Who’s really trading stablecoins: the true depth of DEX markets

These bots execute anywhere from thousands to millions of trades per month, operate almost every day, and typically trade across around ten different venues at once. Individual trades are usually worth only a few hundred dollars.

How the JaredFromSubway MEV Bot Lost $7.5 million without being hacked

A sophisticated attacker tricked Ethereum's infamous JaredFromSubway MEV bot into approving malicious smart contracts, ultimately draining roughly $7.5 million. GetBlock AML Research breaks down one of the most unusual DeFi exploits of the year.

Читать дальше

Their job is to capture small price differences between exchanges while also helping keep asset prices aligned across markets.

The second category is liquidity aggregators.

These include services such as Jupiter on Solana, and 1inch, 0x, CoW, and KyberSwap across EVM networks. When a user submits a swap order, the aggregator automatically splits it into multiple smaller transactions and routes them across different liquidity pools to secure the best possible execution price.

As a result, a single user trade can turn into several separate blockchain transactions. Both arbitrage bots and aggregators have been part of the DEX market for years.

The third category is proprietary automated market makers (AMMs).

This has become the most important new trend in recent months. The best-known examples are HumidiFi, SolFi, and BisonFi. These venues trade exclusively with their own capital and do not have a public-facing user interface. Instead, they publish quotes directly on-chain and interact only with aggregators.

A user never chooses such a venue directly — their order is routed there automatically if the aggregator determines it offers the best price. The concept itself is not entirely new. Lifinity operated on a similar model back in 2022.

What is new is the current wave of such systems, which now process a meaningful share of trading on Solana. Much of this has been enabled by Solana’s architecture, where computing costs are so low that automated systems can update quotes several times per second.

Machine behavior is hard to confuse with human behavior

Once you look closely at wallet activity on-chain, the differences become obvious. Automated trading systems operate almost nonstop — typically 26 to 28 days per month — and use anywhere from six to fourteen trading venues at the same time.

Tether has frozen over $5 Billion in USDT: here’s how

Most frozen USDT is held on the TRON network, yet Ethereum users are far more likely to recover their funds. Meanwhile, the fate of more than $3 billion remains officially unexplained.

Читать дальше

Retail users behave very differently. Most of them appear on the market only two or three days per month and use just one or two exchanges. That is why it is often possible to distinguish a human-controlled address from an automated one even without knowing who owns the wallet.

Addresses that display machine-like behavior make up only a tiny fraction of all wallets, yet they account for the overwhelming majority of total trading volume.

By contrast, addresses that behave like regular users represent the vast majority of wallets but generate less than one-tenth of overall trading volume.

The biggest speculators are not people at all

Another important finding from the study concerns the nature of the trades themselves.

The more automated an address is, the less likely it is to swap one stablecoin for another — and the more likely it is to trade volatile crypto assets.

For example, the most active bots on EVM networks spend about 90% of their activity trading volatile digital assets. Retail users, by comparison, do this in only about 30% of cases. Much more often, they are simply swapping one digital dollar for another.

This means that much of what looks like speculative trading on the surface is actually being carried out by automated algorithms engaged in arbitrage or liquidity provision.

In other words, most of these trades are not made by people trying to profit from price swings. They are executed by programs designed to keep prices aligned across different venues.

The result is a striking paradox: regular users are the ones most often using stablecoins for their original purpose — moving and exchanging digital dollars.

Which Stablecoins Automated Systems Prefer

The mix of market participants varies significantly depending on the stablecoin in question. This is one of the study’s more interesting findings because it helps separate assets with genuine user-driven demand from those whose trading activity is largely created by automated systems.

As expected, the largest stablecoins are heavily dominated by machines. Around 70% of USDC trading volume comes from automated addresses. For USDT, the figure is roughly 59%. The picture looks very different, however, for some newer digital dollars.

One example is USD1, issued by World Liberty. It turned out to be the stablecoin most dependent on automated trading: about 82% of its total trading volume is generated by bots. That matters because high trading volume alone does not necessarily mean a token is widely used by real people.

In the case of USD1, a large share of the activity appears to come not from organic demand, but from trading algorithms. A similar pattern can be seen in EURC, Circle’s euro-denominated stablecoin, where machines account for about 80% of trading volume. The explanation there is somewhat different.

In EURC’s case, the skew is driven less by retail adoption of a “digital euro” and more by automated arbitrage between euro- and dollar-based assets.

Which stablecoins are used more often by real users

On the other end of the spectrum are assets where human participation is meaningfully higher.

How cryptocurrency transactions are tracked in 2026: a real investigation case

A cryptocurrency transaction became the key lead in an investigation that resulted in the arrest of a daycare worker. Here’s how law enforcement agencies are working with digital assets in 2026.

Читать дальше

The clearest examples are RLUSD, issued by Ripple, and the decentralized stablecoin DAI. Roughly one-third of trading volume in these assets comes from ordinary user wallets, while automated systems account for only about 17%.

USDe stands apart as a separate case. Its participant profile differs from both of the groups above. Activity here is driven less by casual users or simple trading bots and more by highly active market participants.

That group includes yield farmers, who place capital into automated financial protocols to earn returns, as well as professional traders who regularly open and manage positions.

That is why, when a new stablecoin appears with impressive trading volume, the key question should not be “How large is the volume?” but rather “Who is actually generating it?”

Blockchain data is what makes that question answerable.

Machines have always been the main participants

It might be tempting to assume that automated trading systems only started to dominate in recent years. But the data tells a different story. Over the full 18 months covered by the study, the share of machine-driven trading barely changed.

On average, automated systems consistently accounted for about 65% of all stablecoin trading volume. Deviations were small. The only meaningful dip came in June 2025, when a fresh wave of retail investors entered the market and rising prices sharply increased activity from ordinary users.

During that period, the share of automated trading temporarily fell to around 43%. But once that retail burst faded, the market quickly returned to its previous structure.

That means machine dominance is not a new trend. It has been a structural feature of the market for quite some time.

The implication is important: if roughly two-thirds of all stablecoin trading has been generated by automated systems for a year and a half regardless of market conditions, then rising trading volume does not necessarily mean the user base is growing.

In many cases, it simply means there are more trading algorithms in the market — or that existing ones are moving more capital as they execute their own strategies.

What actually changed: the market became smaller, faster, and less visible

While the share of automated trading stayed broadly the same over the past 18 months, the infrastructure behind it changed significantly. The study highlights several major shifts.

Trades got much smaller

The average size of stablecoin trades has dropped sharply.

Across EVM networks, the average trade fell from about $2,940 to $700. On Solana, the decline was similar, from $1,550 to $380. At the same time, the total number of trades barely changed.

In June alone, the market recorded around 300 million stablecoin trades below $1,000. Of those, about 185 million took place on EVM chains and another 113 million on Solana.

This trend is especially visible among liquidity aggregators. For example, the 0x API alone processed more than 2 million trades, with a median trade size of just $10.

In other words, modern automated trading systems increasingly prefer to execute a huge number of very small trades rather than a smaller number of large ones.

The number of trading bots surged

At the same time, the number of highly active automated addresses rose sharply. At the beginning of 2025, EVM networks had roughly 2,700 addresses making more than 2,000 trades per month. That number later climbed to nearly 12,000, before stabilizing at around 7,200.

So there are now many more automated trading systems in the market. Each of them is trading more frequently, while the average trade size has become much smaller.

Trading venues became almost invisible

Another major change affected the exchanges themselves. Traditional automated market makers such as Uniswap, PancakeSwap, and Raydium still account for around 72% of total trading volume. Another roughly 10% comes from specialized stablecoin pools, with Curve being the best-known example.

The fastest-growing category, however, is the proprietary automated market makers mentioned earlier.

These venues now account for around 18% of all trading volume, and nearly all of that activity is concentrated on Solana. They have no public website, no open liquidity pool, and no conventional user interface.

The only way to access them is through an aggregator, which routes a user’s order there automatically if it offers the best execution path. One of the clearest examples is HumidiFi. Just a few months after launch, it was already handling roughly one-third of all DEX trading volume on Solana.