Sunday, October 11, 2026

Brazilian fintechs deploy AI credit models to scale 100+ products while maintaining risk margins

Nu Holdings deployed its nuFormer AI credit model in 2025, launching over 100 products while keeping risk-adjusted margins stable. Inter & Co added 7 million clients at industry-leading funding costs. AI-driven underwriting allows Brazilian fintechs to scale credit faster than banks in developed markets still relying on manual processes.

Brazilian fintechs deploy AI credit models to scale 100+ products while maintaining risk margins
Image generated by AI for illustrative purposes. Not actual footage or photography from the reported events.

Nu Holdings deployed its nuFormer AI credit model to production in 2025, launching over 100 products and features across Brazil, Mexico, and Colombia while maintaining stable risk-adjusted net interest margins quarter-over-quarter. The deployment shows how AI-driven underwriting enables rapid expansion without proportional credit losses, a capability that distinguishes emerging market fintechs from slower-moving incumbents in developed economies.

Inter & Co welcomed 7 million new clients in 2025, its strongest annual performance, while maintaining funding costs at 65.6% of CDI, Brazil's benchmark interbank rate. Newer client cohorts transact faster and more frequently than older ones, suggesting AI models improve client quality at acquisition—a pattern mirroring fintech growth in India and Southeast Asia where mobile-first banking bypassed legacy infrastructure.

Traditional credit underwriting in markets like the U.S. and Europe still requires manual document review, income verification, and days-long processing. AI models analyze hundreds of data points in seconds, including transaction patterns and behavioral signals invisible to human reviewers. This speed advantage matters most in emerging markets where credit access remains limited and competition for quality borrowers intensifies.

Real-time portfolio monitoring provides a second edge. AI systems flag deteriorating credit conditions before they materialize as losses, allowing lenders to adjust terms proactively. Traditional systems rely on monthly reviews, creating lag between risk emergence and response—a vulnerability exposed during recent rate cycles in both developed and emerging markets.

The technology optimizes client acquisition costs. Automated decisioning reduces underwriting expenses per application while improving approval accuracy. Higher approval rates for creditworthy applicants increase conversion; better rejection of risky applicants reduces future losses. Both effects improve unit economics, critical in markets where fintechs compete against established banks with lower cost of capital.

Nu's stable margins despite rapid product expansion suggest AI models maintain credit discipline under growth pressure. Manual underwriting teams often lower standards when acquisition targets rise. Algorithmic decisioning applies consistent criteria regardless of volume, an advantage as Brazilian fintechs expand into Mexico, Colombia, and other Latin American markets.

The competitive gap widens between AI-equipped fintechs and traditional banks. Incumbents in developed markets carry legacy systems designed for branch-based lending and batch processing. Retrofitting AI requires replacing core infrastructure, a multi-year project. Neobanks built cloud-native architectures from inception, allowing rapid model deployment—a structural advantage particularly pronounced in markets like Brazil where regulatory frameworks accommodate digital-first banking.

Brazilian fintechs now process millions of credit decisions monthly through automated systems, a scale impossible with human underwriters. The technology transforms credit from a manually rationed resource into a dynamic product that scales with demand while maintaining risk controls.


Sources:
1 Globe Newswire, "Olympians Inspire Expands School Assembly and Leadership Workshop Programming Featuring Elite Athlet" (March 23, 2026)
2 Yahoo Finance, "Asian shares decline as hopes dim for resolution in Iran after Trump's latest comments" (March 23, 2026)
3 Yahoo Finance, "NU's Efficiency Edge: A Key Factor Driving Its Premium Narrative" (March 20, 2026)
4 Yahoo Finance, "BC-Most Active Stocks" (March 20, 2026)
5 Nasdaq, "AI-Driven Fear Slashed Toast Stock by 43%, Even as Free Cash Flow Hit Records" (March 23, 2026)

In this story · Knowledge Files

What we know · the intelligence behind this page
Live from the substrate
What we're seeing
Agentic AI Rewires Enterprise Software: Platform Incumbents, Governance, and a Funded Startup Wave
Enterprise software is being rebuilt around autonomous AI agents. Incumbents and large platforms (SAP with its Autonomous Suite and Joule, Zeta with AthenaOS/AIM/Athena MCP, Meta with its new Enterprise Platform) are racing to own the agent layer. Meanwhile, seed and Series A money flows to finance-office and vertical startups (Dextr, Latitude, Dentira, Light), and consolidation continues through acquisitions (Tiny–Oso Cloud, Harvey–Guardrails AI). Investor commentary stresses that AI is better at disrupting around the edges of systems of record than at replacing them, that it should not be trusted with finance calculations, and that governance must be enforced by the system rather than left to agents.
Our read on the data ›
Signals we're tracking
EPKINLY Regulatory-Clinical Success Cascade
High probability of expanded label indications, additional combination approvals, and competitive positioning strength in follicular lymphoma market. Predicts positive commercial uptake and potential accelerated review for related indications.
Patterns we're watching ›
Where sources disagree
ING Group
Both facts record the same metric (shares_outstanding) for ING Group at the identical observation date (2025-12-31). FACT A states 2,902,437,688 shares; FACT B states 2,902 million shares (2,902,000,000). The difference is 437,688 shares (~0.015%). This is a genuine value conflict, though the discrepancy appears to result from FACT B rounding to the nearest million while FACT A provides the precise count.
We flag conflicts openly ›
Recently verified
✓ Checked against the original source
4,986
facts traced to their source — and we flag the ones that don't hold up.
101 entities tracked4,986 facts checked against source5,369 source documents archived
Query this data → isubstrate.com