On-chain investigationEthereumTornado Cash

Tornado Cash is coming back?

In 2022 the United States sanctioned software that nobody owned — code with no company behind it, no office and no off switch. Three years later a court disagreed, and the Treasury backed down. We read seven years of Ethereum records to find out whether the sanction worked, and what came back once it was lifted.

At a glance
Tornado Cash is a crypto mixer — software that breaks the link between the wallet money goes in from and the wallet it comes out to. In August 2022 the U.S. Treasury sanctioned it: code with no owner and no off switch. In March 2025, after a federal appeals court ruled such code is not property the government can block, it reversed the decision. We read every deposit and withdrawal across all 19 Ethereum pools since 2019. The sanction worked: nine in ten deposits stopped within two months, without a line of the code changing. Lifting it worked just as fast. But the mixer that came back is a different one — more money inside it than ever before, and about a third fewer people using it.
293,472
Deposits audited, 2019–2026
+6.2%
Deposits vs pre-sanction peak — a record
−33%
Fewer depositors than before the sanction
273,304 ETH
Sitting in the pools, unwithdrawn

In August 2022 the United States Treasury sanctioned a piece of software that nobody owned.

It had blacklisted a crypto mixer before — Blender.io, three months earlier — but that was a business, with operators who could be served notice and a service that could be switched off. Tornado Cash was different. It was a set of programs running on Ethereum that mixed people's money together so nobody could tell whose was whose, and by design nobody controlled it. No company, no office, no off switch. The code would keep running whether Washington liked it or not.

So the real question was never whether the code could be stopped. It was whether people would stop using it. Sanction a company and the company dies. Sanction software that nobody controls and you are betting on something softer: that the humans around it — the users, the developers, the middlemen who make it usable — will quietly decide the legal risk is not worth it.

For two and a half years, that bet looked like a win. Then a federal appeals court ruled that code with no owner is not "property" the government can block, and in March 2025 the Treasury took Tornado Cash off the list. The software had not changed. The law around it had.

That leaves an unusually clean experiment, and a question worth answering with evidence rather than opinion: does sanctioning software on a blockchain actually work? We read every deposit and withdrawal Tornado Cash has processed on Ethereum since the day it launched — seven years, ending this August — to see what the sanction did, and what came back once it was lifted.

The short answer is that it worked, completely, and then stopped working the moment it was lifted. But what came back is not what left. The money returned, and there is more of it sitting in Tornado Cash today than at any point in its history. The people did not return. Roughly a third of the crowd is simply gone — and for a tool that hides you by putting you in a crowd, that is the part that matters.

01 — The premise

What a mixer actually does

Everything on Ethereum is public. Every payment you have ever made sits in an open ledger that anyone can read, forever, attached to an address that becomes easy to link to you the moment you touch an exchange. For most people that is fine, right up until it isn't — until a landlord, an employer or a stranger who once sold you something can read your balance and your spending like a bank statement left face-up on a table.

A mixer is the workaround. You put money in. Later you take the same amount out to a fresh address, and the ledger shows no connection between the two. The same trick that protects someone who would rather their salary weren't public also protects someone moving the proceeds of a hack. That is the entire policy problem, in one sentence.

Tornado Cash does it with a queue and a secret. You deposit a fixed amount, and everyone deposits the same fixed amounts, so no single deposit stands out from the pile. In return your browser keeps a secret — a note. Later you use that note to prove you are owed a withdrawal without revealing which deposit was yours, and the money goes to whatever address you name. Usually a middleman called a relayer sends that final transaction for a fee, because a brand-new wallet has nothing in it to pay network fees with, and topping it up from your old wallet would give away the very link you just paid to break.

Here is the part that matters, and it is not the cryptography. What actually hides you is other people.

Think of it as a crowd. If a thousand deposits are sitting in the pool unclaimed when you take your money out, an observer knows your withdrawal came from one of a thousand — and has no way to narrow it down. If only nine are sitting there, the same observer has a very short list. Same code, same maths, completely different outcome, and the only thing that changed was how many other people were standing there with you.

Privacy here is not a property of the software. It is a property of the crowd.

And a crowd can be counted. You cannot see who is in it, but every deposit and every withdrawal is public, so you can measure exactly how big it is.

Figure 1One cycle through the mixer
Funding walletholds the moneydeposit 100 ETHPool contract2,481 unspent notesproof + nullifierRelayerpays the gas100 ETH outClean walletno prior historysecret note travels off-chain — never recordedthe on-chain link this severs — the reason the tool exists
Money enters in a fixed denomination and leaves in the same fixed denomination, to an address with no connection to the sender. The only thing tying the two together is a secret note held off-chain by the depositor. The pool’s unspent notes are public and countable, which is why the strength of the hiding place can be measured — and why this audit is possible at all.
02 — How to read this

What "coming back" could mean

"Is it back?" sounds like a yes-or-no question. It is not, because there are at least five separate things you could measure, and right now they do not agree with each other.

How often is it used? How much money passes through it? How many people use it? How much is parked inside it? And how much of Ethereum's overall traffic does it account for?

Two of those five are at all-time highs. The other three sit well below where they were before the sanction. Take the first two on their own and Tornado Cash has more than recovered. Take the other three and it plainly has not. Both readings come from the same ledger — which is why what follows uses all five rather than picking a favourite.

Record+6.2%Deposits per year, against the pre-sanction peak
Record+47%Money held in the pools, against the previous high
Below peak62%Value flowing through, as a share of the old peak
Below peak67%People using it, as a share of the old peak
Below peak67%Share of Ethereum's traffic, against the old peak
03 — 2019 to 2022

The rise

Tornado Cash went live in December 2019 and grew with the bull market, the way most things in crypto did. By the spring of 2022 it was handling tens of thousands of deposits every quarter, from tens of thousands of different wallets.

The wallets are the number to hold on to, not the deposits. The crowd was genuinely large, and genuinely spread out: a great many people, most of them passing through only a handful of times and then leaving. That is precisely the shape a privacy tool wants. A lot of strangers, none of them staying long enough to stand out.

Figure 2Seven years of deposits, by quarter
PRE-SANCTIONSANCTIONEDPOST-DELISTING06,00012,00018,00024,0002019 Q40.7M at deposit prices">2020 Q289.0M at deposit prices">2020 Q4,356.6M at deposit prices">2021 Q2,115.9M at deposit prices">,420.8M at deposit prices">2021 Q4,009.1M at deposit prices">,421.3M at deposit prices">2022 Q22022 Q423.0M at deposit prices">69.1M at deposit prices">2023 Q234.9M at deposit prices">87.9M at deposit prices">2023 Q42024 Q22024 Q462.2M at deposit prices">2025 Q2,114.6M at deposit prices">2025 Q42026 Q28 Aug 2022 — designated21 Mar 2025 — delistedDeposits per quarter, all 19 pools. Final bar is partial (1 Jul – 12 Aug 2026).
Every deposit into all 19 Tornado Cash pools, quarter by quarter. The designation on 8 August 2022 cut activity by about 90% within two months; the delisting on 21 March 2025 is followed by six consecutive quarters at or above the sanctioned-era ceiling. The last bar covers 43 days, not a full quarter. Hover any bar for exact deposit count, ETH and dollar value.
04 — August 2022

The hammer

On 8 August 2022 the Treasury added Tornado Cash to the sanctions list — the same list used for arms dealers and drug cartels, except this entry was software.

The effect was immediate and brutal. Within two months, nine out of every ten deposits had stopped. The number of wallets using it fell by roughly the same amount.

The most revealing part is not in the deposits, though. It is in the relayers: the paid middlemen who send withdrawals on other people's behalf, and who — unlike the software — have names, addresses, and something to lose. Three of the largest processed their final transaction within days of the announcement. One stopped on the day itself. Seven of the ten biggest never came back at all.

The sanction never touched the code. The contracts ran throughout, exactly as before, indifferent to the whole thing. What the sanction removed was the people willing to stand next to them — and the crowd left with them.

Nine in ten deposits stopped within two months. Not one line of the code had changed.August to October 2022
05 — 2022 to 2025

The quiet years

For the next two and a half years Tornado Cash ran at about a quarter of its old pace, largely ignored.

Parts of it simply died and never recovered. The pool that handled Bitcoin took its last genuine deposit a month before the sanction even landed. The dollar-stablecoin pools went two full years without a single deposit between them — not a slow decline, just nothing at all.

The lowest point came right at the end, in the first months of 2025. Then, on 21 March, the Treasury took it off the list.

06 — April 2025

The turn

The response came almost immediately. March had been an ordinary month, indistinguishable from every month of the year before it. April — the first full month after the sanction lifted — saw deposits jump by about two thirds, and the number of separate wallets using the mixer more than double.

It kept climbing from there. By the summer of 2026, Tornado Cash was recording its busiest month ever: busier than anything it managed at the height of the bull market, when it was legal everywhere and nobody had heard of an SDN list.

Deposits and wallets are charted separately below. They rise together in April 2025 — and then they stop rising together.

Figure 3The turn, month by month
03,0006,0009,00012,000delisting → 21 Mar 2025Deposits per month05001,0001,5002,0002024-012024-042024-072024-102025-012025-042025-072025-102026-012026-042026-07Unique depositing wallets per monthJan 2024 – 12 Aug 2026. Shared x-axis, separate scales. Final month partial (12 days).
before delistingafter delisting
Two measures on the same timeline: how often the mixer was used (top) and how many distinct wallets used it (bottom). Both step up in April 2025, the first full month after the delisting — deposits +69%, wallets +134%. They are drawn as separate panels because they are counted in different units; putting them on one pair of axes would invent a relationship between them. August 2026 covers 12 days.
07 — The record, qualified

What came back

A record is a record. But three things sit inside this one, and each makes it smaller than it looks.

The first is size. Most of the new deposits are tiny. The mixer only accepts fixed amounts, and the smallest of them — a tenth of an ether, pocket change — has gone from about a sixth of all deposits before the sanction to more than a third after it. In the record month, two thirds of the deposits were that smallest size, and together they carried well under one percent of the money. The count went up partly because the deposits got smaller.

Figure 6What the record is made of
Pre-sanction era150,226 deposits16.3%83.7%Post-delisting era84,035 deposits36.2%63.8%July 2026 (record month)11,419 deposits64.5%35.5%In July 2026 the 0.1 ETH pool carried 65% of deposits and 0.6% of the ETH (737 of 117,619).
0.1 ETH pool1, 10 and 100 ETH pools
Share of all deposits that went into the smallest pool, 0.1 ETH. It has more than doubled since the sanction and dominates the record month outright — while contributing almost none of the value. Deposit count rose partly because the unit got smaller: average deposit size fell from 23.2 ETH to 13.5.

The second is Ethereum itself. The network roughly doubled in size over the same stretch. Measured against the traffic around it, Tornado Cash is running at about two thirds of the share it once held. It grew. Ethereum grew faster.

Figure 4Intensity, once you divide by the chain
0601201802402019 Q42020 Q22020 Q42021 Q22021 Q42022 Q22022 Q42023 Q22023 Q42024 Q22024 Q42025 Q22025 Q42026 Q2designateddelistedpeak intensity 225.8 (Q2 2022)Deposits per million Ethereum transactions. Final bar partial (43 days).
Ethereum roughly doubled in size between 2022 and 2026, so raw counts flatter the present. Measured against the chain it runs on, the mixer’s best post-delisting quarter reaches 151.9 deposits per million transactions against a 2022 peak of 225.8 — about two-thirds. This is the chart on which the record disappears.

The third is who is actually doing the depositing. That needs a section of its own.

Deposits climbed steadily, month after month. The money did not. It arrived in lumps, with two months carrying most of it. Usage looks like a trend. Value looks like a series of events.

Half-yearDeposits/dayETH/dayWallets/month
2024 H150.71,459346
2024 H243.41,306276
2025 H1 delisting76.91,418473
2025 H2159.22,464902
2026 H1162.11,845830
2026 Jul – 12 Aug partial357.93,3741,884 (Jul)
08 — The balance

Where the money sits

There is more money inside Tornado Cash today than at any point in its history, comfortably more than before the sanction. And it is not passing through. It is parked.

You can see that in the way the balance moves. Money arrives in enormous blocks and leaves in enormous blocks, a quarter of the year at a time, rather than flowing steadily through. For a large share of what is in there, the mixer is behaving less like a tumbler and more like a car park with a privacy guarantee attached.

There is a wrinkle in what is waiting to be claimed, too. Most of the unclaimed deposits are the small ones. The pool holding almost none of the money holds most of the tickets.

Figure 5The anonymity set: money sitting in the pools
085,000170,000255,000340,000pre-sanction peak 215,261 ETH (Q3 2021)all-time high 316,375 ETH — Q4 20252019 Q42020 Q22020 Q42021 Q22021 Q42022 Q22022 Q42023 Q22023 Q42024 Q22024 Q42025 Q22025 Q42026 Q2Cumulative ETH deposited minus ETH withdrawn, all four ETH pools.
Everything deposited and not yet claimed, quarter by quarter. This is the pile a withdrawal hides inside, so a higher line means — arithmetically — better cover. It peaked at 316,375 ETH in Q4 2025 and stands at 273,304 ETH, 27% above anything reached before the sanction. The swing from +156,594 ETH in Q4 2025 to −99,293 ETH in Q1 2026 is money parking, not passing through.
09 — Concentration

Who is actually using it

Here is the part that changes what the record means.

Before the sanction, the ten biggest depositors accounted for a small slice of the money going in. Today those same top ten move roughly one ether in every eight. Widen the lens to a hundred wallets and they account for more than four tenths of everything deposited. The crowd shrank by a third, and what replaced it is far heavier.

Figure 7Concentration doubled
0%10%20%30%40%50%6.5%13.1%Top 10 wallets16.4%31.3%Top 50 wallets24.1%42.2%Top 100 walletsShare of all ETH deposited, by the largest depositing wallets in each era.
pre-sanctionpost-delisting
How much of the money comes from the biggest wallets. Every band roughly doubled after the delisting: the hundred largest depositors now supply 42.2% of all ETH entering the pools, against 24.1% before the sanction. A mixer works by mixing many parties’ funds; these are the same parties returning.

The middlemen tell the same story. There are around half as many relayers as before the sanction, and the twenty largest now handle four withdrawals out of every five. Fewer people are relaying their own transactions, which means more of them are routing through a small group of intermediaries who can see both ends of the trade.

EraDistinct relayersTop-5 shareTop-20 share
Pre-sanction3,11429.4%60.3%
Sanctioned1,95734.0%66.1%
Post-delisting1,61940.8%81.4%

One of those relayers has been running continuously since a month before the sanction landed, straight through it, without a gap. Others use custom-generated addresses that are marginally cheaper to transact with at scale — the sort of optimisation nobody bothers with unless they are doing this professionally, in volume. This is infrastructure, not a crowd of volunteers.

Ten wallets now move one ETH in every eight. The top twenty relayers handle four withdrawals in five.Post-delisting era, 21 Mar 2025 – 12 Aug 2026

The night of 5 November

Just after midnight one night last November, sixteen wallets that had never done anything before started doing the same thing at the same time. Each had been handed its money by exactly one other wallet, in exactly one transaction, with nothing left behind. All of it inside a three-hour window. The wallets that funded them had themselves been funded the same way sixteen days earlier — long enough for a trail to go cold, short enough to be deliberate.

Then look at the amounts. Three of the sixteen deposited exactly the same round number. Two more matched each other. Independent people do not arrive at identical round totals by coincidence; numbers that round are numbers somebody chose. By the end of the day it was the biggest single day Tornado Cash had ever seen, and about a third of everything deposited in those three months.

It happened twice more in 2026, in the same shape. We are not naming anyone here. Identifying who is behind these flows means following each chain back through single-use wallets until it reaches an address we can put a name to — a separate piece of work, and one worth doing properly rather than in a footnote. What this audit can say honestly is what the shape says. But the structure is not ambiguous. The largest deposits arriving at a tool built for many small independent users are coming from a handful of coordinated ones running scripts.

Figure 8The 5 November fan-out
19–20 Oct 2025origin fundingstaging walletsheld 16 days, untouched5 Nov 2025, 00:25–03:21 UTC16 depositors, one transfer eachthe pools1,109 deposits106,940ETH in 24hlargest day everThree wallets deposited exactly 5,200 ETH; two exactly 5,100 —a fixed total being divided, not sixteen independent decisions.16-day dwell
The structure of the largest day in the protocol’s history: sixteen chains, each one wallet funding exactly one wallet in exactly one transaction, staged sixteen days in advance and executed inside a three-hour window. No entity attribution was performed and none is implied — this describes shape and timing only.
10 — The tell

Nine

One last pattern, and it is the strangest thing in the data.

Count how many times each wallet has used the smallest pool since the delisting and you get the smooth curve you would expect from a real population: most people once, fewer twice, fewer still three times, tailing away. Except at exactly nine, where hundreds of wallets stop dead. At ten, it falls off a cliff.

Figure 9A spike at exactly nine
08001,6002,4003,200123456789101112Wallets by number of 0.1 ETH deposits, 21 Mar 2025 – 12 Aug 2026. x-axis: deposits per wallet.
Counting how many times each wallet used the smallest pool produces a smooth decay — until nine, where 526 wallets pile up above the trend, and ten, where the count collapses by 94%. Wallets stopping at nine finish in an average of 1.2 days; wallets going past ten spread theirs over 18 or more. The cause is undetermined; the pattern is a repeating signature in a system whose entire purpose is to remove them.

Those wallets behave differently, too. The ones that stop at nine do all nine in a single sitting, usually within a day. The ones that carry on past ten spread theirs over weeks.

We do not know why. It could be a tool or interface that caps a session at nine. It could be people funding a wallet with a single ether, depositing nine tenths of it and holding back the rest for fees. Both explanations fit the evidence equally well, and the data cannot separate them.

What matters is the effect. Hundreds of users have stamped the same recognisable signature on their own behaviour — and a signature is exactly what a mixer exists to remove. They partitioned themselves out of the crowd they were paying to join.

11 — What it means

Does sanctioning software on blockchains work?

On this evidence, yes. Nine in ten deposits stopped within two months, without a single line of the code changing, because the sanction never needed to reach the code. It reached the relayers, the websites, and the ordinary user's appetite for legal risk. Two and a half quiet years followed.

It also stops working the moment it is lifted, and just as quickly. The recovery dates to a single month — April 2025, the first full month under the new rules — and it shows up in the number of deposits and the number of users at the same time.

But the thing that came back is not the thing that left. The old Tornado Cash was a large, dispersed crowd of occasional users. Today's is a much smaller group that uses it far more often, with a handful of operators moving a large share of the money, four fifths of all withdrawals passing through twenty intermediaries, and single days on which one script accounts for a third of a quarter's volume. The ordinary users left and mostly did not come back. The professionals returned, with better tooling.

Which leaves the mixer's own promise in an awkward place. There has never been more money inside it, and never more unclaimed deposits waiting in the queue, so on paper the crowd has never been larger. But a crowd in which ten addresses supply an eighth of the value, most of the tickets are pocket change, hundreds of wallets share an identical fingerprint, and every exit runs through the same twenty doors is not the crowd the arithmetic assumes.

Anonymity was never a property of the pool. It is a property of the people standing in it — and there are far fewer of them than there used to be.

12 — Reference

The full series

Every quarter of the seven years, so the charts above can be checked against their own numbers. "Held" is the running total of everything deposited and not yet withdrawn.

QuarterDepositsETH inValue inETH outHeld
2019 Q45292,465$0.3M1,3871,078
2020 Q13,06323,371$4.4M19,6084,841
2020 Q23,73852,464$10.7M47,8229,483
2020 Q34,582115,161$40.8M100,21024,434
2020 Q414,071384,959$189.0M335,70473,689
2021 Q118,905515,150$804.4M439,662149,177
2021 Q222,109536,289$1,356.6M482,991202,475
2021 Q317,955388,226$1,115.9M375,440215,261
2021 Q414,662348,766$1,420.8M413,108150,919
2022 Q118,066345,663$1,009.1M336,312160,270
2022 Q222,498609,614$1,421.3M556,261213,623
2022 Q312,245214,651$302.0M340,31287,962
2022 Q43,00463,200$84.5M50,300100,862
2023 Q13,70877,211$123.0M64,314113,759
2023 Q24,13690,939$169.1M78,601126,097
2023 Q34,53176,733$134.9M82,937119,893
2023 Q44,20797,158$187.9M88,884128,167
2024 Q14,169129,703$403.8M131,418126,452
2024 Q25,057135,834$459.7M116,034146,252
2024 Q34,157151,343$389.5M139,747157,848
2024 Q43,82789,037$272.5M97,657149,228
2025 Q14,39861,573$162.2M64,753146,048
2025 Q29,520195,035$448.5M187,267153,816
2025 Q311,805130,140$523.3M124,175159,781
2025 Q417,488323,238$1,114.6M166,644316,375
2026 Q116,082159,141$392.4M258,434217,082
2026 Q213,260174,869$349.7M159,927232,024
2026 Q3 (43d)15,391145,099$269.2M103,821273,302

Four ETH pools. The final quarter covers 1 July – 12 August 2026 only. Value is priced at daily rates on the deposit date. The running total ends two ETH below the 273,304 quoted elsewhere: that is rounding at each quarter boundary, not a discrepancy in the underlying data.

How this was measured

  • All 19 Ethereum Tornado Cash pools, 16 December 2019 to 12 August 2026, from full-archive transaction and transfer data.
  • Pool membership was verified rather than assumed: each contract confirmed by exact-denomination match. The 100 ETH pool resolves to 47,868 deposits summing to precisely 4,786,800 ETH.
  • Every figure comes from a single frozen snapshot ending on the last complete day, so the numbers are reproducible rather than drifting with the live chain.
  • Deposit and withdrawal totals reconcile against the archive's independently-built balance ledger to 0.05 ETH on 5.64 million — nine parts per billion.
  • Prices come from on-chain DEX trades matched on token contract address, never on ticker symbol. Symbol matching pulls in counterfeit tokens named "USDC" and "USDT", and produced ETH prices in the trillions before it was caught.
  • Hundreds of spoofed tokens have been sent into the pool contracts, including a forged DAI and homoglyph imitations of ETH. All excluded.
  • Withdrawals are counted by distinct transaction hash, never by transfer leg — each withdrawal emits a second leg for the relayer fee.
  • No entity attribution. This audit measures flows, not identities: tracing these chains back to named exchanges or services is a separate investigation. Large coordinated flows are described structurally — timing, topology, amounts — and name nobody.
AssetDepositsNative totalValue at depositStatus
ETH277,1635,637,032 ETH$13.16BActive, record usage
DAI9,454459,130,800$459MBest year since 2022
WBTC2,39813,077$533MEffectively dead
USDT2,8472,021,310$2.0MRevived after delisting
USDC1,286940,200$0.9MRevived after delisting
cDAI324619,425,000$13.3MDead since Oct 2021

The 100,000 DAI pool took $20.2M in the first six weeks of Q3 2026; 2026 to date, at $39.9M, nearly matches 2023, 2024 and 2025 combined at $46.3M. About 99% of post-delisting deposits route through the Tornado Router at 0xd90e2f925DA726b50C4Ed8D0Fb90Ad053324F31b, so the front-end infrastructure is alive and maintained.

Bitquery Investigations · Data: Bitquery full-archive Ethereum dataset · Snapshot frozen 12 Aug 2026 · Figures at deposit-time prices · No entity attribution