Mexico cannot count its own lenders

Mexico publishes two official counts of its own non-bank lenders and they differ by 559. Neither arrives with a loan book. Small firms here borrow from suppliers first, banks second, and third from a tier no authority measures. The market does not lack lenders. It lacks countable ones.

The most repeated account of Mexican finance in 2026 reverses its own source

A trade piece published in July 2026, and cited steadily since, surveys Mexican financial services and reports remittances of more than US$70 billion and rising, citing Banco de México. Banxico's release of 3 February 2026 puts 2025 remittances at US$61,791 million, down 4.6% from US$64,746 million in 2024, the first annual decline since 2013. Mexico has never received more than US$70 billion in a year. The level is wrong and so is the direction: the number offered as evidence of growth is the first fall in twelve years.

The interesting part is not the error. It is that a survey of Mexican finance long enough to carry that claim contains no figure about credit at all, not a loan book, not a delinquency rate, not a count of lenders. The picture most allocators hold of this market is assembled from numbers nobody traced back. What follows is traced, and the tracing is the argument.

The largest source of credit to Mexican firms is other firms

Banxico surveys companies every quarter about the funding they actually used. In the first quarter of 2026, on a sample of 647 firms, 64.3% used supplier credit, 25.0% used commercial bank credit, 12.1% used other companies in their own corporate group, 1.1% used development banks, 1.1% used foreign banks and 0.2% used the debt market. Trade credit reaches 2.6 times more firms than the banking system does, and commercial bank penetration has fallen from 30.5% in the first quarter of 2024.

The gap that defines this market opens on firm size. Among firms with 100 employees or fewer, 23.0% used commercial bank credit in the quarter against 40.4% of larger firms, a difference of 17.4 percentage points inside one survey. The binding constraint is not refusal. Of the 86.9% of firms that took no new bank credit that quarter, 79.2% never applied, 3.9% were awaiting a decision and 3.8% were rejected. Demand is never tested, so nobody knows the true approval curve for a Mexican SME, the lenders included.

In Mexico, issuing a bond is what buys a lender a supervisor

Per the CNBV's own definition, a Mexican non-bank lender acquires a prudential supervisor in exactly three ways: an ownership link to a bank, a savings institution or a cooperative; funding itself with debt securities registered in the Registro Nacional de Valores; or applying voluntarily. Lending money to thousands of small firms is not on that list. Raising money from the public is.

The vocabulary, once, so the rest of this reads. A SOFOM, sociedad financiera de objeto múltiple, is Mexico's general purpose non-bank lending company: it lends, leases and factors, and it takes no deposits. Factoring is the purchase of a firm's unpaid invoices at a discount for immediate cash. The entidad regulada, or ER, is supervised prudentially by the Comisión Nacional Bancaria y de Valores. The entidad no regulada, or ENR, which is the majority form, registers with CONDUSEF, the financial consumer protection agency, and is supervised by the CNBV for anti-money-laundering purposes only. The securities in the second route are certificados bursátiles, universally shortened to CEBURES, and where they are backed by a ring-fenced pool of loans held in a trust rather than by the issuer itself, the operation is a bursatilización, a securitisation.

Fitch states the mechanism from the outside: «Corporate governance requirements for Sofomes apply only to bank-linked entities, with additional financial requirements for Sofomes issuing local-market debt». The unregulated tier is unmeasured because it never issued, and it never issued because it is unmeasured.

The central bank counts its non-bank lenders from a consumer protection registry

Banco de México reports that at December 2025, the latest date available to it, the Buro de Entidades Financieras held information on 2,336 SOFOM ENR with operating status, including those lending only to related parties. The footnote to that table says where the count comes from: RECO, CONDUSEF's registry of fees and commissions, a consumer protection database rather than a prudential return. The same footnote labels the series preliminary and records corrections going back to June 2023.

The ambiguity is printed inside one table. Banxico's own count of unregulated SOFOMes carrying a positive loan book is 1,777 at March 2026, in the same document that carries the 2,336. Two official readings of one sector, 559 entities apart, one of them lagged six months, in a report published in June 2026.

The gap is neither new nor narrowing, and the basis moved underneath it: Banxico records in the June 2026 edition that its entity count now means entities in operation, where earlier editions meant entities in operation with assets above zero. Anyone charting that series as a clean trend is selling a slope the source does not support.

A registry entry certifies form filling, not lending

SIPRES, the Sistema de Registro de Prestadores de Servicios Financieros, is CONDUSEF's register of financial service providers. Mexico's Supreme Court has described what losing an entry means: the entity lost its registration because it «omitio validar su informacion corporativa por dos trimestres consecutivos», and the effect of cancellation is the loss of its status as a financial entity (SCJN, ejecutoria 32843). Registration attests that a corporate-data box was ticked in two consecutive quarters. It attests to no capital, no loan book, no audited accounts and no loss performance.

The distance between the register and the market was measured at maximum stress, and that measurement is eleven years old. In February 2015 CONDUSEF reported 3,025 SOFOMes in operation, of which 500, or 16%, were recorded as not locatable; two months later SIPRES listed 3,012 operating, 138 had applied to renew their registration and 15 had obtained it. Neither figure is a current count and neither should ever be quoted as one. The regulator reached the same answer on a smaller sample four years later: of 77 SOFOM ENR entering cancellation, 59, or 77%, had no active loans at all.

The CNBV declines to warrant the count of the entities it supervises for anti-money-laundering purposes. Its own page states that the sector total for SOFOM ENR comes from CONDUSEF and that, because CONDUSEF is updating the systems holding it, «las cifras pueden variar», with SIPRES recommended as the principal source.

We tested that principal source. Of the SOFOM ENR records SIPRES listed as operating when we queried the public register on 5 August 2026, 56.8% had not updated their own record in more than a year, 34.0% in more than two years and 19.9% in more than three, and six entities listed as operating simultaneously carried the register's own not-locatable flag. That is our computation on the register's published last-updated field, offered as such.

One consequence deserves stating flatly. The last time any Mexican authority published a consolidated loan book and delinquency ratio for this tier was 13 June 2019: 1,652 SOFOM ENR, MXN 547,806 million of total loans, an IMOR, índice de morosidad or past-due ratio, of 5.1%, and just over 6 million credit contracts. No newer consolidated figure has been published as of 5 August 2026.

The one tier with public, monthly, audited numbers barely lends to anyone directly

NAFIN publishes its lending monthly: at the close of January 2026 it had extended MXN 32,304 million of credit, guarantees and induced balance, up about 40% year on year, of which second-floor credit through financial intermediaries was MXN 20,665 million and first-floor direct credit MXN 503 million. The second-floor book is 41 times the direct book. The state is a wholesaler here, and the operative question for any lender in this market is whether it qualifies as an intermediary.

The guarantee stack is where state money does most of its work. Under Plan México, NAFIN and Bancomext committed up to 120,000 million pesos for Mipymes, with a 70% guarantee on credits up to 20 million pesos in priority sectors and an 80% guarantee on first-time credits up to 5 million pesos. NAFIN caps the end-borrower rate on its Plan México simple credit at 14.50% fixed, subject to each intermediary's policy. One caveat belongs on the same page: on 13 May 2026 S&P revised the outlook to negative from stable on twelve Mexican financial institutions following a sovereign action, NAFIN, Bancomext, Banobras and FEGA among them, while affirming the ratings.

Commercial banks did not leave the SME, they grew everything else faster

Bank MSME loan stock reached MXN 565.4 billion at end-2024, up 57% in nominal pesos from MXN 359.3 billion in 2013. Over the same period the MSME share of outstanding business loans fell from 24.9% to 13.0%, and the MSME share of new business lending fell from about 23% in 2013 and 2014 to 6.03% in 2024. Price moved with share: the average MSME rate was 15.59% against 10.71% for large firms in 2024, a spread of 4.9 percentage points, while MSME non-performing loans stood at 3.4% in both 2023 and 2024, the lowest level since collection began. The OECD attaches its own warning that national measurement of SME non-performing loans varies.

The stock alone tells the wrong story. The share and the spread tell the right one, and the policy target sits on top of them: Plan México sets a goal of financing 30% of SMEs by 2030, against the 10.7% of MSMEs that had financing from any source in 2023. Tripling penetration in six years is the demand for capital, stated by the state, and it cannot be executed through a tier whose loan books cannot be seen.

The unregulated tier is the fragmented one, and it is the one nobody reports on

Where its money comes from is published in aggregate and nowhere else. Unregulated non-bank intermediaries drew 45.0% of their funding from commercial banks and 25.0% from development banks, with the development bank component down 15.0% over the period. Across all non-bank intermediaries, debt issuance was 15.3% of total funding at March 2026, and total funding to the unregulated group contracted for a second consecutive month at the close of March 2026.

The banks themselves are barely exposed. At March 2026, commercial banks' loans to and holdings of securities of non-bank intermediaries were 1.8% of their total assets and development banks' 5.6%, which Banxico states is small enough that a default there would not be a systemic risk. A tier the domestic banking system funds to 1.8% of its balance sheet has to find the marginal peso somewhere else.

One non-bank tier behaves like an asset class, and it is the one that issues paper

AMSOFAC, the association of leasing, credit and factoring companies, has 54 members, reports a portfolio of about MXN 317 billion at the first quarter of 2026, and about MXN 150 billion of debt securities outstanding on Mexico's two exchanges. The tier is legible because it issues, and it issues because it is legible. Serfimex Capital, a SOFOM ENR, placed SERFICB 25 for MXN 400 million under a revolving programme, rated HR AAA (E), with a disclosed historical default rate of 2.3% on the originated and serviced portfolio. A published default rate on a named book is what issuance buys. Nothing comparable exists for a SOFOM that does not issue.

Factoring itself has no register to screen. The standalone empresas de factoraje financiero licence class shows zero entities in operation in SIPRES when queried on 5 August 2026, because the activity moved inside the SOFOM wrapper after the 2006 reform of the LGOAAC. In the register a factor is indistinguishable from any other SOFOM ENR. Mexican factoring turnover was EUR 31,410 million for 2024, growing 6.1% in 2025, penetration is about 2.5% of GDP against about 16% in Chile, and 3% of Mexican firms have ever used factoring since they began operating.

Fintech lending is counted in companies and never in pesos

Mexico had 795 local fintechs and 316 foreign entrants operating in the country, with the lending vertical at 170 local and 35 foreign, 205 in all. The report carries no breakout of how many of those lend to SMEs, and no aggregate loan book, origination volume, delinquency rate or ticket size across its 45 pages. The sector's own census cannot say how much anyone lends.

Almost none of them are authorised entities. The Ley Fintech authorises two types: IFPE, electronic payment funds institutions, and IFC, crowdfunding institutions. Between 2018 and the end of 2025 the interinstitutional committee received 198 applications and granted 89 authorisations, 62 IFPE and 27 IFC. Banxico counted 25 crowdfunding institutions in operation at March 2026, holding 0.0% of financial system assets. The rest of the 205 operate as SOFOM ENR, with no obligation to publish a loan book or a delinquency ratio.

The three legible SME lenders are legible for reasons that have nothing to do with Mexico. MercadoLibre's merchant loan book stood at US$2,285 million gross against a US$901 million allowance at 31 March 2026, and its Mexican funding is disclosed instrument by instrument, four trusts totalling US$1,215 million priced at TIIE plus 2.20% to 2.80%. None of that is Mexican disclosure. The most transparent SME lender operating in Mexico is transparent because the United States securities regime compels it. Covalto files quarterly CNBV statements and carries a public rating because it acquired a bank in 2021. And the sector's best known SME lender states the inversion out loud: «We don't have to ask for financials or audited financials», its chief executive told the IFC in July 2026, describing 98.6% automated credit decisions. It solved its borrowers' information problem with data and reproduced the same problem one level up, toward its own funders.

The 2022 losses arrived as disclosure events, and the collateral kept paying

Four of the six Mexican non-bank lenders that ever issued international debt defaulted, and in each the public record shows a document being corrected or withheld rather than a delinquency line rising. One issuer missed a CHF 170 million bond on 9 February 2022 having filed no financial statements after the fourth quarter of 2021, and its own Chapter 15 disclosure of 12 March 2025 sets recovery to unsecured creditors at 19% to 20%. Another announced on 20 April 2021 that it would restate its 2018 and 2019 accounts. A third missed MXN 85 million of local certificates and a US$14.5 million coupon on 17 January 2023 after disclosing errors in its records of financial assets. A fourth stopped paying on US$2,300 million on 9 August 2022, entered concurso mercantil, the Mexican court-supervised insolvency procedure, and restarted origination in April 2025. Bloomberg Linea put the aggregate loss to foreign bondholders across the episode at close to US$5 billion. The full sequence is set out in the eleven dark months.

A rating agency supplies the contradiction that makes this an information story rather than a credit story: «Mexican ABS from NBFIs remained resilient throughout the COVID-19 pandemic, and they continue to perform well». The securitised pools kept paying while the corporate issuers that originated them failed. Mexican SME credit was not a bad asset class. It was an unreportable one.

What came back after 2022 was not trust, it was structure

Aggregate appetite for Mexican credit recovered in full. ECLAC recorded ninety Latin American corporate issuers placing US$89.3 billion in international bond markets in the first nine months of 2025, up 63% year on year, with Mexico taking 34.4% of the total. The non-bank tier did not participate: Fitch measured non-bank financial institutions at about 7.4% of Mexican financial system assets and under 1% of total debt market issuance as of June 2026.

So the money arrives wrapped. IDB Invest describes what investors, itself included, fell back on: «secured facilities with visibility and control of the flow of funds from collateralized loans, corporate guarantees, the use of special purpose vehicles and more sophisticated structures like master trusts with a lockbox mechanism and warehouse facilities». Five pieces of engineering do the work, and each replaces one thing the funder cannot verify.

• A warehouse line is a secured revolving facility whose proceeds may only fund newly originated loans meeting agreed eligibility criteria.
• A Mexican fideicomiso or special purpose vehicle holds those loans outside the originator's insolvency, reached by cesión de derechos, the assignment of credit rights.
• A master collection trust receives all collections directly, so cash never passes through the originator's own account.
• Aforo, or overcollateralisation, follows from advance rates of 70% to 90% of portfolio value, which is 10 to 30 points of cushion.
• A pre-contracted substitute servicer stands ready to collect if the originator stops.

State money now guarantees the same architecture. The SOFOMCB 26X trust pools five originators, carries partial guarantees from FIRA and NAFIN of up to 18% each in pursuit of a AAA local rating, and is anchored by IDB Invest at US$57.26 million equivalent. Every element of that stack is a purchased substitute for a loan-level view of a live book.

The distance between a measured lender and an unmeasured one is a reporting spread

Put two 2025 transactions side by side. A Mexican non-bank placed US$120 million of senior unsecured notes at a 15.50% dollar coupon, 1.7 times oversubscribed, in about July 2025 (Pareto Securities). Weeks later a rated multilateral funding the same end market placed MXN 2,500 million in Mexico at TIIE de Fondeo plus 31 basis points, with a book covered 1.9 times. Both put money into Mexican SME credit. What separates them is what a buyer can verify.

The institutions that supervise the system record the same absence

The IMF puts it in the language of surveillance: «The data provided to the Fund has some shortcomings that somewhat hamper surveillance». The Financial Stability Board puts it as a hole in the global statistics rather than a Mexican failing: in regulatory and statistical data «a specific category for private credit funds is missing», and identifying the most important non-bank market participants is «a key data gap».

The development bank that writes cheques into this sector measured it and published the result. IDB Invest ran the first sector-wide corporate governance self-assessment of Mexican non-bank financial institutions with three industry associations, covering more than 150 institutions, and stated the premise plainly: «The comparative diagnosis of NBFIs is relevant because, unlike regulated banks, public and comparable information on governance practices is very limited». Transparency scored 51% and management 38%, fewer than 30% of institutions include independent board members, and an executive succession plan is in place at 29% of lenders with a book under MXN 250 million, 41% between MXN 250 million and MXN 1,000 million, and 55% above MXN 1,000 million. Measurability scales with the size of the book, not with the quality of the credit.

Even the tier's most quoted statistic is unmeasured. ASOFOM says 60% of Mexican SMEs get their first credit from a SOFOM; IDB Invest says 65%. Neither publishes a methodology, ASOFOM contributed to the IDB Invest study, so the two are not independent, and the single most repeated fact about this tier is a trade claim asserted twice at two different values.

The sector diagnosed itself correctly and then priced the cure. «Nadie nos va a venir a rescatar de la falta de fondeo, tenemos que probarle a los mercados financieros que somos empresas viables», said Javier Garza Hoeffer, then vice president of ASOFOM and listed in August 2026 as its national president, describing a benchmark whose results would be made public so that funds would begin to know the association. Three years on, the benchmark exists, and the full study and its comparative databases are a membership benefit. The measurement was built, and then scoped to members.

The only official price comparison ever made is five years old

Banxico's Reporte de Indicadores Básicos de Creditos a las PyMEs is the only official product that ever gave SME loan volumes, rates, borrower counts and concentration. Four editions exist; the last was published on 15 November 2022 with data to September 2021, and there is none newer as of 5 August 2026. It holds the only bank-versus-non-bank price comparison a Mexican regulator has published: commercial bank loans carried an average rate 200 basis points below loans granted by regulated SOFOMes. Its scope was commercial banks plus regulated SOFOMes. The unregulated majority was never in it, and is priced by nobody.

Nor is the domestic institutional peso arriving to fill the gap. Afores held 51.5% of assets in government debt in the first quarter of 2026 and keep structured instruments below 10% of portfolios against a 30% regulatory ceiling, while CKDs outstanding stood at MXN 256,370 million at September 2025, down 23.3% in real terms year on year (SHCP). One Mexican pension manager has crossed into this asset class in a disclosed transaction, in a MXN 7,422 million financing of 18 December 2024. One absence belongs here as data too: across 2024 to 2026 we found no disclosed facility to a Mexican non-bank lender from FMO, DEG, Proparco, FinDev Canada, CAF or BLADEX. Most of the development finance universe is not in this market at all.

What closes the gap is a loan tape, not another survey

A loan tape is the loan-by-loan file a credit investor uses to reunderwrite a book: origination date, amount, rate, term, collateral and arrears bucket for every contract, refreshed on a fixed cycle and reconciled to the servicer's cash. It is the unit of evidence institutional capital consumes, and it is the one artefact this tier has never been required to produce. What a fund asks for inside it is a separate piece.

Effect is not a bank, a lender, a broker-dealer, an investment adviser or a credit rating agency. Nothing here is investment advice or a recommendation on any security, and no outcome is promised. Every figure above carries its source and its as-of date so that a reader can check it rather than take it.