More USDT was issued without a sustained rise in balances held by smart-contract accounts on Ethereum, according to a Bank for International Settlements working paper published Sept. 15, 2026. On Tron, those accounts held about 1% of USDT through most of the study's historical series. Together, the findings challenge the assumption that a larger stablecoin supply automatically means more capital has entered decentralized finance.

The BIS paper tracks where tokens sit, not why every holder owns them. Its holder-balance chart stops before 2026 on its date axis, so its percentages cannot be read as September 2026 measurements. That timing matters as current dashboards continue to show large USDT balances on both networks.

On Ethereum, smart-contract accounts held more than 20% of the network's USDT during part of 2021 and 2022. Their share hovered around 15% to 20% until late 2024, then moved down to roughly 10% to 15% as issuance expanded. The drop is a change in the proportion of tokens in contracts, not a finding that the absolute balance kept falling. The BIS authors say the issuance growth did not bring a sustained increase in contract holdings.

A lower share can result when newly issued tokens accumulate outside contracts even if the amount in contracts stays near its earlier level. The study's Ethereum series reflects that distinction: it shows far more USDT in non-contract accounts as issuance grew, without a comparable sustained increase in smart-contract holdings. The percentage change alone is therefore a poor way to infer a withdrawal of dollars from DeFi.

Figure 10 of the study places Ethereum contract-held USDT at roughly $10 billion to $15 billion toward the end of its plotted period, and Tron's at around $1 billion or less. These are approximate chart values. Ethereum's dollar balance fluctuated in the low tens of billions while the share shrank; on Tron, contract balances remained a small slice of a much larger supply. The two percentages have different chain-specific denominators and cannot be treated as a single measure of DeFi adoption.

The researchers reconstructed USDT holdings from Ethereum and Tron transfer event logs. They identified smart-contract accounts from contract deployments, classified other addresses as externally owned accounts, and cross-checked token supply against mint, burn and blacklist-destruction events. Following the token itself gives a different view from adding up deposits reported by DeFi protocols, where the same tokens may be counted more than once.

Protocol-level total value locked measures assets assigned to particular DeFi applications. The BIS reconstruction instead follows one token across addresses on two networks, including holdings outside those applications. It is better suited to asking how USDT is divided between account types, while protocol TVL can describe the scale of selected venues. Neither turns a balance into a verified description of the holder's purpose.

That distinction improves the measurement, but an account type is still an imperfect guide to economic use. A smart contract may hold USDT for a bridge, wrapper or custodian rather than a DeFi lending or trading strategy. An externally owned address may be used for payments, savings, remittances or exchange custody. Tron's roughly 1% contract-held share therefore does not show that the remaining tokens were spent as payments. Nor does Ethereum's falling share prove that DeFi use contracted.