How AI Helped the IMF Track Two Trillion in Cross-Border Stablecoin Transfers

In a bold new paper from the International Monetary Fund’s Research Department, Marco Reuter introduces a pioneering methodology for tracking international stablecoin flows, an area long seen as opaque and unquantifiable. Supported by research dialogue from leading institutions like the Bank for International Settlements (BIS), Financial Stability Board (FSB), European Union (EU), Financial Action Task Force (FATF), and the U.S. Treasury, the paper represents a significant leap in global crypto surveillance. As stablecoins like USDT and USDC become essential tools in cross-border finance, capital flight, and even unofficial dollarization, regulators and economists alike have been desperate for a way to illuminate the flows behind this vast digital financial machinery. Reuter’s approach answers that call, combining machine learning, blockchain data, and artificial intelligence to create the first robust regional flow estimates of stablecoin activity.

Cracking Wallet Geography with Machine Learning

One of the study’s core achievements is its success in solving the problem of wallet pseudonymity. While crypto transactions are publicly visible on blockchains, wallets themselves are pseudonymous, identified only by cryptic hexadecimal strings. Reuter’s team found a workaround by drawing on Ethereum Name System (ENS) domains, human-readable names like “pijiu.eth” and using a large language model to identify cultural or linguistic markers. This process revealed, for instance, that Arabic-script names were likely tied to the Middle East, while Chinese terms pointed to Asia. In parallel, wallets interacting heavily with regional crypto exchanges, like Indodax in Indonesia or Bitso in Mexico, were classified by geography.

These two datasets provided a labeled training pool of over 346,000 wallets. Using this data, the team trained a gradient-boosted decision tree model with features like transaction time-of-day, daylight saving adherence, favored exchanges, and ERC-20 token interactions. The model achieved a prediction accuracy of 65%, a significant improvement over the 20% expected from random guessing. It was then used to classify over 20 million active self-custodial wallets that had conducted stablecoin transactions in 2024, producing a never-before-seen map of global crypto flows.

Trillions in Motion: Mapping the Global Flow of Stablecoins

The results are both staggering and revealing. In 2024 alone, Reuter estimates 138 million stablecoin transactions totaling $2.019 trillion. Asia and the Pacific topped the chart with $407 billion in inflows and $395 billion in outflows, followed closely by North America, which logged $363 billion in inflows and $417 billion in outflows. But when viewed relative to GDP, two regions stood out: Latin America and the Caribbean (7.7%) and Africa and the Middle East (6.7%) saw the highest levels of stablecoin activity, suggesting a growing reliance on digital dollars in economies facing inflation, capital controls, or remittance dependence.

Intriguingly, stablecoin usage is largely international. Intra-regional flows accounted for just 12%–14% of transactions in Latin America and Africa, compared to 34% in North America. This suggests that in developing regions, stablecoins function more as cross-border financial tools than domestic currency substitutes. Transaction sizes reflect income levels: North America averaged $35,016 per transaction, while Asia and the Pacific lagged at $11,493. USDT, the dominant stablecoin in emerging markets, was more widely used than USDC, which retained favor in advanced economies like the U.S. and Europe.

The U.S. as Global Supplier of Digital Dollars

A central finding of the study is that North America is the primary exporter of stablecoins. In 2024, it supplied $54 billion in net outflows to the rest of the world. These flows were not static; they increased during periods when the U.S. dollar strengthened relative to other currencies, confirming that stablecoins act as a shadow mechanism for global dollar liquidity. A deeper country-level analysis of China underscores this role. Using a specially trained secondary model with 79% accuracy, the paper estimates that China alone received $18.6 billion in net stablecoin inflows in 2024, 100 times more than the $0.18 billion estimated by the commercial Chainalysis dataset.

The discrepancy is attributed to methodological weaknesses in Chainalysis, which assumes users do not use VPNs and that transaction sizes are the same across regions. In China, where crypto trading is banned and VPN use is widespread, these assumptions fail dramatically. The IMF study shows that Binance, inaccessible in China without a VPN, remains the top platform used by Chinese wallets, reinforcing the superiority of Reuter’s method for capturing crypto flows in restrictive environments.

Crypto, Crisis, and Capital Controls: What the Future Holds

The study also revealed that stablecoin flows are vulnerable to systemic shocks in the traditional banking world. During the March 2023 U.S. banking crisis, which saw the collapse of crypto-friendly banks like Silicon Valley Bank and Signature Bank, stablecoin flows from North America plummeted by nearly 10 standard deviations. This sudden drop was captured using a difference-in-differences analysis and illustrates just how dependent digital assets still are on fiat infrastructure for issuing and redeeming stablecoins.

The paper is more than a research paper; it is a blueprint for regulatory oversight, economic analysis, and digital finance policymaking in the age of blockchain. By breaking through the barrier of pseudonymity, Reuter has laid the groundwork for understanding not just where stablecoins flow, but why they matter. The methodology offers a powerful lens for future research, including applications in country-level analysis, crypto asset stock distribution, and even decentralized finance (DeFi) usage patterns. As crypto assets continue to expand their role in global finance, this study ensures policymakers now have the tools to keep up.