Upstart (UPST) Bank of America Global Technology Conference 2025 summary

AI and model differentiation

• Sustained investment in a large team of machine learning researchers and alternative data collection has enabled continuous model innovation and improvement since 2012.

• Model improvements focus on algorithm design, new data sources, compute and memory enhancements, and accumulating repayment data, leading to consistent accuracy gains.

• Macro calibration enhancements now allow the model to adjust to economic shocks within two quarters, down from eight, improving resilience and credit performance.

• Time-aware modeling enables dynamic interaction between macro conditions and borrower characteristics, increasing adaptability to future economic changes.

• Improved credit performance builds confidence among capital markets and loan buyers, supporting business growth and capital resilience.

Product expansion and strategy

• Expansion into auto, HELOC, and small dollar products is driven by borrower needs and opportunities to improve approval rates and process efficiency.

• New products are designed to serve consumers across their credit life cycle, leveraging data advantages for better underwriting.

• Strategic focus during market stress led to prioritizing auto, HELOC, and small dollar products, which are now gaining momentum and ready to scale.

• Leadership views the small dollar product as the most strategic and auto as the most audacious, with HELOC complementing the existing portfolio.

Financial outlook and risk management

• Profitability in the second half of the year depends on stable macro conditions and continued model improvements.

• Forward flow agreements with funding partners involve extensive due diligence and risk-sharing, with Upstart retaining a low to mid single-digit percentage of origination risk.

• Overperformance in benign periods is intended to offset underperformance during macro shocks, creating a macro insurance layer.