TokenPost.ai
Decentralized finance (DeFi) was built to remove intermediaries, but new research argues it has instead created a fresh layer of middlemen—one that is increasingly concentrated. A report from CEPR and a University of Turin-linked research group warns that the market structure emerging around Ethereum (ETH) and on-chain trading is pulling DeFi back toward ‘recentralization’, undermining one of the sector’s core narratives.
The paper, titled Can Blockchain Decentralize Money, Contracts, and Finance? and published in February 2026 by a team including Prof. Bruno Biais, extends earlier concerns about stablecoins’ tendency toward centralized control. Its sharper critique, however, focuses on DeFi: the study argues that while the technology removes familiar gatekeepers, it recreates familiar frictions—information asymmetry and order manipulation—under new names and with crypto-native tooling.
Old market problems, new labels
The report highlights two recurring patterns.
First is the transfer of value from slower participants to faster ones—a dynamic the authors frame as a classic information advantage problem. Automated market makers (AMMs), popularized by Uniswap’s 2018 design, use algorithms to quote prices and execute swaps against pooled liquidity supplied by users known as ‘liquidity providers’. While AMMs eliminate the need for a traditional dealer, the report argues liquidity providers still bear losses when better-informed or faster traders trade against them.
In crypto markets this effect is widely discussed as ‘impermanent loss’, a term that suggests a peculiarly blockchain-era tradeoff. The report contends the economic substance is older: liquidity providers are effectively underwriting trading opportunities for informed counterparties, and the costs resemble traditional losses driven by information gaps rather than an entirely novel phenomenon.
Second is transaction reordering—more openly adversarial in nature. Because many blockchains expose pending transactions in a public queue (the mempool) before they are finalized, bots can detect a large trade and insert their own transactions around it. In a typical ‘sandwich attack’, a bot buys ahead of a victim’s trade, pushes the execution price against them, then sells immediately after—capturing a spread while the victim receives worse pricing.