In October 2021, a Kenyan motorcycle-taxi financier borrowed 5 million dollars from a blockchain that runs exactly as written when it is called, holding and moving funds according to its code, without an intermediary able to change the outcome.">smart contract holding reserves of two or more tokens, supplied by users, that a decentralized exchange uses to execute trades automatically at a price set by a formula.">pool of crypto depositors. Tugende Kenya paid its interest, 53,400 dollars a month according to TechCrunch, until May 2023, then stopped. A review had already traced 1.9 million dollars to its sister company in Uganda, against the loan terms. The file closed in November 2024 with a final payment of 460,000 dollars. In June 2026 token holders of the lender, Goldfinch, voted to stop development and only collect what is left.
That story holds most of the lesson on DeFi lending in emerging markets. The funding was on-chain and the ledger was public. The judgement, the covenants, the collection and the legal claim all sat off-chain, where a smart contract cannot reach.
It follows our lesson on crypto remittances in Africa, because a reader who sends money home is also a potential lender to the same economies. So the question here is plain: what would your dollars fund, and who carries the loss?
Credit without collateral: why someone has to judge the borrower
Ordinary DeFi lending works because the borrower posts more crypto than they take out, a mechanism explained in our guide to crypto lending and borrowing. If the collateral falls, a bot sells it. Nobody needs to know the borrower’s name. A clinic in Nairobi or a shop in Manila has no pile of ether to lock, which is the reason it needs a loan in the first place. Take the crypto collateral away and a human has to decide who is likely to repay.
The need is real. The IFC and World Bank’s MSME Finance Gap report of March 2025 puts the unmet formal credit demand of small firms in emerging and developing economies at 5.7 trillion dollars, about 19 percent of their GDP, on 2019 data. That is a stock, not a yearly flow. On the household side, the World Bank’s Little Data Book on Financial Inclusion 2025 shows that 70.4 percent of adults in sub-Saharan Africa borrowed in 2024, and only 12.4 percent did so formally. People borrow all the time. They borrow from relatives, shopkeepers and savings groups, as the lesson on DeFi in Africa showed for financial services in general.
The crypto pitch from 2021 on was simple: yield-hungry depositors on one side, capital-starved small businesses on the other, and a blockchain to remove the middlemen. Depositors did arrive. The record turns at the next step: reaching small firms and getting repaid.
How DeFi lending in emerging markets is built: three layers, one on-chain
Every product in this lesson, whatever its branding, has three layers. Reading any pool starts with naming who sits in each.
- The origination layer finds the borrower, checks the books, sets the terms and monitors the covenants. It is done by people, in the borrower’s country or through a credit fund, and it never touches a block.
- The funding layer is the part on the chain: depositors send USDC to a contract, receive a token for their share, and see repayments arrive.
- The enforcement layer is a loan agreement under a national law, a security interest over assets, a bank account, and eventually a lawyer in the borrower’s jurisdiction.
A contract can stop paying out, mark a loan down on schedule and record who owns what. It cannot read a bank statement, notice that money moved between two sister companies, or repossess a motorbike. The designs differ mostly in who performs the first layer. Goldfinch used single-borrower pools with a junior tranche of “Backers” who were meant to do the homework, and a Senior Pool that spread deposits across all of them. Jia’s first pool on Huma, capped at 100,000 dollars, had Jia itself as the lender on the ground. Credix ran a marketplace for institutional lenders.
On-chain micro-loans: Jia and the small-ticket experiments
The closest thing to true on-chain microcredit was Jia. TechCrunch reported in May 2023 that it had raised a 4.3 million dollar seed round, plus 1 million dollars of committed on-chain liquidity, to lend 200 to 5,000 dollars to small businesses in Kenya and the Philippines at about 2 to 6 percent a month. That is 24 to 72 percent a year in simple terms. It sounds harsh until you divide a few dollars of interest by the cost of serving a 200-dollar loan; Jia’s chief executive, Zach Marks, told TechCrunch the rate was about a third of what a typical consumer fintech lender charged. Borrowers who repaid earned points, meant to become tokens redeemable against Jia’s profits.
Jia said in 2023 that a 100,000-dollar pool on Huma had funded more than 500 borrowers, and in February 2024 it told CoinGeek it had passed 2 million dollars of originations and 2,000 business owners in the Philippines and Kenya. These are company numbers relayed by the press, not audited ones, and other Jia posts give different borrower counts for the same early weeks. Even at face value, two million dollars is a rounding error against 5.7 trillion.
Jia today (site read on 2 November 2026) describes something else: invoice financing in the Philippines through Jia Financing, Inc., a financing company regulated by the Philippine SEC, advancing up to 90 percent of an invoice for fees of 1 to 5 percent. The page mentions neither an on-chain pool nor Kenya. It claims a loan loss rate under 1 percent, undated and unaudited as far as the page shows. The page does not say when or why the model changed. What it describes is a regulated lender, with no public pool in sight.
EthicHub took a similar path. DefiLlama still describes it as a protocol funding smallholder farmers, and shows 184 dollars of TVL with about 1.05 million dollars listed as borrowed on 2 November 2026. Its home page now sells specialty green coffee from Mexico, Colombia, Brazil and Honduras and shows no lending offer. Do not treat either name as a live DeFi credit product.
Huma itself is a good example of drift. DefiLlama showed it at about 369.9 million dollars of staking or liquidity, measured at current market prices. It is the main size metric in DeFi.">TVL on 2 November 2026, but it now describes its yield as coming from payment flows, and nothing in that figure shows small African or Asian businesses being funded.
The identity problem: Kiva Protocol in Sierra Leone
Kiva, the microlending charity, attacked the root of the problem rather than the funding. Its 2019 announcement described a national digital identity platform in Sierra Leone, built with the government and two UN agencies, on which citizens could hold a verifiable ID, with nationwide credit reporting planned alongside the central bank. No credit file, no underwriting: the logic was sound. The project did not last. A banner on the same page now reads: “Kiva will sunset operations of Kiva Protocol, effective June 30, 2022.” Kiva’s ordinary microloans carry on without it.
Tokenised private credit: Goldfinch, Centrifuge and Credix
Private credit means loans made outside banks and public bond markets, held to maturity by funds. Tokenised private credit puts the investor’s share of those loans into a token. The token can move in seconds. The loan underneath cannot.
Goldfinch: a Senior Pool that lent to credit funds
Goldfinch was the flagship. Andreessen Horowitz led a 25 million dollar round in January 2022, and GFI, its token, reached 32.94 dollars that month, according to Protos. Depositors put USDC into the Senior Pool. Backers took the first loss on individual pools in return for a higher rate. Warbler Labs, the company behind the protocol, handled monitoring and early recovery work until an adviser appointed by the Goldfinch Foundation took over most pools in 2024.
It was not an African lender, and it would be wrong to read it as one. The money mostly went to intermediaries. The risk review published by Warbler in October 2023 described 39 underlying companies in 20 countries, with an average exposure of 2.5 million dollars. Those included lenders in India, Colombia and Africa, carbon credit developers, and a US asset-backed vehicle. ALMA, a credit fund, alone had raised 39 million dollars. Protos cites a mid-2022 claim of more than 200,000 end borrowers in 18 countries; those people stood several steps away from the contract, and the figure is the project’s own pitch.
Three named failures and a group still in recovery
Tugende came first. The July 2023 update says the 5 million dollar loan was a two-year bullet with monthly interest, secured on all assets of the Kenyan company, with a loan-to-value covenant under 80 percent. It was funded entirely by the Senior Pool. The 1.9 million dollar transfer to Uganda surfaced in a December 2022 review of Tugende’s reporting, and the community was told in February 2023. Warbler blamed inflation, energy costs and a hiring push in 2022 for the Ugandan company’s troubles. The contract then did what contracts do well: it wrote the pool down on a timetable, in steps, over 120 days. The Senior Pool’s trailing yield fell from 7.81 percent to a projected 1.50 percent.
A term sheet in September 2023 suggested the net write-down could fall from 3.95 percent to under 0.79 percent. It fell through when the equity financing did not arrive. The final update of 16 December 2024 gave the result: 460,000 dollars, made up of 250,000 of loan recovery and 210,000 of closing fees, on top of roughly 1 million dollars of interest paid during the loan’s life. Separately, token holders had voted in September 2023 to put 1 million dollars from the treasury toward the hole. Our arithmetic: about 1.46 million dollars came back after default against 5 million of principal, and more than two-thirds of that came from Goldfinch’s own treasury rather than from the borrower.
Stratos was the second. In February 2022 the protocol lent 20 million dollars to Stratos, a US credit fund, so this was not an emerging-market loan at all. Warbler said in October 2023 that 13 million had gone to Threecolts, 5 million to REZI, a US apartment-rental company, and 2 million to POKT, a digital asset position that was not in the original proposed use of funds. The last two were expected to go to zero. Warbler disclosed that Stratos was an equity investor in Warbler and its founder an early Goldfinch adviser, and it chose to backstop those losses for the Senior Pool and Backers, excluding its own position and Stratos’s. Stratos sold the Threecolts loan at par in December 2023 for 13.04 million dollars, and Warbler’s backstop was due to add about 2.96 million to repay the Senior Pool’s 16 million that month, with Backers to claim later. So depositors here were saved by the private decision of a company tied to the borrower, which no pool design promises.
The third is the one that should worry anyone who trusts a risk report. On 20 October 2023, after looking through to the ultimate borrowers, Warbler wrote that apart from issues already disclosed, “everything looks good”. Five months later, just before the 10.15 million dollar Lend East pool matured on 3 April 2024, the company’s chief executive told Warbler it expected to repay about 4.25 million, which Warbler called “a loss of approximately 58% of the principal”. Lend East, a lender to fintechs in Southeast Asia and India, had said in its 2022 profile that it had seen no delays or defaults. Two problems were already known: a default by Oriente, 15 percent of the pool, and an unauthorised intercompany loan of about 7 percent. Notices of default went out in April 2024 and, from the Goldfinch Foundation, again on 25 October 2024. Warbler said in May 2024 that it believed the company was withholding funds to extract a settlement. These are one side’s allegations, and Lend East’s answer is not in the forum record. That record shows no court finding; an August 2025 proposal still refers to Lend East litigation.
There was a catch in the paperwork as well. Warbler told Backers in April 2024 that it was not a lender or agent under the Lend East documents, so its power to act was limited, and individual lenders had to sign up to appoint it. The claim was spread across many wallets, and no single creditor held a clear mandate.
The fourth borrower is really a group. A June 2024 update listed nine pools still outstanding. In July 2026, the team wrote that the legacy deals were “either waiting on payments or actively in recovery”. That post does not separate them and points readers to a Discord channel, so the forum gives no pool-by-pool result for Almavest, Cauris or Addem Capital.
| Borrower or pool | Where | Amount | Date | Outcome (latest documented) |
|---|---|---|---|---|
| Tugende Kenya | Kenya, group based in Uganda | 5.0M dollars | Oct 2021 | Default from June 2023; closed Nov 2024 with 460,000 dollars from Tugende, plus 1M from Goldfinch’s treasury |
| Stratos | US credit fund | 20.0M dollars | Feb 2022 | 7M expected at zero (Oct 2023); Threecolts loan sold at par for 13.04M (Dec 2023); Warbler backstop for the rest, excluding its own and Stratos’s positions |
| Lend East #1 | Emerging Asia | 10.15M dollars | Due 3 Apr 2024 | About 4.25M expected (Apr 2024); default notices; no court finding in the forum record |
| Almavest (4 pools) | Multi-country | 36.81M dollars | Due Nov 2024 to Apr 2025 | No pool-by-pool outcome published; legacy deals waiting on payments or in recovery (July 2026) |
| Cauris (3 pools) | Africa focus | 17.27M dollars | Due Nov 2024 to Feb 2025 | As above |
| Addem Capital | Asset-backed pool | 10.0M dollars | Due Apr 2025 | As above |
From Prime to wind-down in 2026
Goldfinch’s next product, proposed in December 2024, moved away from emerging-market borrowers. Goldfinch Prime would offer exposure to private credit funds run by large institutional managers, targeting 9 to 12 percent net of fees, and it was limited to non-US persons. It did not catch on. GIP-87, posted on 12 June 2026, said Prime “has not achieved the level of adoption needed to justify continued investment”. The Snapshot vote closed on 23 June with about 1.05 million GFI in favour and none against.
The plan moves recovery rights to a new US trust under Ted Gavin, Goldfinch’s chief restructuring officer, pays Warbler a fixed 150,000 dollars, stops development and keeps the old app running for at least six months after the last expected borrower payment. Prime investors were redeemed in full by 7 July 2026, with an extra month’s interest. Legacy lenders were not. Recovery, one of the authors wrote, could take “another 2 years or more”.
DefiLlama showed Goldfinch’s TVL at 2.25 million dollars on 2 November 2026, against a peak of 53.5 million on 21 February 2022. Its “borrowed” reading is now zero, but that does not mean the loans were repaid; it reflects what the data feed counts. Protos reported GFI below 7 cents in June 2026, down 99.8 percent. In December 2025 a forgotten 2020 test contract was also exploited for about 330,000 USDC.
Centrifuge: from Tinlake pools to tokenised funds
Centrifuge’s own history page says it began on Ethereum with Tinlake, tranched pools that financed real-world assets, and it frames its mission around the barriers small and medium enterprises face. By 2022 the customers were bigger: BlockTower Credit funded 220 million dollars of assets through it, with 150 million of senior capital from MakerDAO. The investor app on 2 November 2026 showed about 745.8 million dollars under management, of which the Janus Henderson AAA CLO fund held 568.3 million and the Janus Henderson Treasury fund 112.8 million. That is about 91 percent in two products, by our arithmetic. Direct fund access starts at 100,000 or 500,000 dollars, and these are the kind of product covered in our lesson on institutional DeFi.
Centrifuge does not describe this as a pivot; that is our reading of those pages. Its one emerging-market line in the app, a Schroder bond fund, showed about 200,000 dollars, and it holds bonds, not loans to small firms. Centrifuge today is not a lender to emerging-market small businesses.
Credix in Brazil: a book with no deposits
Credix built a Solana marketplace for private credit, and its site now presents it as credit infrastructure for Brazilian businesses. DefiLlama shows a TVL peak of 5.0 million dollars on 13 June 2022 and, on 2 November 2026, next to nothing in the pools against 10.64 million dollars of credit still marked outstanding. An issue filed on DefiLlama’s code repository on 8 October 2026 asks whether that sum can still be collected. It notes the figure has grown from 5.4 million and reads that as interest accruing on an unpaid book. That is one contributor’s reading, and public documents do not settle it. Keep the names apart: CrediX, a lending protocol on the Sonic chain, lost about 4.5 million dollars in an exploit on 4 August 2025, according to the hack records the same issue cites. That loss is recorded on Sonic, not against this Solana book, and public documents do not establish what, beyond the name, the two share.
On-chain tontines: the honest version of the idea
A tontine is the opposite of Goldfinch: no outside borrower, no foreign capital. A group pays a fixed sum each round and one member takes the pot, in turn, until everyone has had it. Cameroon has tontines and njangi, Kenya has chamas, Ghana has susu, Nigeria has ajo. Findex 2025 found that 28.1 percent of adults in sub-Saharan Africa saved informally in 2024, through a savings club or a person outside the family. For anyone early in the rotation the pot is an interest-free loan from friends; for anyone late it is forced saving. It runs on reputation, because everyone knows where you live. On the rules that apply to crypto in Cameroon and the CEMAC zone, see the dedicated guide: the social practice and a token version are different legal questions.
What a contract can and cannot enforce
Writing this as code is easy, and it solves one real problem: the treasurer. A contract holds the contributions, pays the pot in a fixed order, and cannot run off with the cash. If the group saves in a dollar stablecoin, members also avoid holding a falling local currency between rounds, with the risks set out in the lesson on stablecoins as dollar access.
It does nothing about the other classic failure. Take an example, with invented numbers: ten members, 50 dollars a month, a 500-dollar pot. The member who takes the first pot can stop paying the next day, leaving a hole of 450 dollars. A village has an answer to that: shame, family and the next tontine. A wallet address has none. The only code answer is a deposit, and to fully cover the risk, the first member must lock 450 dollars up front. At that point the pot is no longer credit, because a person who has 450 dollars spare does not need to borrow 500. Some designs add a default pool or partial deposits, which only shrink the problem.
What exists as of November 2026
The evidence is thin. The projects that appear are small and self-described, many of them hackathon entries on Celo, Base or Hedera, some aimed at diaspora groups. Those opened for this lesson show no audited usage figures, and none is recommended here. Add the ordinary risks: a contract bug can take everyone’s money at once, no deposit insurance applies, and members need a phone, data and a way in and out of the stablecoin.
Results and defaults: what the numbers say
Goldfinch’s own verdict in GIP-87 is short: about 100 million dollars of loans, and “a number of borrower pools experienced serious performance issues”. No audited loss total is on the public record. Protos reported cumulative losses past 18 million dollars in June 2026 and quoted a depositor putting the mismanaged money above 50 million, an unverified claim. The gap partly reflects what is counted: several pools have not been resolved. The governance side offers some hard points: GIP-67, a proposal to sell up to 1.2 million GFI to soften losses, passed in July 2024 by only 691,184 to 611,966, and in late 2025 token holders voted to pay it in GFI rather than USDC, a token that had lost nearly all its value. By August 2025 the Foundation had reported 534,000 dollars that year in legal, adviser and liquidator fees.
Lender yield against borrower cost
A comparison of rates shows where the money goes, and why it did not cover the risk.
| Rate | Figure | Date | Note |
|---|---|---|---|
| Jia borrower, small business | About 2 to 6 percent a month (24 to 72 percent a year, simple) | May 2023 | Company figures via TechCrunch; short loans, high servicing costs |
| Tugende, interest paid | About 12.8 percent a year | 2021-2023 | Our own calculation, not a contract rate: the 53,400 dollars a month reported by TechCrunch, on 5M dollars |
| Goldfinch Senior Pool, depositor | 7.81 percent trailing 12 months | July 2023, before the Tugende write-down | Falls to 1.50 percent after it |
| Goldfinch Prime target | 9 to 12 percent net of fees | Dec 2024 | Not emerging-market credit |
| 13-week US Treasury bill | 4.08 percent coupon equivalent | 30 Oct 2026 | Treasury.gov daily rates; shown for scale, different years |
Our 12.8 percent is a computation from a press figure, so read it as an order of magnitude. The gap between it and 7.81 is not all profit. It pays for protocol fees, Backer yield and pool mix, and the split is not published in the posts cited here. The simpler point: a spread of a few points a year cannot absorb a loan that returns under a tenth of its principal. “These loans were all very likely underpriced,” Ryan Rodenbaugh, founder of Wallfacer Labs, told DL News in August 2023. The forum’s yield figures exclude GFI rewards, whose dollar value has since nearly vanished.
Why on-chain did not mean lower default
Transparency arrived late and in the wrong place. The chain showed every payment, but the problems began in places it could not see: an intercompany transfer, a position outside the agreed use of funds, a portfolio update. Three of the failures above involve money moving to a related party or an unapproved purpose, and each surfaced through people reading reports, not through code.
Currency is the other weight. Pools lent in USDC. Tugende’s loan book was 89 percent Ugandan, and a small business in Nairobi or Lagos earns local money. If the shilling or naira falls, a dollar loan grows in the terms that matter to the borrower. The forum posts cited here do not blame currency for these defaults, so treat this as a general risk rather than a cause. Local-currency credit from a local lender and dollar credit from a global pool are still different products.
How to judge an on-chain credit pool before putting money in
This is education, not a recommendation of any pool. Add these checks to the general method in our protocol due diligence guide and the map of DeFi risk types; a smart-contract audit answers none of them.
- Name the underwriter, and find out who pays them and whether they own part of the borrower or are owned by it.
- Ask what secures the loan, in which country, and whether a local court has ever enforced that kind of security for a foreign lender.
- Find out who holds the legal claim, and who has authority to sue when the token is spread across thousands of wallets.
- Check how often reports arrive and who verifies them, because a monthly update written by the borrower is not monitoring.
- Read the concentration: one borrower or one fund manager can be most of a pool.
- Look for transfers to related companies being expressly barred, and for what happens if they occur.
- Compare the currency of the loan with the currency the end borrower earns.
- Test the exit, since credit tokens rarely have a deep market and redemptions depend on repayments arriving.
- Compare the yield with a Treasury bill on the same day, and ask what event would produce a total loss.
Rules for readers in the US, UK and EU
In the US, a token giving a share of a loan pool can be a security, and privately placed offers are often limited to accredited investors. Under Rule 501 of Regulation D, that means, for a person, net worth above 1 million dollars or income above 200,000 dollars (300,000 jointly) in each of the last two years with the same expected this year; the net worth test excludes the primary residence. Goldfinch Prime simply barred US persons from its on-chain product. A retail American reader should expect to be shut out of most of these offers. For the IRS, digital assets are property: income paid in tokens is generally taxed when received, and swapping tokens is a disposal. See our guide to US crypto taxes. How a written-down pool token is treated is a question for a tax professional.
In the UK, the FCA says that since 8 October 2023 firms promoting cryptoassets must be authorised or registered by it, or use another lawful route such as approval by an authorised firm, and that this applies to firms based overseas too. Promotions need prominent risk warnings and must not inappropriately incentivise people to invest. Its investor guidance adds that with an unauthorised firm you are likely to miss out on FSCS or Financial Ombudsman cover. Tax is in our UK guide.
In the EU, loans to businesses through a platform fall under the European crowdfunding regulation, which requires authorised providers and covers offers up to 5 million euros per project owner over 12 months. It does not apply to loans to consumers. An on-chain pool gives you none of those protections unless its operator holds that licence.
What this means for borrowers in emerging markets
For the Kenyan clinic or the Manila shop, the record is sobering. The chain moved capital across borders cheaply and fast. It did not make a loan cheaper to assess, or easier to collect, and those are the costs that make small loans expensive. Most of the capital reached intermediaries rather than small firms, and the survivors moved toward what can be priced and controlled: invoices from known buyers, regulated lenders, and funds of large managers. If your remittance dollars went into such a pool, they would fund a lender to a lender, and your claim would depend on someone else enforcing it abroad.
The next lesson in this path turns from lending to a single transaction: a constructed case of a merchant in Douala receiving USDC from a foreign client, with the real fees and rules that apply. On Goldfinch, the next thing to watch is the first report from the new recovery trust.
