There is a particular type of meeting that has started happening in Ulaanbaatar this year. American officials, Korean executives, and Mongolian ministers sit around a long table and discuss minerals that none of them mine yet. The buildings get nicer. The translation gets better. The minerals stay in the ground.
The current projection is that Mongolia holds something like two and a half percent of the world's rare earth reserves by 2026 estimates — not a dominant share, but a meaningful one in a market where China currently produces close to ninety percent of refined output. The United States and South Korea agreed in 2025 to fund a Rare Metals Cooperation Center based in Ulaanbaatar. KT Corporation signed a memorandum with Monnis Group, one of the country's largest holding companies, to cooperate on extraction. The Mongolian state-backed body MNREC has stated its goal of producing the country's first mixed rare earth oxides by 2028.
What does not exist yet, in Mongolia, is a single operating rare earth mine.
That is not a defect in the story. It is the story. The international agreements, the cooperation centers, the corporate memoranda — all of these are operating against a country that has the reserves on the geological map and almost none of the industrial substrate that would turn reserves into shipments. The processing facilities are unbuilt. The water rights for processing are uncalculated. The road network from the deposit zones to a port — and Mongolia does not have a port — runs through either Russia or China, both of which have an interest in what Mongolia's third-country strategy looks like.
For builders watching the AI infrastructure conversation from outside the bidding war for North American grid capacity, this is the part of the supply chain that is rarely visible. The conversation in San Francisco is about GPUs and data centers. The conversation that determines whether the GPUs get built happens a layer down, in the magnets and the processed oxides, and that conversation is increasingly held in places like Ulaanbaatar — but it is held about events that will not happen for two to five years.
What does it mean to build software from a place that is, simultaneously, geologically central to the next decade of compute and industrially peripheral to it?
For us, it means three things in our actual operations.
It means latency math. Our nearest hyperscaler region is in either Tokyo or Frankfurt depending on the workload, and the round-trip number is non-trivial. We design for it. We push more state to the edge than a Silicon Valley team would, because the alternative is a one-second cold start on a request that should be quiet. The same constraint that makes Mongolia an interesting place to mine from makes it a particular place to write from.
It means timezones that work in our favor more often than they work against us. American clients sleep while we ship, which means a client who calls Monday morning gets work finished Friday evening their time — work the team was awake to fix on Saturday. This is unglamorous and well-known to anyone who has worked across distance, and it remains underpriced by every procurement team that treats geography as a discount lever rather than a delivery model.
It means an unusual angle of vision on what is happening to the supply chain. When we tell US clients that the rare earths in their next batch of magnets will probably touch Mongolia somewhere between the year we are writing this and 2030, they hear a geopolitical observation. From here, it is an operational one. The same country whose names they cannot pronounce on the deposit map is the country whose engineers wrote their inventory system.
We have argued in other posts that frontier deals are not procurement signals and that the grid has the last word on what gets built. The rare earth question sits one layer below both. The grid is fed by the magnets. The magnets are fed by the oxides. The oxides, increasingly, are going to come from places whose first processing facility is still on the drawing board.
The country that has the reserves and no mines is not a defect of the supply chain. It is the supply chain, at the stage it is in now, a short drive from our office.
The short of it.
Mongolia holds an estimated two and a half percent of global rare earth reserves, a new US-South Korea cooperation center in Ulaanbaatar, multiple corporate MoUs with extractors, and zero operating mines. The country is geologically central to the next decade of compute and industrially peripheral to it — a vantage point we use in client work, particularly around supply-chain timing. The first oxides are expected by 2028; the build decisions that depend on them are being made now.