MELI — Thread 2026-08-08 (EN)

1/11 MELI reported Q2. Revenue up 50% YoY, the fastest pace in four years. And the stock fell 8% on the print. Why? Operating income dropped 17%. The market read "margins destroyed." I read the full shareholder letter and I think it's wrong.

2/11 What MELI is: LatAm's e-commerce + fintech ecosystem in one app. Marketplace (GMV $22B this quarter) + Mercado Pago (TPV $101B, +56%) + logistics + ads + a $16B credit book. The whole thing sits inside one login. That "all in one" is the entire point.

3/11 Thesis 1: the margin drop is deliberate, not an accident. Op margin 6.7% vs 12.2% a year ago. They cut the free-shipping threshold in Brazil, added PIX discounts and ramped card issuance. Same playbook as 2016 and 2024. Worked both times.

4/11 Thesis 2: the biggest bear risk was credit. It didn't show up. The book grew 75% to $16B and 15-90 NPLs sit near record lows: 4.6% on the card, 7.0% total. You don't scale a loan book 75% with delinquencies at lows by luck. That's underwriting.

5/11 Thesis 3: card NIMAL printed -2.5% and sounds awful. It's cohort dilution — 2.6M new cards issued vs 1.6M a year ago. Provisions front-loaded, yield back-loaded. A cardholder is 2-3x likelier to stay ecosystemic, and that user generates 70% more GMV.

6/11 Anti-thesis 1: what if the compression is structural, not transitory? Op income fell in absolute dollars, not just in margin. And H1 adjusted free cash flow collapsed to $158M from $512M a year ago. If the engagement payoff never lands, it's value destruction.

7/11 Anti-thesis 2: Argentina is decelerating (FX-neutral GMV +38%, soft consumption) and Mexico's contribution margin compressed ~4pts QoQ on POS device costs. Net debt rose $1.7B to fund the credit book. You're growing by levering up. In LatAm that's never free.

8/11 My variant: consensus sees margin compression as a problem. I see a deliberate arbitrage — eat the credit drag today to lock in an ecosystemic user worth multiples more in lifetime value. The disagreement isn't this quarter's margin. It's the terminal one.

9/11 One piece the market ignored: AI is funding the reinvestment. Product development fell from 8.4% to 7.2% of revenue even as AI spend rose ~$80M. Code submissions +110% YoY, agents reviewed 500K+ submissions. They're scaling output without scaling engineers.

10/11 What I'm watching next 2-3 quarters: does card NIMAL inflect back up as cohorts mature? That's the whole thesis. Plus op margin bottoming and adjusted FCF recovering. If margins keep sliding and cash doesn't turn, the market was right and I was buying an excuse.

11/11 Bottom line: a quality business whose margins got punished for deliberate reinvestment the market is misreading. I hold one share, underwater, and I'm not adding on this print — I want to see the margin inflect first. Position: long. Not financial advice.