collapse 60–90%. And how could we forget Klarna's valuation plummet from its 2021 $45.6bn peak to a low of $6.7bn just over a year later. Klarna, after 20 years, finally did IPO a few days ago at a $15bn valuation.

The boom had quickly become a bust. Yet COVID also left behind something enduring: it triggered a widespread aspiration for financial freedom. Retail investors poured into stocks and crypto, curiosity spread toward stablecoins, and adoption accelerated especially in emerging markets like South America and Eastern Europe. This cultural shift pushed fintechs and regulators alike to focus harder on reaching underserved populations.

Another COVID-driven change was the step-change in online user experience. With physical channels shut down, digital became the only gateway, and consumers demanded seamless, frictionless products. Fintechs responded by stripping away unnecessary steps and building intuitive web and mobile interfaces that made customers more willing to share data. This shift fed directly into a more powerful flywheel effect: easier onboarding brought more users, more users generated richer data streams, and richer data allowed faster improvements to credit models. Unlike the 2000s or even the pre-COVID 2010s, when borrower data trickled slowly through separate front-end applications and back-end models, today user interactions update risk models almost in real time. This velocity is enabled by open banking APIs, cloud computing, and highly efficient chips and networks that minimize latency. The result is a feedback loop where lending products improve continuously with scale.

At the same time, embedded lending has opened new data frontiers. BNPL providers like Affirm and Klarna don’t just underwrite off bureau data — they tap directly into ecommerce platforms, capturing high-fidelity consumption and repayment signals that incumbents never had. Open banking standards have also created tailwinds: APIs now allow lenders to analyze borrower cash flow directly from bank accounts, a far more dynamic risk signal than a static credit score. Those that capitalize on these standards stand to build lasting moats.

The macro backdrop is also shifting in fintech’s favor. After two years of punishing rate hikes, monetary policy has leveled off, with cuts expected on the horizon. Traditional banks face a headwind as lower rates compress their net interest margins, but fintechs gain a tailwind: lower borrowing costs increase loan demand, and because their revenues are largely tied to origination fees rather than spreads, margins hold steady. Meanwhile, ABS investors — hungry for equity-like returns with more protection — are increasingly open to fintech-sourced consumer credit, especially when powered by AI models that promise finer borrower separation.

Finally, there’s been an attitudinal change in banks themselves. Where incumbents once fought fintechs, they now increasingly partner with them. A player like Pagaya (PGY) illustrates this shift: banks originate loans, but Pagaya’s AI underwriting approves many borrowers the banks would otherwise reject. The bank keeps the customer relationship, but the loan is immediately sold to ABS investors, turning what was once a “no” into a four-way win for the borrower, the bank, the investor, and Pagaya itself.

From FICO to Fintech: A Structural Shift in Credit Risk

For many consumers, especially younger borrowers, the FICO score has become increasingly disconnected from their real credit performance, paving the way for fintech to grow in importance for the consumer credit markets. Take a 20-year-old who wants to buy a car: they may have steady income and a spotless record of repaying BNPL purchases, yet none of this data feeds into their FICO profile. With no traditional history, their score defaults to very low, and the bank — lacking integration with BNPL or cash-flow data — rejects them outright. This is generally because FICO, and the banks' application of it, is backwardlooking, checking the previous loan repayment that might be a few years old rather than ingesting more present credit history available in BNPL and other fintech areas. The problem is compounded by the fact that many banks still rely on outdated FICO models, often FICO 8 from 2009, because migrating to newer versions is costly and disruptive to their custom scorecards.