1.6 million SOL in stake accounts nobody comes back for

Contents

What a dormant stake account is

When you stake SOL natively to a validator, the coins don’t sit in your wallet — they move into a separate stake account, its own address on chain. Your wallet only holds the rights to it: the right to manage the delegation (staker) and the right to take the money out (withdrawer).

Exiting native staking takes two actions:

• deactivate the delegation — over one epoch (about two days) the stake stops working;
• withdraw the funds from the stake account back to your wallet — the Withdraw instruction.

A dormant, or zombie, account is one where the first action happened and the second never did. The money sits in a state of “no longer working, not yet collected.” It earns nothing at all.

Our criteria: deactivation happened at least 10 epochs ago (so we don’t count people simply moving stake between validators), and the account holds more than 1 SOL (so we don’t count dust).

This money is not locked. It is available to the owner at any moment. There just isn’t an owner around — or they’ve forgotten it exists.

Scale: many people, little money

As of epoch 1028 (September 4, 2026), sitting dormant:

33,864 accounts. 1,637,284 SOL. 26,553 distinct owners.

The distribution is more interesting than the total:

The median dormant account holds 3.44 SOL. The typical forgotten stake isn’t somebody’s serious capital — it’s a leftover nobody bothered with. Someone tried staking, exited halfway, moved on. Or switched wallets. Or lost a device.

Meanwhile, 18 accounts hold a third of all the dormant money.

A third of dormant accounts were put to sleep by the network, not the owner

This is the most technical part, and to us the most interesting.

In April 2022 the Solana stake program gained an instruction called DeactivateDelinquent (PR #23932, by Michael Vines). It lets anyone, without the owner’s signature, deactivate stake delegated to a validator that hasn’t voted for at least five consecutive epochs.

Sounds odd — someone else’s stake, without permission? — but the reasoning was purely engineering, and the PR states it plainly: there was no mechanism to clean up stake delegated to abandoned vote accounts. The author’s example was devnet, where 9% of stake had been abandoned for close to a hundred epochs, and the only way to clear it was a network restart.

Why it matters. Solana’s leader schedule is built from active stake weight. A validator that stopped voting long ago but still carries a large delegation keeps receiving leader slots — and those slots get skipped, no blocks produced in them. Dead stake literally takes throughput away from the network. DeactivateDelinquent removes it from the calculation, and the slots go to operators who actually run.

There’s no reward for calling the instruction — it was designed as a common good, a matter of network hygiene.

Here’s what we found in the data. We took a random sample of 120 dormant accounts and checked which instruction had deactivated each one:

• deactivate (the owner did it themselves) — 70%
• deactivateDelinquent (someone else did it) — 30%

Extrapolated across the set, that’s roughly 10,000 accounts (confidence interval 7,400–12,900).

The reason is visible in the raw data too, without any sampling. Of the 33,864 dormant accounts:

That is 7,798 accounts (23%) sitting on validators that stopped operating, affecting 5,305 distinct owners. Within that group specifically, our sample puts the share of third-party deactivation at 80%.

The cleanup is done by bots, and you can see them in the transactions: the most active one signs as GiYGtz42bh1sLKvCxw3PNzvS6FAwB4ciNejLi5CpHPgr and processes about twenty accounts per transaction. There’s a second one with a fitting address: ByeByeS4EhEhAPmqE2nULzwzx9yK1Ee47We3TCQ5Bwys.

So a third of these dormant accounts aren’t somebody forgetting step one. They’re a network hygiene mechanism working exactly as designed. The network stopped spending resources on dead stake.

One thing worth keeping in mind: a bot can remove someone else’s delegation, but it cannot take the money — Withdraw requires the owner’s signature. The network does precisely its half of the job. The other half is left to whoever owns the stake.

The largest private collection of dormant accounts

The wallet HXdYQ5gixrY2H6Y9gqsD8kPM2JQKSaRiohDQtLbZkRWE holds the record: 1,200 dormant accounts in our list. At first it looks like chaos — 2.3 SOL apiece. But look at all its stake accounts and a deliberate strategy appears: maximum diversification, both of yield and of single-validator failure risk.

In total it holds 1,572 stake accounts worth 3,650 SOL — and exactly one account per each of 1,571 validators, about 2.25 SOL each. The accounts were created between October 2021 and June 2023.

Then validators started shutting down, and the mechanism described above kicked in. We checked individual accounts: they were deactivated by that same GiYGtz42… bot via deactivateDelinquent. The most recent cleanup ran on September 5, 2026 — after we took our second network snapshot.

Where it stands today:

• 142 accounts still active (328 SOL) — validators alive, stake working;
• 1,429 accounts deactivated (2,763 SOL) — mostly by other people’s hands, as validators died off;
• 1 account holding 557.81 SOL was created and never delegated at all — it has just been sitting there since 2023.

One more detail visible in the transactions. While an account is active, MEV rewards trickle in every epoch via the Jito Tip Router — literally fractions of a thousandth of a SOL at a time. Trivial individually, but across fifteen hundred accounts and several years it adds up. We measured it: above the delegated principal and the rent-exempt reserve, these stake accounts hold 30.9 SOL of accumulated rewards, and 26.7 SOL of that sits in accounts that have already gone dormant. That’s money withdrawable right now, without unstaking anything. Once an account is deactivated, the trickle stops.

What the strategy was — even coverage of the whole validator set for decentralization, farming validator airdrops, or monitoring infrastructure — we don’t know. But someone put real work into building it, and two thirds of the construction now stands dead.

The account that woke up after three years

On April 26, 2023, someone removed the delegation from a stake account whose address begins with 8k4MoNEY. It held 563,253 SOL.

Then nothing happened. A month, a year, two years. We recorded it in our July 10, 2026 network snapshot as the largest dormant account on Solana — while preparing our Solana Staking Report: development vectors and analytics from epoch 1000.

Then on July 20, 2026, at 06:17 UTC, a single transaction took all 563,253 SOL out.

Three years and three months — and ten days after we counted it. We checked the history: nothing had been done with that stake the entire time. The only activity was occasional token spam landing on a “rich” address, as tends to happen with large wallets.

What it was — a recovered seed phrase, a settled inheritance, unlocked corporate access — we don’t know. But the story captures the point: this isn’t lost money. It’s money nobody is picking up.

A few more findings

The oldest dormant account is 5.5 years old. Account 9S1gnNB1yQy3j1uTqrLVX6UqzQYKpHbdBtwHb4vjUfYd was deactivated at epoch 16 — April 2020, the first weeks of Solana mainnet. It still holds exactly 10 SOL.

We dug up whose it is. The stake sat on a validator called LKValidator, and this account’s withdrawer key matches the node’s own identity — meaning the operator staked 10 SOL on their own validator. Their on-chain profile still carries the keybase handle tdssollk: “tds” stands for Tour de SOL, the very first testnet incentive program and the ancestor of the one discussed below. That node cast its last vote on April 24, 2020 and never came back online. The ten SOL have been sitting there ever since.

Dormant accounts pile up where staking is easiest:

This isn’t a complaint about them — rather the opposite. Ledger and Phantom offer staking in two clicks straight from the wallet, which is where mainstream users arrive. Everstake has long worked actively with small delegators across most proof-of-stake networks and has a broad base of smaller clients. The lower the barrier to entry, the more people wander in casually and then forget.

The cost of forgetting is around 113,500 SOL. That’s the rewards these accounts didn’t earn while idle (simple math: principal × 5.5% annual × years idle). Almost 7% on top of what’s sitting there. The record for a single account is 12,678 SOL of forgone yield.

19 dormant accounts genuinely cannot be withdrawn. They still carry a lockup — a time lock set by whoever issued the funds. The nearest one unlocked on September 7, 2026; the furthest waits until January 2028. This is the only part of the set where the owner’s silence means nothing: they physically have no access.

To be clear and not mislead: these are the longest lockups among dormant accounts, not in the network overall. There is considerably more live, working stake under an active lockup — 698 accounts, including two of 600,000 SOL each unlocking in February and March 2028. Funds like that simply don’t tend to get forgotten.

What changed in two months

We took two snapshots: July 10 and September 4, 2026, 28 epochs apart, and compared the fate of every account.

By volume, the set shrank by half — 49.6% of the money was withdrawn.

By account count, it barely moved — only 7.6% saw any activity. For 92.4%, nothing changed at all.

One figure resolves the contradiction: 95.7% of all the withdrawn money came from a hundred accounts, and a third of it from one (that same 8k4MoNEY). “Half of it woke up” means a handful of very large owners finally got back to their wallets. The mass of small forgotten accounts sits exactly where it sat.

From this comes a simple pattern. The odds of someone coming back for the money depend on two things:

• on the amount: among accounts holding 1–10 SOL, 6% saw movement in two months; among accounts over 500,000 SOL, 100%;
• on how long it’s been: among accounts dormant 10–50 epochs, money was withdrawn in 28% of cases; among those past 600 epochs, 0.7%.

The smaller the amount and the longer it sits, the likelier it is never collected. After roughly a year of dormancy the odds fall close to zero.

A trail from 2022: 269 testnet accounts

A separate story, one every long-time validator remembers.

From 2022 through March 2024 the Solana Foundation rewarded validators for running testnet — the Tour de Sun 22 program, TdS22 for short. It paid operators for running a testnet node on approved hardware in one of a set of quality data centers; the program existed to encourage geographic and hardware diversification of the network.

The payout worked like this. The Foundation created a separate stake account, put the payment into it, made the validator its sole authority, and applied a 12-month lockup. The validator would typically delegate those funds to their own node. The program closed on March 31, 2024, with the final payment issued on April 8.

Here is the full content of one of the earliest such transactions, from November 17, 2022:

Then a neat technical detail came into play. These accounts were frequently split into smaller pieces using the Split instruction. A child account inherits the lockup together with the custodian address — the key entitled to lift that lockup. That turned out to be a perfect marker: however many times an account was divided, the custodian stays the same. We followed the marker and reassembled the whole family.

Result: 269 dormant accounts are TdS22 payouts. Together 58,342 SOL, across 77 owners.

It’s worth showing where we drew the line. The final TdS22 payment went out on April 8, 2024, and the lockup was 12 months — so no account from that program can unlock later than April 8, 2025. Four accounts in our initial set failed that test: their lockups run into 2025 and 2026. They may well have been issued by the same Foundation infrastructure, but they cannot be attributed to TdS22, so we excluded them — along with their 15,093 SOL.

The proportion is telling: 0.79% of dormant accounts by count, but 3.56% by money. Testnet accounts run noticeably larger than average — a median of 70.8 SOL against 3.44 SOL across the whole set.

And an observation that surprised us: they barely move. Over two months only 844 SOL was withdrawn from them — 1.4%. This while the set as a whole lost half its money in the same period.

The explanation is probably simple. This is compensation for work done four years ago, for some people on a project long since closed. The keys to those accounts are wherever they were left in 2022.

We built a search page

All 33,864 accounts are searchable in our Zombie Stake checker. Enter your wallet address and see whether a forgotten stake account is registered to it. Search by both withdrawer and staker: in 373 accounts these are different keys, and you may only remember one of them.

A few important caveats:

• We have no access to these funds. Nobody does, except the holder of the withdrawer key. This is simply public blockchain data organized into a usable form.
• Finding yourself on the list is not the same as having a claim. The key gives the right, not the table entry.
• We do not “unlock” or “recover” anything. If the key is lost, this page won’t help — and neither will anyone else. Anyone promising otherwise is scamming you. Publishing a list like this inevitably attracts fraudsters; please keep that in mind.
• Data is as of epoch 1028. If you withdrew later, your account may still appear.

Why any of this matters

The obvious takeaway — “check your old wallets” — is actually the least interesting one.

Here’s the more interesting part. Exiting a stake in Solana takes two actions, and the second one never reminds you of itself. No notification, no accruals ticking up to catch your eye. The account simply goes quiet and stops working.

And in roughly a third of cases even the first action isn’t taken by a person but by a bot — because the validator they trusted their money to shut down. That, importantly, isn’t a failure but a mechanism working as intended: the network clears dead stake out of the leader schedule and stops spending throughput on it. The side effect is that it moves someone else’s money into a state the owner won’t learn about until they go looking.

And 33,864 times, someone stopped between the first action and the second. Half of them, judging by the amounts, will never think about it again. But 199 accounts hold between one and ten thousand SOL, and eighteen hold more than ten thousand. Some of those people simply don’t know.

Instead of a conclusion

This research is another reminder that there is no perfect solution — only the one that better fits a particular owner’s needs.

Liquid staking has real strengths in this specific story. Had the same funds been in a pool, their owners would:

• not be losing yield because the delegation was removed long ago and nobody got around to withdrawing;
• not depend on the fate of a single validator — the pool redistributes stake itself when a node stops working;
• see an LST token right in their wallet, rather than a separate stake account that is easy to forget exists.

To be fair: liquid staking isn’t the absence of risk, it’s a different set of it — smart contract risk, secondary market liquidity, the LST-to-SOL exchange rate. But this one particular problem — “the money sits there doing nothing because I forgot about it” — simply doesn’t arise.

JPool Research · in collaboration with the SolBrothers validator · epoch 1028 snapshot (September 4, 2026)
Data: two complete snapshots of every stake account on Solana — epoch 1000 (July 10, 2026) and epoch 1028 (September 4, 2026), 1,497,112 and 1,433,252 accounts respectively. Snapshots taken via getProgramAccounts with direct parsing of account state, including lockup fields. Deactivation method verified on a random sample of 120 accounts.