Upstart (UPST) The Citizens JMP Technology Conference 2025 summary

Business model and technology

• Applies machine learning to credit underwriting, fraud detection, and borrower acquisition since 2014, focusing on automation and reduced process friction.

• Operates a low-capital-intensity platform, charging fees to lenders and institutional investors, enabling rapid growth and profitability.

• Expanding AI-driven lending models to new segments, building on momentum after challenging years in the lending space.

AI and competitive differentiation

• Uses advanced, non-linear proprietary models tailored for credit's time series challenges, not just off-the-shelf solutions.

• Collects and leverages unique borrower data, such as education and employment, to enhance risk assessment.

• AI models are deployed in production, continuously learning and making real-time decisions.

• Generative AI has not changed internal processes; differentiation comes from adapting models to credit and building proprietary datasets over 11 years.

Data, learning, and macro adaptation

• Proprietary data collection over 11 years provides a competitive edge, as new entrants would need years to replicate this dataset.

• Models are designed to adapt to time series data, handling the uncertainty of loan outcomes over time.

• Macro events like the pandemic and inflation have tested models, emphasizing the need for rapid detection and response to changing conditions.

• Focus is on ranking borrower risk and reacting quickly to macro shifts, rather than predicting specific future events.