NU Holdings AI Push: Can NuFormer Strengthen Its Competitive Edge?

Nu Holdings' Q2 2026 credit portfolio rose 37% year over year to $39.4 billion, while risk-adjusted net interest margin expanded to 12.4% from 9.5% in Q1. Gross revenue increased 39% to $5.9 billion and net income reached $1.1 billion as NuFormer AI was deployed across underwriting, growth campaigns and customer support, where AI agents now handle over 60% of Brazilian conversations. NU shares have gained 29.4% over three months, though the stock trades at a 14.38x forward P/E versus the industry's 11.55x, and consensus 2026 EPS estimates rose $0.03 to $0.86.

Analysis

NU's AI narrative matters only if it converts into a durable reduction in loss formation rather than merely lower servicing expense. The key operating leverage is asymmetric: better approval selection can raise loan growth while reducing provisions, whereas support automation primarily protects the efficiency ratio. Investors should focus on 30+ and 90+ day delinquency vintages, cost of risk, and incremental credit-margin performance in Mexico as the next independently verifiable evidence over the next 1-3 quarters.

Competitive intensity is likely to absorb part of the benefit. ITUB has lower funding costs, a deeper corporate relationship base and regulatory familiarity; AI-enabled onboarding and servicing could let it narrow NU's experience advantage without matching NU's growth-at-all-costs credit exposure. MELI remains the more consequential medium-term threat in underbanked consumers and merchants because commerce, payments and advertising create proprietary transaction signals and customer-acquisition subsidies that a standalone bank cannot fully replicate.

NU's premium to domestic-bank comparables leaves little room for a credit normalization surprise after a sharp share-price move. The near-term setup is therefore more valuation-sensitive than technology-sensitive: another margin expansion or sustained benign delinquencies can support multiple durability, but any sequential rise in provisions, funding costs, or Mexican/Colombian losses would challenge the premise that model-driven underwriting is structurally superior. Over 6-18 months, successful SME and Colombia deployment could widen the addressable credit pool, but these are precisely segments where downturn loss curves are least seasoned.

Contrarian view: automation may be a sector table stake, not a moat. Model performance trained on a long benign or idiosyncratic consumer-credit history can decay rapidly if Brazilian real rates, employment, or regulatory constraints change borrower behavior. The market should not capitalize claimed inference savings until disclosure shows that savings are retained rather than competed away through lower pricing or higher marketing spend.

Key Decisions for Investors

• Maintain a tactical neutral-to-underweight NU for the next 1-3 months after the rerating; do not add on AI headlines. Reassess long exposure only if the next two quarterly reports show stable or improving 90+ day delinquency and cost of risk alongside credit growth above 25%, which would validate underwriting leverage.
• Express relative competitive risk via long ITUB / short NU in equal dollar amounts over 3-6 months. ITUB offers lower-beta exposure to Brazilian financial digitization and should outperform if AI commoditizes service while funding-cost and credit-cycle discipline reassert themselves; exit if NU's risk-adjusted margin remains above 12% with no deterioration in delinquency cohorts for two quarters.
• For higher-beta LatAm fintech exposure, prefer MELI over NU on a 6-18 month horizon: Mercado Pago's merchant and marketplace data can improve lending economics while its ecosystem supplies cross-sell channels. Risk is a broad LatAm consumer slowdown, which would impair both; size as a relative overweight rather than an outright macro bet.
• Set an earnings alert around sequential provision expense, Mexican unsecured-credit losses, and customer-acquisition cost. A 100-150 bp deterioration in risk-adjusted margin or a material guidance reduction would likely trigger both earnings-estimate cuts and compression toward domestic-bank valuation levels; conversely, avoid pressing shorts if these metrics remain stable while SME credit scales.