Upstart Holdings, Inc. · NASDAQ:UPST
Upstart — Investor Day (May 2025)
153 slides from Upstart's investor day presentation, May 14, 2025
Slide 1: Upstart AI Day 2025
Upstart Al DAY 2025
Slide 2: Welcome
WELCOME
Slide 3: Agenda
D AV E G I R O U A R D, C O - F O U N D E R & C E O The Opportunity PA U L G U , C O - F O U N D E R & C T O Our AI Advantage Break C H A N TA L R A P P O R T, C M O Our Category-of-One Business A G E N D A S A N J AY D AT TA , C F O Pro tability and Resilience Executive Q&A fi
Slide 4: Forward Looking Statements & Non-GAAP Financial Measures
FORWARD LOOKING STATEMENTS & NON-GAAP FINANCIAL MEASURES During the course of AI Day, and throughout this presentation, the executive team of Upstart Holdings, Inc. (the “Company”, “we”, or “our”) may make “forward-looking” statements that are based on our management’s beliefs and assumptions and on
Slide 5: Upstart AI Day 2025
& Upstart Al DAY 2025
Slide 6: Dave Girouard
DAVE GIROUARD Co-founder & Chief Executive Of cer fi
Slide 7: Welcome
WELCOME ~ Al DAY 2025
Slide 8: Today's Goals
TODAY’S GOALS Help you understand the size of the AI lending opportunity Help you understand why Upstart will win Convince you to buy our stock and hold it forever
Slide 9: Why We Do This
WHY WE DO THIS ~ Al DAY 2025
Slide 10: We Did Not Start as an AI Company
WE DID NOT START AS AN AI COMPANY
Slide 11: Upstart's Success Is a Win for Consumers, Lenders and the Economy
UPSTART’S SUCCESS IS A WIN FOR CONSUMERS, LENDERS AND THE ECONOMY
Slide 12: Mission-Driven Companies Stop at Nothing
MISSION-DRIVEN COMPANIES STOP AT NOTHING TO ACHIEVE THEIR GOALS
Slide 13: >$25 Trillion: The Market
>$25 TRILLION in credit originated globally each year THE MARKET AUTO PERSONAL is unimaginably large COMMERCIAL SMALL AND CARD BUSINESS INDUSTRIAL REAL HOME E S TAT E Based on internal estimates. Categories listed are sample lending opportunities.
Slide 14: At Least $1 Trillion
AT LEAST $1 TRILLION in revenue is available to the platform that originates and services that credit Based on internal estimates.
Slide 15: AI Will Bring Unprecedented Change and Disruption
AI WILL BRING UNPRECEDENTED CHANGE AND DISRUPTION TO MOST MARKETS Education Healthcare Transportation Media & Entertainment Financial Services Manufacturing Professional Services Defense & Warfare Retail & E-commerce Energy & Utilities
Slide 16: AI Will Bring Unprecedented Change and Disruption
AI WILL BRING UNPRECEDENTED CHANGE AND DISRUPTION TO MOST MARKETS Education Healthcare Transportation Media & Entertainment Financial Services Manufacturing Professional Services Defense & Warfare Retail & E-commerce Energy & Utilities
Slide 17: Credit Is a First-Tier Candidate
CREDIT IS A FIRST-TIER CANDIDATE
Slide 18: What can I help with?
What can I help with? List the global industries that are likely to be most impacted by AI Attach Search Reason Voice
Slide 19: Industries Most Impacted by AI: Healthcare
List the global industries that are likely to be most impacted by AI Here’s the list of global industries expected to be most impacted by AI, based on projected economic value, automation potential, and transformative use cases: 🏥 1. Healthcare • Impact Areas: Diagnostics, drug discovery…
Slide 20: Industries Most Impacted by AI: Financial Services
List the global industries that are likely to be most impacted by AI 🏦 2. Financial Services • Impact Areas: Fraud detection, underwriting, trading algorithms, robo-advisors, customer service. • Why: Data-rich, regulated, and already tech-heavy-Al boosts speed and precision.
Slide 21: Why Do Loans Cost So Much?
HOW DOES AI IMPACT THE PRICE OF A LOAN? Customer acquisition WHY DO LOANS People & data Credit losses COST SO MUCH? Pro t fi fi
Slide 22: How Does AI Impact the Price of a Loan?
HOW DOES AI IMPACT THE PRICE OF A LOAN? R E S U LT S T O D AT E Customer acquisition Conversion improvements reduce CAC 50% REDUCTION I N C A C D E S P I T E > 5 X G R OW T H 1 People & data Automation eliminates hard cost 66% REDUCTION (especially people) I N O N B O A R D I N G P E O P L E C O S T
Slide 23: Why Does Getting a Loan Suck So Much?
WHY DOES GETTING A LOAN SUCK SO MUCH?
Slide 24: Loan Application Friction Steps
Doc Emp Wai In-p End ume ting erso app less ntat verifi loye for n licat requ ion catio r a de clos ions ests n ca cisio ing lls n
Slide 25: How Does AI Impact the Borrowing Experience?
Doc Emp Wai In-p End ume ting erso app less ntat verifi loye for n licat requ ion catio r a de clos ions ests n ca cisio ing lls n HOW DOES AI IMPACT THE BORROWING EXPERIENCE?
Slide 26: ~90% of Upstart Loans Have None of These
~90% OF UPSTART LOANS HAVE NONE OF THESE Doc Emp Wai In-p End ume ting erso app less n verifi loye licat requ tation catio r a de for cisio clos n ions ests n ca n ing lls
Slide 27: Upstart Is Building the Foundation Model for Credit
UPSTART IS BUILDING THE FOUNDATION MODEL FOR CREDIT
Slide 28: Foundation Model for Credit: Separation
UPSTART IS BUILDING THE FOUNDATION MODEL FOR CREDIT S E PA R AT I O N
Slide 29: Foundation Model for Credit: Automation
UPSTART IS BUILDING THE FOUNDATION MODEL FOR CREDIT S E PA R AT I O N A U T O M AT I O N
Slide 30: Foundation Model for Credit: Calibration
UPSTART IS BUILDING THE FOUNDATION MODEL FOR CREDIT S E PA R AT I O N A U T O M AT I O N C A L I B R AT I O N
Slide 31: Foundation Model for Credit: Generalization
UPSTART IS BUILDING THE FOUNDATION MODEL FOR CREDIT S E PA R AT I O N A U T O M AT I O N C A L I B R AT I O N G E N E R A L I Z AT I O N
Slide 32: Foundation Model for Credit: Personalization
UPSTART IS BUILDING THE FOUNDATION MODEL FOR CREDIT S E PA R AT I O N A U T O M AT I O N C A L I B R AT I O N G E N E R A L I Z AT I O N P E R S O N A L I Z AT I O N
Slide 33: 2025 Priorities
10x our advantage in AI 2025 Prepare our funding supply for rapid growth PRIORITIES Return to GAAP pro tability in 2H Giant leaps toward best rates, best process for all fi
Slide 34: Incredible Progress Toward Best Rates
INCREDIBLE PROGRESS TOWARD “BEST RATES” U P S TA R T W I N R AT E S I N P E R S O N A L L O A N S V E R S U S D I G I TA L P E E R S Closest peer Upstart 60 51% 51% 47% 47% 45% 45 45% 30 15 720+ 660-720 <660 Source: Based on internal analysis of third-party data from 4/9-5/6.
Slide 35: Upstart AI Day 2025
& Upstart Al DAY 2025
Slide 36: Paul Gu: Our AI Advantage
PAUL GU Co-founder & Chief Technology Of cer OUR AI ADVANTAGE fi
Slide 37: Profit, Growth, Credit
PROFIT GROWTH CREDIT
Slide 38: Profit Credit Growth
PROFIT CREDIT G R OW T H
Slide 39: Traditional Credit Risk Modeling References
NAEEM SIDDIQI Christian Bluhm, Ludger Overbeck, and Christoph Wagner Introduction to Credit Risk Modeling Seon Edition . Cre dit-Ri sk Chapman & Hall/CRC FINANCIAL MATHEMATICS SERIES a Al DAY 2025
Slide 40: Preliminary Scorecard and Logistic Regression
Preliminary Scorecard In general, the nal scorecards produced in this stage consist of between 8 and 15 characteristics, although in larger banks with a lot of data, I have seen those with 18 to 30 characteristics. This is done to ensure a comprehensive risk pro le that includes all the reasonable v
Slide 41: Logistic Regression Logit Equation
In general, the nal scorecards produced in this stage consist of between 8 and 15 characteristics, although in larger banks with a lot of data, I have seen those with 18 to 30 characteristics. This is done to ensure a comprehensive risk pro le that includes all the reasonable variables, as well as a
Slide 42: Designing a Scorecard
continuous, such as modeling loss given default, linear regression is used. The rest of this section will deal with using multiple logistic regression to predict a binary outcome (good/ bad). Logistic regression, like most other predictive modeling methods, uses a set of predictor characteristics to
Slide 43: Scorecard Business Goals
There are, however, some business goals that need to be considered when developing scorecards. Age Delq at Bureau Time at Res Score Odds 18-24 20 0 55 0-6 12 300 30 25-29 32 1-2 40 7-18 25 301 31 30-37 46 3-5 12 19-36 38 302 33 38-45 56 6+ 5 37+ 50 303 34 45+ 80 • • • • R/O Burden Region Inq 6 mth 3
Slide 44: Training Data Points
0 0 T R A I N I N G D ATA P O I N T S 0 0 0 2013 0
Slide 45: Training Data Points: Education Variables
0 0 T R A I N I N G D ATA P O I N T S 0 0 Education variables 0 2013 2014 0
Slide 46: Training Data Points: Ensemble of Algorithms
1,835 0.06M 3,209 5,111 7,536 T R A I N I N G D ATA P O I N T S 10,718 14,595 Ensemble of algorithms 2013 19,276 2014 2015 24,936
Slide 47: Training Data Points: Instant Approval
70,990 0.12M 85,396 101,986 120,666 T R A I N I N G D ATA P O I N T S 0.06M 141,220 163,642 Instant approval 2013 187,828 2014 2015 2016 213,385
Slide 48: Training Data Points: XGBoost with Censoring
359,729 0.6M 393,690 429,261 466,560 T R A I N I N G D ATA P O I N T S 0.25M 505,552 546,207 XGBoost with censoring 2013 588,407 2014 2015 2016 2017 633,623
Slide 49: Training Data Points: Revenue Science
935,682 1.2M 1,014,655 1,100,430 1,192,974 T R A I N I N G D ATA P O I N T S 0.6M 1,293,367 1,398,110 Revenue science 2013 1,506,118 2014 2015 2016 2017 2018 1,617,930
Slide 50: Training Data Points: Loan-Month
2,278,397 3M 2,437,208 2,603,130 2,774,643 T R A I N I N G D ATA P O I N T S 1.5M 2,954,635 3,140,165 Loan-month 2013 3,331,248 2014 2015 2016 2017 2018 2019 3,529,927
Slide 51: ML+AI arXiv Papers per Month
= sm] a, fe) = = ov a n a v Q oO a — is) a ML+Al arXiv papers per month publication year The Unreasonable Effectiveness of Recurrent Neural XGBoost Win {Every Mactine earning Competition? bh werbest ails tn radar oat a aiid LY Al DAY 2025
Slide 52: But Loans Are a Survival Problem
BUT LOANS ARE A “SURVIVAL” PROBLEM
Slide 53: Cumulative Default Probability
14% 12% Cumulative default probability 10% P O S T- L O A N M O N T H B O R R OW E R A DEFAULT 12.65% Probability of defaul t 8% O PT I M A L A P R 1 6 . 70 % CDFS 6% B O R R OW E R B 12.65% Probability of defaul t O PT I M A L A P R 1 3 . 0 9 % 4% AV E R A G E T I M I N G C U R V E 2% 0% 0 10 20 30
Slide 54: Training Data Points: Loan-Month
2,278,397 3M 2,437,208 2,603,130 2,774,643 T R A I N I N G D ATA P O I N T S 1.5M 2,954,635 3,140,165 Loan-month 2013 3,331,248 2014 2015 2016 2017 2018 2019 3,529,927
Slide 55: Training Data Points: Neural Networks
4,700,563 6M 4,983,465 5,286,411 5,610,899 T R A I N I N G D ATA P O I N T S 3M 5,962,012 6,337,137 Neural networks 2013 6,727,755 2014 2015 2016 2017 2018 2019 2020 7,112,578
Slide 56: Training Data Points: LDAs and SHAP
9,095,274 12M 9,555,590 10,042,546 10,566,872 T R A I N I N G D ATA P O I N T S 6M 11,130,093 11,736,692 LDAs and SHAP 2013 12,384,768 2014 2015 2016 2017 2018 2019 2020 2021 13,100,301
Slide 57: Training Data Points: Proprietary Loss Function
18,003,683 30M 19,289,044 20,701,366 22,259,813 T R A I N I N G D ATA P O I N T S 15M 23,966,263 25,777,893 Proprietary loss function 2013 27,670,5792014 2015 2016 2017 2018 2019 2020 2021 2022 29,645,983
Slide 58: Generic Loss Function
GENERIC LOSS FUNCTION Loan that defaults in month 1 Loan that defaults in month 60 Loan that defaults Loan that prepays Loan predicted to default in Loan predicted to default in month 60 and defaults in month 59 month 1 and defaults in month 59
Slide 59: Proprietary vs Generic Loss Function Accuracy
0.1100 1.18 0.1075 1.14 0.1050 Accuracy on nancial objective Predicted default in month 60 1.10 0.1025 Proprietary loss function 1.06 0.1000 1.02 0.0975 Generic loss function 0. 98 0.0950 0.0925 0. 94 0.0900 0. 90 0.04 0.05 0.06 0.07 0.08 0.09 0.10 0.11 0.12 Predicted default in month 1 fi
Slide 60: Training Data Points: Proprietary Loss Function
18,003,683 30M 19,289,044 20,701,366 22,259,813 T R A I N I N G D ATA P O I N T S 15M 23,966,263 25,777,893 Proprietary loss function 2013 27,670,5792014 2015 2016 2017 2018 2019 2020 2021 2022 29,645,983
Slide 61: Training Data Points: Dynamic Macro
39,798,572 50M 41,823,890 43,831,287 45,815,630 T R A I N I N G D ATA P O I N T S 25 M 47,764,837 49,700,997 Dynamic macro 2013 51,602,936 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 53,472,105
Slide 62: Training Data Points: APR as a Feature
62,376,516 70M 64,092,180 65,795,540 67,481,671 T R A I N I N G D ATA P O I N T S 35M 69,141,906 70,793,166 APR as a feature 2013 72,406,909 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2 0 24 74,012,080
Slide 63: Upstart Matrix Compression
UPSTART MATRIX COMPRESSION UPSTARTXGB REDUCED MEMORY CONSUMPTION 30-60X Bank Loan Loan month Borrower ID Credit score Education Time Outcome balance amount index 10000 714 BA $4000 $10000 2023-01 1 0 10000 714 BA $4000 $10000 2023-02 2 0 10000 714 BA $4000 $10000 2023-03 3 0 Data repeats Data is uni
Slide 64: UpstartXGB Can Reduce Memory Consumption 30-60x
UPSTARTXGB CAN REDUCE MEMORY CONSUMPTION 30-60X Bank Loan Loan month Borrower ID Credit score Education Time balance amount index 10000 714 BA $4000 $10000 0 0 10000 0 0 0 0 2023-01 1 10000 0 0 0 0 2023-02 2 ML algorithms Novel 10000 0 0 0 0 2023-03 3 rewritten to “block sparse … … … … … … … support
Slide 65: Training Data Points: APR as a Feature
62,376,516 70M 64,092,180 65,795,540 67,481,671 T R A I N I N G D ATA P O I N T S 35M 69,141,906 70,793,166 APR as a feature 2013 72,406,909 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2 0 24 74,012,080
Slide 66: Training Data Points: Embeddings
87,686,453 100M 89,396,912 91,124,212 T R A I N I N G D ATA P O I N T S 50M Embeddings 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2 0 24 2025
Slide 67: Why This Matters
WHY THIS MATTERS S E PA R AT I O N A U T O M AT I O N C A L I B R AT I O N G E N E R A L I Z AT I O N P E R S O N A L I Z AT I O N
Slide 68: Why This Matters: Separation
WHY THIS MATTERS S E PA R AT I O N
Slide 69: Upstart vs Industry Pricing for Random Loan Sample
U P S TA R T V S I N D U S T R Y P R I C I N G F O R R A N D O M L O A N S A M P L E Industry range (Min. to max.) Upstart 40 35 30 Loss variance 25 20 15 10 5 625 649 664 669 674 679 685 691 696 702 708 715 720 725 732 739 750 763 777 786 798 825 Credit score Source: Based on internal analysis of t
Slide 70: 5 Years Ago: Traditional Model
TRADITIONAL MODEL THE SCENARIO 5 YEARS • FICO 666 • 5.8 years of credit history AGO • $61k annual income Decision: REJECT Probability of default is 32%
Slide 71: 5 Years Ago: Traditional vs Upstart Model
OUTCOME: 21 ON-TIME PAYMENTS BEFORE FULLY PAYING OFF LOAN TRADITIONAL MODEL U P S TA R T M O D E L THE SCENARIO 5 YEARS • FICO 666 • 5.8 years of credit history • JD from University of Pittsburgh • Lawyer for the City of Philadelphia AGO • $61k annual income • Low debt use relative to others in simi
Slide 72: Outcome: On-Time Payments Before Fully Paying Off Loan
OUTCOME: 21 ON-TIME PAYMENTS BEFORE FULLY PAYING OFF LOAN Predicted default Predicted prepayment 0.0600 P R E PA I D H E R E 2.5000 Predicted Prepayment % 2.0000 Predicted default % 0.0450 1.5000 0.0300 1.0000 0.0150 0.5000 0.0000 0.0000 6 02 0 6 2 0 6 02 -0 -0 -0 -0 -1 -1 - - 22 20 22 20 24 21 23 1
Slide 73: Traditional Model vs Upstart Model
TRADITIONAL MODEL U P S TA R T MODEL
Slide 74: Same Approval Rate, Lower Default Rate
50% A P P R OVA L R AT E TRADITIONAL MODEL 9% D E FA U LT R AT E 50% A P P R OVA L R AT E U P S TA R T MODEL 5% D E FA U LT R AT E
Slide 75: Same Default Rate, Higher Approval Rate
27% A P P R OVA L R AT E TRADITIONAL MODEL 7.5% D E FA U LT R AT E 67% A P P R OVA L R AT E U P S TA R T MODEL 7.5% D E FA U LT R AT E
Slide 76: Why This Matters: Automation
WHY THIS MATTERS S E PA R AT I O N A U T O M AT I O N
Slide 77: Fully Automated Loans
25% 50% 75% 100% Q 1' 1 8 Q 2' 18 Q 3' 18 Q 4' 18 Q 1' 1 9 Q 2' 19 Q 3' 19 Q 4' 19 Q 1'2 0 Q 2' 20 Q 3' 20 Q 4' 20 Q 1'2 1 Q 2' 21 Q 3' 21 Q 4' 21 Q 1'2 2 Q 2' 22 Q 3' 22 Q 4' 22 Q 1'2 3 Q 2' P E R C E N TA G E O F L O A N S F U L LY A U T O M AT E D 23 Q 3' 23 Q 4' 23 FULLY AUTOMATED LOANS Q 1'2 4
Slide 78: Fully Automated Loans: Conversion to Loans
FULLY AUTOMATED LOANS % A P P L I C AT I O N S CONVERSION TO LOANS A U T O M AT I C A L LY A P P R OV E D % Applications automatically approved % Applications manually reviewed 100% 100% 75% 75% 50% 50% 25% 25% 20 1 21 21 21 2 22 22 22 3 23 23 23 4 24 24 24 5 25 20 1 21 21 21 2 22 22 22 3 23 23 23 4
Slide 79: Why This Matters: Calibration
WHY THIS MATTERS S E PA R AT I O N A U T O M AT I O N C A L I B R AT I O N
Slide 80: Consumer Credit Post-COVID
20 21 0.4% 0.8% 1.2% 1.6% -0 4 20 21 -0 6 20 21 -0 8 20 21 -1 0 20 21 -1 2 20 22 -0 20 2 22 -0 4 20 22 -0 20 6 22 -0 8 20 22 -1 0 20 22 -1 2 20 23 -0 2 Personal loan 20 23 -0 4 20 23 -0 6 Card 20 23 -0 8 20 23 -1 0 20 Auto 23 -1 20 2 24 -0 20 2 6 0 - 9 0 D AY D E L I N Q U E N C I E S ( T R A N S U
Slide 81: How Our Model Did
HOW OUR MODEL DID 0.6 59% 59% 58% 58% 55% 55% 0.45 41% 41% L O S S VA R I A N C E 38% 38% 30% 30% 0.3 28% 22% 21% 21% 16% 0.15 12% 13% 13% 0% 0 -5% -5% -3% -7% -8% -8% -5% -7% -0.15 2022-Q1 2022-Q2 2022-Q3 2022-Q4 2023-Q1 2023-Q2 2023-Q3 2023-Q4 2024-Q1 -15% 2024-Q2 -16% 2024-Q3 -0.3
Slide 82: Today's Model
TODAY’S MODEL I N C O M PA R I S O N 0.6 59% 59% 58% 58% 55% 55% 0.45 41% 41% L O S S VA R I A N C E 38% 38% 28% 30% 30% 0.3 28% 22% 22% 21% 21% 16% 0.15 12% 12% 13% 13% 0 0% 0% -5% -5% -3% -5% -7% -8% -8% -5% -7% -7% -8% -8% -7% -0.15 -15% -15% -16% -16% -0.3 2022-Q1 2022-Q2 2022-Q3 2022-Q4 2023-Q1
Slide 83: Why This Matters: Generalization
WHY THIS MATTERS S E PA R AT I O N A U T O M AT I O N C A L I B R AT I O N G E N E R A L I Z AT I O N
Slide 84: Model Features by Product
MODEL FEATURES BY PRODUCT SDL AUTO REFI AUTO RETAIL HELOC Education Ensemble of Instant Revenue Loan Neural LDAs and Proprietary Dynamic APR as a Embeddings variables algorithms approval science month networks SHAP loss function macro feature
Slide 85: Why This Matters: Personalization
WHY THIS MATTERS S E PA R AT I O N A U T O M AT I O N C A L I B R AT I O N G E N E R A L I Z AT I O N P E R S O N A L I Z AT I O N
Slide 86: Reduction in Charge-Off, Increase in Revenue
TA R G E T E D TA R G E T E D TA R G E T E D L O A N TA R G E T E D D E B T L OW- C O S T H I G H -T O U C H M O D I F I C AT I O N S SETTLEMENT INTERVENTIONS INTERVENTIONS REDUCTION IN INCREASE 1% CHARGE-OFF 1% IN REVENUE AT CONSTANT COST* O PT I M I Z E D AGENT & AI AGENT PAY M E N T D U E B O R R
Slide 87: Profit Credit Growth
Short-term credit Home PROFIT CREDIT G R OW T H Auto Unsecured installment
Slide 88: Upstart AI Day 2025
° & Upstart ’ AILDAY 2025
Slide 89: Chantal Rapport: Our Category-of-One Business
CHANTAL RAPPORT Chief Marketing Of cer & SVP of Growth OUR CATEGORY-OF-ONE BUSINESS fi
Slide 90: Credit Matters
CREDIT MATTERS — wh
Slide 91: New Realities of Lending
NEW REALITIES OF LENDING Banks losing market share
Slide 92: New Realities of Lending
NEW REALITIES OF LENDING Banks losing market share Loans will be shopped
Slide 93: New Realities of Lending
NEW REALITIES OF LENDING Banks losing market share Loans will be shopped AI is a commitment
Slide 94: New Realities of Lending
NEW REALITIES OF LENDING Banks losing market share Loans will be shopped AI is a commitment Private credit emerges
Slide 95: The Business Model That Wins
THE BUSINESS MODEL THAT WINS Leading Intelligent Winning foundation model marketplace brand
Slide 96: Leading Foundation Model
U P S TA R T ’ S W I N N I N G B U S I N E S S M O D E L LEADING FOUNDATION MODEL
Slide 97: What Are the Proof Points of Leading AI?
WHAT ARE THE PROOF POINTS OF LEADING AI? Pricing power Positive selection Accelerated learnings Superior product
Slide 98: Personal Loan Take Rate
WHAT ARE THE PROOF POINTS OF LEADING AI? Fee% Pricing power 10% Personal loan take rate 8% Positive selection 5% Accelerated learnings 3% 2021 2024 Superior product Period
Slide 99: Proof Points of Leading AI: Loss Variance
WHAT ARE THE PROOF POINTS OF LEADING AI? Industry range (Min. to max.) Upstart Pricing power 40 35 Positive selection 30 Loss variance 25 20 Accelerated learnings 15 10 5 625 649 664 669 674 679 685 691 696 702 708 715 720 725 732 739 750 763 777 786 798 825 Superior product Credit score Source: Bas
Slide 100: Proof Points of Leading AI: Loss Variance
WHAT ARE THE PROOF POINTS OF LEADING AI? Upstart Pricing power 40 35 Positive selection 30 Loss variance 25 20 Accelerated learnings 15 10 5 625 649 664 669 674 679 685 691 696 702 708 715 720 725 732 739 750 763 777 786 798 825 Superior product Credit score Source: Based on internal analysis of thi
Slide 101: Proof Points of Leading AI: Loss Variance
WHAT ARE THE PROOF POINTS OF LEADING AI? Upstart Pricing power 40 35 Positive selection 30 Loss variance 25 20 Accelerated learnings 15 10 5 625 649 664 669 674 679 685 691 696 702 708 715 720 725 732 739 750 763 777 786 798 825 Superior product Credit score Source: Based on internal analysis of thi
Slide 102: Proof Points: Increase in Accuracy
WHAT ARE THE PROOF POINTS OF LEADING AI? Pricing power Positive selection 1 % Increase Accelerated learnings in accuracy Superior product
Slide 103: Proof Points: Accuracy and Conversion
WHAT ARE THE PROOF POINTS OF LEADING AI? Pricing power Positive selection 1 : 13 % Increase % Increase Accelerated learnings in accuracy in conversion Superior product
Slide 104: Proof Points: Faster Learnings
WHAT ARE THE PROOF POINTS OF LEADING AI? Increased Pricing power conversion Positive selection consumers product Better More . Accelerated learnings Superior product Faster learnings
Slide 105: Best Rates, Best Process for All
WHAT ARE THE PROOF POINTS OF LEADING AI? Pricing power BEST RATES, Positive selection BEST PROCESS Accelerated learnings FOR ALL Superior product
Slide 106: Marketplace Capital
U P S TA R T ’ S W I N N I N G B U S I N E S S M O D E L MARKETPLACE CAPITAL
Slide 107: Lender and Consumer
LENDER CONSUMER
Slide 108: Competition Among Diverse Capital Sources
INTELLIGENT MARKETPLACE COMPETITION AMONG DIVERSE CAPITAL SOURCES
Slide 109: Marketplace Auction Creates Structural Advantages
MARKETPLACE AUCTION CREATES STRUCTURAL ADVANTAGES SCALABLE EFFICIENT F L E XI B L E Serve more Optimize rates Deploy more borrowers across and returns adaptable capital more credit tiers dynamically sources
Slide 110: The Winning Brand
U P S TA R T ’ S W I N N I N G B U S I N E S S M O D E L THE WINNING BRAND
Slide 111: Something for Everyone
C U R R E N T P R O D U C T C OV E R A G E Super prime B R A N D A DVA N TA G E SOMETHING Prime FOR EVERYONE Non prime Personal Home Auto re Relief loans f
Slide 112: Lending for a Lifetime
OF OUR CUSTOMERS… B R A N D A DVA N TA G E 76% LENDING FOR of applicants have cars A LIFETIME 40% of applicants own homes 30% of borrowers are likely to take out another loan within 5 years Source: Based on internal analysis.
Slide 113: We Live for Speed
B R A N D A DVA N TA G E WE LIVE FOR SPEED
Slide 114: >90% of Personal Loans Approved Automatically
B R A N D A DVA N TA G E >90% OF PERSONAL LOANS APPROVED AUTOMATICALLY Source: Based on internal analysis.
Slide 115: The First Instant Auto Refinance
B R A N D A DVA N TA G E THE FIRST INSTANT AUTO REFINANCE Source: Based on internal analysis.
Slide 116: HELOC >2x Faster Than the Industry
B R A N D A DVA N TA G E HELOC >2X FASTER THAN THE INDUSTRY Source: Based on internal analysis.
Slide 117: The Reviews Speak for Themselves
B R A N D A DVA N TA G E THE REVIEWS The process was very easy, quick, & legit! SPEAK FOR The process was very easy & quick…[and] the interest rate!!! I haven't seen an interest rate offer this good in YEARS. I THEMSELVES will be able to pay off debts without my payment being eaten up by interest. T
Slide 118: Unlimited Potential
UNLIMITED POTENTIAL a Al DAY 2025
Slide 119: Automobile Refinancing
BRINGING IT TOGETHER AUTOMOBILE REFINANCING
Slide 120: A $50 Billion Market?
A SO BILLION MARKET? a Al DAY 2025
Slide 121: Meet Fran
a Al DAY 2025
Slide 122: Applies for a Personal Loan
11:53 AM Applies for a personal loan
Slide 123: Loan Processed
11:53 AM Applies for a personal loan 12:03 PM Loan processed
Slide 124: Applies for an Auto Refi Loan
11:53 AM Applies for a personal loan 12:03 PM Loan processed 12:05 PM Applies for an auto re loan fi
Slide 125: Refinances Her Car
11:53 AM Applies for a personal loan 12:03 PM Loan processed 12:05 PM Applies for an auto re loan 12:14 PM Re nances her car fi fi
Slide 126: Customer Feedback
FEEDBACK I am extremely grateful to you. You can't imagine how much help you have given me. I would recommend you to all my family and friends without a doubt the best. Thank you.
Slide 127: Accelerating Early Product Growth
19x ACCELERATING EARLY PRODUCT GROWTH 1x Q1 2024 Q1 2025 Auto re originations fi
Slide 128: Where We Are Today
W H E R E W E A R E T O D AY TRUSTED BY MILLIONS Over 3M borrowers served CATEGORY-LEADING AI 8x better risk separation MARKETPLACE AUCTION 100+ partners funding full credit spectrum EXPANDING PRODUCT BREADTH 4+ lending products GROWING LIFETIME VALUE 0% CAC second loans
Slide 129: Where We Are Today and Our Vision
W H E R E W E A R E T O D AY OUR VISION TRUSTED BY MILLIONS ALWAYS-ON CREDIT Over 3M borrowers served For 100% of Americans CATEGORY-LEADING AI EVERYTHING STORE FOR CREDIT 8x better risk separation To borrow, manage, and optimize all credit MARKETPLACE AUCTION GENERATIONAL BRAND 100+ partners fundin
Slide 130: The Future of Credit Is Here
o THE FUTURE OF CREDIT IS HERE a AI DAY 2025
Slide 131: Upstart AI Day 2025
& Upstart Al DAY 2025
Slide 132: Sanjay Datta: Profitability and Resilience
SANJAY DATTA Chief Financial Of cer PROFITABILITY AND RESILIENCE fi
Slide 133: The Market Opportunity Is Vast
THE MARKET OPPORTUNITY IS VAST WHY DO LOANS COST SO MUCH? WHY DOES GETTING A LOAN SUCK SO MUCH? Customer acquisition People & data In-p ers clos on Wai ing tin a de g for Emp cisio loy n verifi Doc catio er Credit losses ume n n ca lls requ tation End ests app less licat ions Pro t fi
Slide 134: Upstart Is Building the Foundation Model for Credit
UPSTART IS BUILDING THE FOUNDATION MODEL FOR CREDIT S E PA R AT I O N A U T O M AT I O N C A L I B R AT I O N G E N E R A L I Z AT I O N P E R S O N A L I Z AT I O N
Slide 135: Our Category-of-One Business
OUR CATEGORY-OF-ONE BUSINESS B E S T R AT E S , B E S T P R O C E S S F O R A L L W H E R E W E A R E T O D AY OUR VISION TRUSTED BY MILLIONS ALWAYS-ON CREDIT Over 3M borrowers served For 100% of Americans CATEGORY-LEADING AI EVERYTHING STORE FOR CREDIT 8x better risk separation To borrow, manage, a
Slide 136: Upstart AI + Business Model = Unique Financial Profile
UPSTART AI + BUSINESS MODEL = UNIQUE FINANCIAL PROFILE Growth Pricing power Margins Operating leverage Pro ts fi
Slide 137: Conversion Rate
UPSTART AI + BUSINESS MODEL = UNIQUE FINANCIAL PROFILE C O N V E R S I O N R AT E Growth 20% Pricing power 15% Margins 10% Operating leverage 5% Pro ts 0% 3 23 3 3 4 24 24 24 5 1'2 2 2 1'2 1'2 2' 3' 4' 2' 3' 4' Q Q Q Q Q Q Q Q Q fi
Slide 138: Personal Loan Take Rate
UPSTART AI + BUSINESS MODEL = UNIQUE FINANCIAL PROFILE P E R S O N A L L O A N TA K E R AT E Growth 10% Pricing power 7.5% Margins 5% Operating leverage 2.5% Pro ts 0% 2 22 2 2 3 23 3 3 4 24 24 24 1'2 2 2 1'2 2 2 1'2 2' 3' 4' 2' 3' 4' 2' 3' 4' Q Q Q Q Q Q Q Q Q Q Q Q fi
Slide 139: CAC and % Fully Automated
UPSTART AI + BUSINESS MODEL = UNIQUE FINANCIAL PROFILE C A C A N D % F U L LY A U T O M AT E D Growth 100% 3% Pricing power 90% 2.5% % Fully automated CAC Margins 80% 2% Operating leverage 70% 1.5% Pro ts 60% 1% 1 21 21 21 2 22 22 22 3 23 23 23 4 24 24 24 1'2 1'2 1'2 1'2 2' 3' 4' 2' 3' 4' 2' 3' 4' 2
Slide 140: Contribution Margin
fi Pro ts Growth Margins Pricing power Operating leverage % Fully automated 0% 20% 40% 60% 80% 60% 70% 80% 90% 100% Q Q 1'2 1' 2 1 1 Q Q 2' 2' 21 21 Q Q 3' 3' 21 21 Q Q 4' 4' 21 21 Q Q 1'2 1' 2 2 2 Q Q 2' 2' 22 22 Q Q 3' 3' 22 22 Q Q 4' 4' 22 22 Q Q 1'2 1' 2 3 3 Q Q 2' 2' 23 23 Q Q 3' 3' 23 23 Q Q C
Slide 141: Monthly Fee Revenue vs. Fixed Cost
UPSTART AI + BUSINESS MODEL = UNIQUE FINANCIAL PROFILE M O N T H LY F E E R E V E N U E V S . F I XE D C O S T Growth Revenue from fees Fixed cost $80 Pricing power $60 $ in millions Margins $40 Operating leverage Pro ts $20 1 2 3 4 5 6 7 8 20 9 0 1 20 2 1 2 3 -0 -0 -1 -0 -0 -0 -0 -0 -1 -0 -0 -0 -0
Slide 142: Net Income (Loss): Steady Progress Last 5 Quarters
UPSTART AI + BUSINESS MODEL = UNIQUE FINANCIAL PROFILE N E T I N C O M E ( L O S S ) : S T E A DY P R O G R E S S L A S T 5 Q U A R T E R S Growth $0M Pricing power . Margins , Operating leverage $-70M 4 24 24 24 5 Pro ts 1'2 1'2 2' 3' 4' Q Q Q Q Q fi
Slide 143: Our Technology + Business Model = Unique Financial Profile
OUR TECHNOLOGY + BUSINESS MODEL = UNIQUE FINANCIAL PROFILE Growth Pricing power Margins Low capital Operating leverage intensity Pro ts fi
Slide 144: Business Model Resilience
BUSINESS MODEL RESILIENCE Borrower diversity Product diversity AUTO LOANS HOME LOANS $240 $70 $50 Originations ($M) $180 $61 $61 $41 $41 Originations ($M) Originations ($M) $53 $43 $38 $43 $120 $27 $27 $35 $25 $26 $26 $17 $17 $60 $18 $18 $13 $13 $18 $13 $7 $7 $8 $8 $0 $0 $0 20 03 4 5 6 7 8 20 9 0 20
Slide 145: Borrower Diversity
BORROWER DIVERSITY S U P E R - P R I M E O R I G I N AT I O N S ( F I C O > 7 2 0 ) $240 T-Prime launched $180 Originations ($M) $120 $60 $0 03 04 05 06 07 08 09 0 1 2 1 2 3 -0 -1 -0 -0 -1 -1 24 - - - - - - 24 - 24 25 25 25 24 24 24 24 24 24 24 20 20 20 20 20 20 20 20 20 20 20 20 20 Monthly from 202
Slide 146: Product Diversity
PRODUCT DIVERSITY AUTO LOANS HOME LOANS $70 $50 $61 $41 $41 Originations ($M) Originations ($M) $53 $38 $43 $27 $35 $25 $26 $26 $17 $17 $18 $18 $18 $13 $13 $13 $8 $8 $7 $7 $0 $0 Q1 '24 Q2 '24 Q3 '24 Q4 '24 Q1 '25 Q1 '24 Q2 '24 Q3 '24 Q4 '24 Q1 '25
Slide 147: Funding Resilience
FUNDING RESILIENCE 0% 35% 65% 2022 2023 2024
Slide 148: Macro-Tooling
MACRO-TOOLING C U M U L AT I V E L O S S VA R I A N C E B Y O R I G I N AT I O N H A L F Actual With current tools 60% 59% 59% >50% 48% 48% 45% 34% 34% 30% 25% 25% of the credit Loss variance 15% 17% 17% underperformance 4% 4% 5% 5% 5% 5% experienced would have been avoided 0% -2% -2% -4% -4% -8% -8
Slide 149: Future Revenue Models
RATABLE FEE SUBSCRIPTIONS REVENUE FUTURE REVENUE MODELS SERVICING REVOLVING FOR ALL
Slide 150: Thank You
a THANK YOU ~ Al DAY 2025
Slide 151: Reconciliation of Non-GAAP Financial Measures
RECONCILIATION OF NON-GAAP FINANCIAL MEASURES ($ IN MILLIONS) Fiscal 2021 Fiscal 2022 Fiscal 2023 Fiscal 2024 Q1 2021 Q2 2021 Q3 2021 Q4 2021 Q1 2022 Q2 2022 Q3 2022 Q4 2022 Q1 2023 Q2 2023 Q3 2023 Q4 2023 Q1 2024 Q2 2024 Q3 2024 Q4 2024 Revenue from fees, net $116.2 $187.3 $210.4 $287.4 $314.0 $258
Slide 152: Upstart AI Day 2025
® Upstart Al DAY 2025
Slide 153: Upstart AI Day 2025
& Upstart Al DAY 2025