
Ethereum Block Builders Vault Hack: Builders Took Most of $3.4 Million
More than half of the $3.4 million stolen from an Ethereum savings vault this week went not to the hackers who found the flaw, but to the companies that assemble Ethereum’s blocks, according to onchain data that lays bare one of the strangest features of modern decentralized finance. The Ethereum block builders vault hack drained K3 Capital’s sBOLD vault in about eight minutes on Wednesday, November 18, 2026, through 97 transactions that manipulated the vault’s accounting to redeem shares at inflated values. Blockchain analytics firm Bitquery traced the flow of funds and found that Ethereum block builders received about $1.78 million in payments from automated bots competing to have their exploit transactions processed first. An automated trading bot that copied the attack extracted $1.91 million, or 56 percent of the stolen funds, and then handed most of that money to the block builders. The original attacker and an associated wallet retained only about $520,000.
The incident is a vivid illustration of maximal extractable value, or MEV, the term for the profit that can be earned by controlling the order of transactions inside a block. On Ethereum, the order in which transactions are processed determines who gets the best price, who captures an arbitrage opportunity, and, in this case, who gets to drain a vulnerable vault first. When a profitable opportunity appears in the public mempool, where pending transactions wait to be included in a block, bots race to execute it first, and they pay the block builders to make sure their transaction is the one that gets in. In the sBOLD attack, that race became a bidding war so intense that the payment to secure a place in a single block reached 44 Ether.
How the sBOLD vault exploit worked
According to Bitquery’s reconstruction, the exploiters manipulated the vault’s share accounting by repeatedly depositing BOLD, introducing wrapped Ether that was initially excluded from the vault’s accounting, and then converting that Ether into BOLD inside the vault before redeeming shares at inflated values. The key flaw was in how the vault measured the value of assets that arrived from outside: the wrapped Ether was not counted when it entered, but the BOLD it was converted into was counted when the shares were redeemed. Each cycle inflated the apparent value of the attacker’s shares, and each redemption withdrew funds that belonged to other, legitimate depositors. Repeating the cycle 97 times over about eight minutes emptied the vault of roughly $3.4 million.
The MEV bidding war that followed
The most remarkable part of the story is what happened after the first attacker found the flaw. The exploit was visible onchain: every transaction was recorded in public, and any automated system watching the blockchain could see exactly what the attacker was doing. Within minutes, an automated trading bot copied the attack and ran it itself, extracting $1.91 million, or 56 percent of the total drained funds. That bot was not the original attacker, and it had not discovered the vulnerability. It simply watched the public blockchain, recognised a profitable pattern, and repeated it faster and more aggressively than the person who found it.
That is where the block builders come in. On Ethereum, a transaction does not get into a block simply because it was sent first. Since the network adopted proposer-builder separation, specialized firms known as block builders assemble the transactions into blocks, choosing the order that earns them the most money, and then pass those blocks to validators, who propose them to the network. Builders compete for the right to have their block chosen, and the way they compete is by bidding: they pay validators for the privilege, and they collect payments from the traders and bots whose transactions they include. A bot that wants its transaction processed before a rival’s will pay the builder to prioritise it, and when several bots are chasing the same opportunity, the price of priority rises rapidly.
In the sBOLD attack, the bidding war pushed the payment for inclusion in a single block to 44 Ether, and the total paid to builders reached about $1.78 million. The copycat bot that extracted $1.91 million handed over the bulk of that sum to the builders, leaving itself with a net profit of only about $130,000, or 6.81 percent of the amount it extracted. The original attacker, who had done the work of finding the flaw, ended up with about $520,000. The builders, who had done none of the work and taken none of the legal risk, collected the largest share. A review of the analysis by CoinDesk found that Titan built most of the blocks containing the bot’s transactions, with BuilderNet, Ultrasound and Quasar also processing them. It is important to be precise: the $1.78 million figure measures incoming payments to builders, not confirmed profits, because builders must themselves pay validators to publish their blocks through the MEV-Boost mechanism.
MEV block builder profits and the economics of Ethereum
To understand why the builders profited, it helps to understand what MEV is and why it exists. Every transaction on Ethereum is a request to change the state of the blockchain, and the order in which those requests are processed can create or destroy value. A large trade on a decentralized exchange moves the price, so a trader who can place their own trade just before it can profit from the movement. A liquidation on a lending protocol creates a discount, so a bot that can claim it first captures the gain. The value that can be extracted by controlling transaction order is MEV, and it is a structural feature of any blockchain where transactions are visible before they are confirmed.
Proposer-builder separation was introduced to manage MEV, not to eliminate it. Before the change, validators themselves assembled blocks, which meant that the largest validators could capture the most MEV and the system tended toward centralization. By separating the roles, Ethereum allowed specialized builders to compete on their ability to assemble profitable blocks, while validators simply chose the most lucrative block offered to them. In practice, it created a new layer of intermediaries who profit from the same ordering power, and the sBOLD attack shows how that layer behaves when a large, sudden opportunity appears. The builders did not create the vulnerability, and they did not write the exploit. They simply sold priority to the highest bidder, which is what the system is designed to do.
What the Ethereum block builders vault hack means for DeFi security
The sBOLD incident carries lessons that go beyond a single vault. The first is about the speed of exploitation in DeFi. The vault was drained in eight minutes, and the copycat bot needed only a fraction of that time to replicate the attack. There is no human response fast enough to stop an automated exploit once it is visible onchain, which means that the only reliable defences are those built into the code before deployment: correct accounting, limits on how much can be redeemed in a single block, and mechanisms that prevent uncounted assets from being converted into shares. Security researchers have long recommended exactly those defences for ERC-4626 vaults, including strictly isolating deposits of tokens that are not on the vault’s whitelist and setting maximum redemption limits per block. Whether K3 Capital had those defences is a question the post-mortem must answer.
The second lesson is about audits. ChainSecurity’s May 2025 audit identified a value-extraction route through self-liquidations, and the firm said the issue was mitigated in a later version of the code. The fact that an attack with a similar flavour succeeded eighteen months later does not necessarily mean the audit was wrong or the fix was not applied: the attack may have used a different mechanism entirely. But it does show the limits of an audit as a guarantee. An audit is a snapshot of a particular version of the code at a particular moment, and it can only find the issues its reviewers think to look for. For users, an audit is a useful signal, not a promise.
What happens next for K3 Capital and sBOLD users
For the depositors whose funds were drained, the immediate questions are practical: how much can be recovered, who bears the loss, and whether the vault will reopen. K3 Capital paused the vault after the incident, and Bitquery has cautioned that its analysis covers activity only through its November 18 snapshot, so the full picture may still be developing. Recovery in cases like this is difficult. Some of the funds went to the original attacker, some were paid to block builders who have already passed much of it on to validators, and the copycat bot kept only a small fraction. Tracing the money is possible, since every movement is recorded onchain, but recovering it from pseudonymous addresses is far harder.
Frequently asked questions
What happened in the sBOLD vault hack? Attackers drained about $3.4 million from K3 Capital’s sBOLD savings vault on Ethereum in about eight minutes on Wednesday, November 18, 2026, using 97 transactions that manipulated the vault’s share accounting to redeem shares at inflated values. Blockchain analytics firm Bitquery traced the flow of funds.
How did Ethereum block builders profit from the hack? Automated bots copying the exploit paid block builders about $1.78 million to have their transactions processed first, in a bidding war that reached 44 Ether for a single block. One copycat bot extracted $1.91 million but kept only about $130,000 after paying the builders, while the original attacker retained about $520,000.
What is MEV? Maximal extractable value is the profit that can be earned by controlling the order of transactions inside a block. Traders and bots pay block builders to prioritise their transactions, and when several bots compete for the same opportunity, the payments can approach the full value of the prize.
Was the vault audited? Yes. ChainSecurity audited sBOLD in May 2025 and identified a way to extract value through self-liquidations, which it said had been mitigated in a later version of the code. It is not yet clear whether the attack used that mechanism, whether the fixes were deployed, or whether a different flaw was involved.
Source: CoinDesk
Related reading on MM Crypto Analysis Web
- Phylax DeFi Censorship Debate: New Tool Divides Ethereum
- Coinbase EU Policy Chief Warns MiCA Suitability Tests Could Push DeFi Users Offshore
- OKXICE Tokenized Securities Venue to Launch November 23 With 70-Plus Assets
More in DeFi on MM Crypto Analysis Web.