Phase Space AI

Research

MercadoLibre [MELI]

MercadoLibre [MELI] — Tier-2 Research

As of 2026-07-30. Spot $1,833.15. Diluted shares 50.697m. Market cap $92.94bn. Sources: FY2025 10-K (filed 2026-02-25, meli-20251231.htm), Q1'26 10-Q (filed 2026-05-08, meli-20260331.htm), Alpha Vantage normalized statements (81 quarters), Alpaca daily bars.


1. The analytical question: is the take rate durable, and is growth volume or monetisation?

MELI is two businesses stapled together and the 10-K discloses both volumes. The decomposition is the thesis.

Disclosed operating metrics

FY2023 FY2024 FY2025 Q1'25 Q1'26
GMV ($m) 44,749 51,467 65,037 13,330 18,951
GMV growth +15.0% +26.4% +42.2%
Total payment volume ($m) 146,738 196,660 277,823 58,303 87,186
TPV growth +34.0% +41.3% +49.5%
Acquiring TPV ($m) 115,953 142,200 188,105 40,317 55,993
Items sold (m) 1,404 1,787 2,429 492 722
Fintech MAU (m) 46 61 78 64 83
Unique active buyers (m) 85 100 121 67 84
NIMAL 36.2% 28.2% 22.4% 22.7% 17.8%

Take rates, computed

Commerce. Commerce revenue includes first-party product sales, which are not a take rate. Both cuts:

Q1'25 Q1'26 Δ
Commerce revenue / GMV (headline) 24.78% 25.69% +0.91pp
Commerce services revenue / GMV (ex-1P) 20.16% 19.73% −0.43pp

Net product revenues were $1,130m in Q1'26 vs $615m in Q1'25 (+83.7%), against total revenue +49.0%. The headline commerce take rate rises only because first-party inventory sales are growing at twice the company rate. On the marketplace-services basis the take rate is falling. 1P revenue is booked gross against inventory cost; it adds revenue optics, not marketplace economics.

The company confirms the mechanism in its own words: the FY2025 commerce services increase was "mainly related to a 26.4% increase in gross merchandise volume and higher flat fee contributions for low gross merchandise volume transactions" — i.e. genuine monetisation on small baskets — but also that shipping carrier costs netted against revenue fell $42m "mainly due to an increase in the share of shipping services where we act as principal, as opposed to agent". A principal/agent flip inside the same line inflates reported revenue with no economics attached. That is the same class of distortion as BKNG's merchant shift and RBA's inventory mix, found here inside a single sub-line.

Fintech.

Q1'25 Q1'26 Δ
Fintech revenue / TPV 4.51% 4.56% +0.05pp
Fintech revenue / Acquiring TPV 6.53% 7.10% +0.57pp

Flat on total TPV. But the fintech "take rate" is not a take rate — it blends payment-processing fees with credit interest income. FY2025's $3,981m fintech revenue increase splits $2,258m Credits / $1,716m Financial services. In Q1'26 the split is $866m Credits / $474m Financial services: 64.6% of fintech's incremental revenue is lending, not payments. Payment monetisation is roughly flat; the fintech growth story is a balance sheet expanding.

Verdict on the analytical question: growth is overwhelmingly VOLUME, and the volume is real. GMV +42.2% and TPV +49.5% in Q1'26 are the load-bearing facts, and neither is an accounting artifact. But the revenue growth rate of +49.0% is flattered by three separate gross-ups — 1P product sales, the shipping principal flip, and credit interest booked gross against provisions in opex — so revenue overstates the rate at which marketplace economics are compounding.


2. Accounting quality — the credit book

This is where MELI can manufacture growth, so it gets the most scrutiny. Findings, in order of severity.

2a. Provisions are the entire margin story, and they are winning

FY2023 FY2024 FY2025 Q1'25 Q1'26
Provision for doubtful accounts ($m) 1,050 1,858 3,091 603 1,244
% of revenue 7.0% 8.9% 10.7% 10.2% 14.1%
Operating margin 14.1% 12.7% 11.1% 12.9% 6.9%

Q1'26 operating margin fell 6.0pp year over year. Provisions rose 3.9pp of revenue over the same period — 65% of the entire compression. MELI attributes the FY2025 rise to "the increase in originations growth at 61% (mainly related to the credit card, consumer and merchant products)."

2b. Credit quality metrics are stable; the economics per dollar lent are not

Loans receivable, gross: $12,508m (2025) vs $6,573m (2024), +90.3%.

2024 2025
Total past due / portfolio 26.8% 26.3%
91–360 days past due (non-accrual) / portfolio 17.5% 16.8%
Allowance / gross loans 25.5% 25.1%
Renegotiations / portfolio 1.4% 1.6%
Write-offs ($m) 885 1,796

This does not support the "loosening credit to manufacture revenue" hypothesis on the stated delinquency metrics. Past-due and non-accrual ratios improved slightly while the book nearly doubled. I looked specifically for the AAOI/AIOT pattern and it is not here.

What is here is NIMAL: 36.2% → 28.2% → 22.4% → 17.8%. Net interest margin after losses has fallen 18.4pp in nine quarters. Management's own definition — credit revenue less funding costs and provisions, over average gross loans — makes this the cleanest available measure of unit economics on the credit book, and it has been cut in half. The stated cause is mix: credit cards went from $2,639m to $5,656m of the book (40.2% → 45.2%), and credit cards are lower-yield. So the growth is real and the risk is contained, but each incremental dollar of loan book earns roughly half what it earned in 2023. A business growing volume 90% while halving unit economics is a different investment from one compounding at a stable margin, and the reported revenue line does not distinguish them.

2c. Receivables factoring — found, quantified, and it is material

Per the DATA_DEFECTS instruction to read the cash-flow statement and notes rather than trusting any DSO variant:

"The gain arising from sales of credit card receivables, net of the costs recognized on sale of those credit card receivables, is $1,124 million and $1,246 million, for the years ended December 31, 2025 and 2024" — FY2025 10-K, Note 2. "$367 million and $255 million, for the three-month periods ended March 31, 2026 and 2025" — Q1'26 10-Q.

MELI runs a receivables-sale programme and books the gain in revenue. Magnitude: - FY2025: $1,124m = 3.89% of revenue; FY2024 $1,246m = 6.00%. - Q1'26: $367m = 4.15% of revenue; Q1'25 $255m = 4.30%.

Two consequences. First, the FY2025 gain fell 9.8% in absolute dollars while revenue rose 39.1% — so this line was a −0.6pp drag on FY2025 growth, and the underlying non-factoring revenue grew faster than reported. That is the favourable direction, and worth saying. Second, no DSO computed from the balance sheet is meaningful here, because the receivable is sold before it ages. Accounts payable also carries $1,371m (2025) / $898m (2024) relating to credit-card receivable sales in Argentina, and there is a supplier-finance programme with $616m (2025) / $425m (2024) of confirmed obligations outstanding. Both are debt-like and neither appears in the screen's net-cash construction.

2d. Retired and redefined metrics — three found

  1. "Number of items shipped" was retired as of December 31, 2024. Rationale given: it has converged with "items sold". Plausible, and unusually well-explained — but it is a retirement.
  2. "Unique Active Users" was replaced by "Fintech monthly active users" and "Unique active buyers" as of January 1, 2024 — a metric split that breaks the pre-2024 series.
  3. Both GMV and "Unique active buyers" were redefined from Q2 2025 to include food-delivery transactions. So the FY2025 GMV of +26.4% and the Q1'26 +42.2% are not like-for-like with FY2024, and the company does not quantify the food-delivery contribution. This is an unquantified upward bias in the single most important operating metric in the memo. Flagged as INDETERMINATE in magnitude; the direction is known.

2e. Argentina, FX and hyperinflation — disclosed, and it matters

Argentina has been accounted for as highly inflationary since 1 July 2018, functional currency USD. Q1'26 disclosures: - Argentine quarterly inflation 9.4%; inter-annual average 32.7%. - Official rate 1,382.00 (31-Mar-26) vs 1,455.00 (31-Dec-25); average rate 1,417.76 vs 1,057.00 — a 34.1% year-over-year devaluation. - Brazil: "average decrease of Brazil's exchange rate against U.S. dollar of 10.1%" — a BRL appreciation, a tailwind to reported USD revenue.

MELI publishes the FX-neutral bridge, which is the disclosure the brief asks for:

Q1'26 As reported FX-neutral
Revenue growth +49.0% +45.5%
Gross profit growth +39.4% +38.5%
Operating expense growth +61.9% +57.8%
Income from operations −19.9% −12.2%

FX contributed +3.5pp of the +49.0% headline revenue growth, and operating income declined on both bases. Note the second row of that table is the important one: FX-neutral opex grew 57.8% against FX-neutral revenue of 45.5%. The margin compression is not an FX artifact.

2f. Customer concentration — clean

"For the three-month periods ended March 31, 2026 and 2025, no single customer accounted for more than 5.0% of our net revenues and financial income."

Tested for the AAOI pattern; not present.


3. Mention frequency (required core metric)

EARNINGS_CALL_TRANSCRIPT coverage was not retrieved within the time box; INDETERMINATE, recorded rather than fabricated. The filing-based substitute, which is auditable: across the FY2025 10-K and the Q1'26 10-Q, "loans receivable" appears 72 times in the 10-K, "credit card" 94 times, "securitiz" 37 times, and "take rate" zero* times in either document. The company does not use the vocabulary of marketplace monetisation anywhere in its own filings; it describes itself in volume and credit terms. That is itself a finding about where management's attention sits.


4. Mechanism — the named driver

Named, evidenced, and forward: the credit card product. It went from $2,639m to $5,656m of gross loans in twelve months (+114%), is now 45.2% of the book, is explicitly named as the driver of both the 61% origination growth and the NIMAL decline, and carries a disclosed off-balance-sheet unused commitment exposure against which CECL is separately provided. Its expansion is the single mechanism that determines whether MELI's margin troughs in FY2026 or keeps falling: credit cards front-load CECL at origination while interest accrues over the life of the balance, so a book growing at 114% structurally understates its own steady-state margin. When origination growth decelerates to the revenue growth rate, the provision line stops outrunning revenue and roughly 3–4pp of operating margin returns without any operational improvement. That is the underwritable claim.


5. Defects found in the screen record and in vendor data

Item Screen / vendor said Correct Magnitude
net_cash −$4,277m (Cash + STI − LongTermDebt) −$5,748m per the company's own net-debt reconciliation (Q1'26 10-Q), lease-inclusive $1,471m, 25.6% understated; EV understated 1.5%
op_margin_pct = 11.1% at margin_period 2025-12-31 labelled as the margin that is FY2025, not TTM. TTM through Q1'26 is 9.6%; latest quarter 6.9% −1.5pp vs TTM, −4.2pp vs latest quarter
terminal_margin 0.111, basis "own trailing operating margin" it is the FY2025 figure on a declining trajectory (14.1→12.7→11.1→9.6), i.e. a ramp endpoint in the wrong direction see valuation
AV OVERVIEW.OperatingMarginTTM 6.91% true TTM 9.6% single-quarter defect confirmed on MELI (6.91% ≈ Q1'26's 6.9%) — same failure recorded for ATRC/BSX/DXCM
normalized_fcf.py trailing FCF margin 37.3% company's own adjusted FCF was −$56m in Q1'26 (−0.6% of revenue) and +$58m in Q1'25 ~38pp, direction manufactures a false PASS — see §6
AV shortLongTermDebtTotal $12,345m 10-Q total debt $12,345m clean, 0.0% error — reported per the "verify per name" instruction
AV SPLITS {"data": []} no splits; 50,697,182 weighted shares matches the 10-Q cover exactly clean
GMV / unique active buyers definitions treated as a continuous series redefined from Q2 2025 to include food delivery, contribution unquantified INDETERMINATE magnitude, known upward direction

6. The FCF defect — the most important number in this memo

normalized_fcf.py --ticker MELI returns a 37.3% trailing FCF margin and a 25.7% "normalized" margin, computed as (operating cash flow − capex) / revenue. Both are wrong for this company, by a wide margin, and in the direction that manufactures a false PASS. From the FY2025 cash-flow statement:

Net cash provided by operating activities                    12,116     41.9% of revenue
  of which: Funds payable to customers                       +5,341     payment float, not earnings
  of which: Amounts payable, credit/debit card transactions  +1,592     payment float, not earnings
Investing: Changes in loans receivable, net                  (6,659)    <- the loan book, in INVESTING
Investing: Property, equipment and intangibles               (1,343)
Basis FY2025 FCF Margin
Tool's basis: OCF − capex $10,773m 37.3%
Less loan-book growth (a real cash outflow, parked in investing) $4,114m 14.2%
Less payment-float growth (a funding source, not earnings) −$2,819m −9.8%
Company's own "adjusted free cash flow", Q1'26 −$56m −0.6%

A 47-percentage-point spread depending on definition. The tool's OCF−capex convention silently assumes that a lending business's loan book is free, because US GAAP puts loan originations in investing. DATA_DEFECTS records the sign-dependent interim-FCF bias; this is a scope-dependent one, and on a fintech it is larger. I use 14.2% as the base and show the full surface in the valuation.


7. Same-quarter DSO

Reported receivables are not the right denominator here because credit-card receivables are sold (§2c) and loans sit in a separate caption. On the accounts-receivable caption alone, Q1'26 receivables rose $887m against $71m in Q1'25 (cash-flow statement), against revenue +49.0% — a working-capital build roughly 12x larger than the prior year on revenue 1.5x larger. Directionally adverse, but uninterpretable as a DSO because the balance is partially sold each period. Recorded as INDETERMINATE with the reason, per the rule that a false disclosure is worse than an absent one.