Bancor (BNT): When 396,720 ETH Flooded the First AMM and Froze Ethereum

1. A name from money that never entered circulation

Before Bancor was a token, bancor was money that never circulated. John Maynard Keynes and E. F. Schumacher conceived it in 1940–42 as a supranational unit for settling trade between central banks. The name came from the French banque or, 'bank gold.' Britain brought the plan to Bretton Woods, where it was not accepted.^[3]

The Bancor Protocol is named in honor of the transnational currency unit for recording and balancing international trade as John Maynard Keynes proposed at the Bretton Woods Conference of 1944.^[2]

— Bancor Protocol Whitepaper, June 2017

The protocol's April 28, 2017 draft v0.77 was followed by draft v1.00 on June 6. Eyal Hertzog, Guy Benartzi, and Galia Benartzi were credited as authors, with contributions and edits from Scott Morris and Bernard Lietaer. Court papers place the Israel-based founders under Bprotocol Foundation, a Swiss nonprofit seeking a 'global standard for intrinsically tradeable currencies.'^[2][1][10]

The design put reserves inside each smart token's contract. Anyone could buy from or sell to that contract: tokens were issued on entry and destroyed on exit. 'SmartTokens do not need to be traded in an exchange in order to become liquid,' the paper said, promising convertibility without a listing.^[2]

2. Bancor reported 396,720 ETH as its website buckled

On June 12, 2017, Bancor ran what coverage called the largest token sale to date. Bancor reported 10,885 participants and 396,720 ETH—about $150 million, or roughly $153 million on the project's site—surpassing The DAO's reported $152 million from the previous year.^[4]

The dollar figure depended on ether's price. Finance Magnates counted 396,619 ETH worth over $142 million at $360 per ether; the $153 million figure reflected higher marks. These were proceeds actually received, not a company valuation.^[5][4]

The disruption was at Bancor, not Ethereum. The one-hour window became three after what the team called 'overwhelming demand and traffic, and massive malicious attacks' on its site and application. Buyers saw errors and pending payments; contemporaneous reports do not say the chain halted.^[4][5]

Bancor created 79,323,978 BNT and sold roughly half; one buyer reportedly took $27 million. The whitepaper earmarked 20% of proceeds for the network token's Ether reserve, the on-chain collateral for its pricing formula.^[4][2]

3. The math: reserves and ratios instead of order books

A smart token's price equaled reserve balance divided by the product of supply and its fixed Constant Reserve Ratio (CRR). When CRR was below 100%, buying added reserves and minted tokens as price rose; selling burned tokens and released reserves as price fell. The formula replaced an order book.^[2]

Token-changer contracts held two or more reserves at a total CRR of 100%, with arbitrage correcting relative prices. BNT was the hub—'the first smart token to be deployed using the Bancor protocol'—and held a single Ether reserve.^[1][2]

The whitepaper described these contracts 'acting as automated market makers': the project's wording for fixed-rule contracts quoting continuous prices without a human market maker. It supports a formula-priced liquidity protocol published in spring 2017, not a claim that Bancor was the first AMM or influenced later designs.^[1]

Uniswap instead lets liquidity providers deposit equivalent values of ETH and the associated ERC-20 token into an exchange contract, then prices swaps by the constant-product rule x·y = k. Swaps change the balances of the two traded assets rather than minting or burning those assets against a CRR reserve. Bancor priced by reserve ratio and supply; Uniswap by pool balances.^[12][1][2]

4. A compromised wallet and the power to freeze BNT

In July 2018, Bancor lost about $23.5 million: $12.5 million in Ether, $1 million in NPXS, and $10 million in BNT, according to TechCrunch. The entry point was not the pricing formula but a privileged upgrade wallet.^[6]

a wallet used to upgrade some smart contracts was compromised.^[6]

— Bancor, in a statement quoted by TechCrunch

Bancor said it froze the stolen BNT but could not freeze the Ether or NPXS, and took its exchange offline. The response demonstrated control over its own token—the tension Charlie Lee criticized as the news broke:^[6]

An exchange is not decentralized if it can lose customer funds OR if it can freeze customer funds. Bancor can do BOTH. It's a false sense of decentralization.^[6]

— Charlie Lee, Litecoin creator (post quoted by TechCrunch)

On June 18, 2020, Bancor disclosed a flaw in a v0.6 contract deployed two days earlier. A bug in three versions led users to grant infinite approvals, and 1inch identified 'a public method' that could exploit them. Bancor said its developers used a controlled exploit to move assets to safety; reports also noted user losses.^[8][7]

5. Bancor and Uniswap, then a lawsuit and a discovery request

Uniswap v1's whitepaper describes ETH-and-token reserves whose product remains invariant during swaps. Bancor minted and burned tokens against a reserve ratio; Uniswap changed the balances of an ETH-token pool under x·y = k. These are different pricing mechanics.^[12][1][2]

The compared primary documents support technical contrast, not a claim that Bancor influenced Uniswap or that one protocol succeeded the other. The record here is deliberately narrower than a lineage story.^[12][1][2]

A later complaint alleged this Bancor timeline: v1 in 2017, v2 in April 2020, v2.1 in October 2020 with impermanent-loss protection, and v3 on May 11, 2022. Plaintiffs alleged that a June 19, 2022 withdrawal spike led to the protection's suspension and losses for withdrawing investors.^[10]

The class action was filed in May 2023. On September 6, 2024, Judge Robert Pitman dismissed without prejudice the claims against the foundation, LocalCoin, and founders for lack of personal jurisdiction and failure to allege domestic transactions—not on the merits. In their May 27, 2025 motion, plaintiffs stated that the court had reinstated the non-moving Bancor DAO and that the clerk had entered default against it; they asked for discovery concerning jurisdiction, domestic transactions, class certification, and damages before a contemplated default-judgment application. The motion was a party request, not a default judgment, merits ruling, or fraud finding.^[9][11][10]