Mekong Capital is not buying AI-native startups. Vietnam’s best-known private equity firm is sitting out a boom that will cost America’s six largest cloud companies more than US$785 billion this year, and the asset carrying its book is a chain of pawnshops. That chain has gone from 11 branches to 972 since Mekong put money in, and it now plans to sell 10 percent of itself on a Vietnamese exchange.
The money went in during 2016, out of the firm’s third fund, into a Ho Chi Minh City lender called F88. It had eleven shops. It handed over small amounts of cash against a motorbike, a set of registration papers or a phone, and held the item until the borrower paid the loan back.
Ten years later there are 972 of them.
The net loan book has grown three hundred times across that stretch. FiinRatings, Vietnam’s credit rating agency, lifted F88 to BBB in October 2025. In May 2025 the State Securities Commission cleared the company to go public, the first lender outside Vietnam’s banking system to get that far.
Set that beside the trade everyone else is making.
Moody’s expects the six biggest American cloud companies to spend over US$785 billion on AI capacity in 2026, and to come close to US$1 trillion the year after. China’s leading technology firms go from roughly US$65 billion in 2025 to between US$140 billion and US$165 billion a year across 2026 and 2027. Moody’s reads their smaller revenue bases and thinner financial flexibility and concludes they will borrow to keep up.1
That is the shape of every seat in this trade below the very top. Borrowed money, spent to stay level with someone whose balance sheet is a country’s.
Mekong has published no doctrine about any of this. The ledger states one anyway, and it is legible enough to copy.
A rider in Ho Chi Minh City who needs cash by Friday against his motorbike papers needs it whether a model in California clears a benchmark that week or fails one.
The question the firm answers before it writes a cheque is whether the demand under a business survives the technology in fashion either working or failing. Call it the indifference test. A rider in Ho Chi Minh City who needs cash by Friday against his motorbike papers needs it whether a model in California clears a benchmark that week or fails one.
The arithmetic sits in the open for anyone who wants to run it against their own numbers. Eleven shops to 972 over ten years is a shade under 57 percent a year, compounded, in a business holding no proprietary technology at all. An operator who cannot name a comparable rate on the thing they are about to build is being asked to accept a worse trade for a more exciting story.
An operator runs the test in one sitting and pays nothing for it. They write down what a customer of the business actually needs, then follow that need into the world where the fashionable technology lands on schedule and into the world where it stalls for three years. A need that survives both is the layer worth owning.
F88’s founder, Phung Anh Tuan, borrowed from a pawnshop two decades ago to raise the money for a cybersecurity company.
The technology business needed the pawnshop. He went and built the pawnshop.
Footnotes
-
The abstention is not total, and the counter-case belongs on the page. Mekong is raising a climate and regenerative-agriculture vehicle, and the Green Climate Fund committed US$50 million to it in September 2026. A firm that takes a thematic bet on soil has an appetite for themes, and is declining the price being asked for this particular one. ↩