Analysis

Mercury's post-Synapse pivot: lessons for crypto-startup banking

By Stephan Kulik · 4 May 2026 · 7 min read

Mercury — the US business neobank used by virtually every tier-1 crypto fund and protocol team for USD operating accounts — spent 2024 and the first quarter of 2025 unwinding its dependency on Evolve Bank & Trust, the FDIC-insured partner bank that had stored customer deposits since Mercury's 2019 launch. The unwind was the direct downstream consequence of the Synapse collapse, the largest fintech-banking failure since the 2023 SVB run, and the path Mercury took provides a clearer-than-most case study for crypto-startup treasurers thinking about counterparty concentration risk.

What Synapse was and why it failed

Synapse was a banking-as-a-service (BaaS) middleware company that sat between fintechs (Mercury, Yotta, Juno, Mainvest, Copper) and FDIC-insured banks (most prominently Evolve). When a fintech app wanted to offer "checking accounts," what they were technically offering was a sub-ledger position inside Synapse's omnibus account at Evolve. Users transacted with the fintech's UI; Synapse held the ledger of who owned what fraction of the omnibus; Evolve held the actual cash on its balance sheet under FDIC.

The model worked while the ledgers reconciled. They stopped reconciling in early 2024. Synapse filed for Chapter 11 in April 2024, then converted to Chapter 7 (liquidation) in May 2024 after Mercury and Evolve disputed the integrity of the ledger Synapse handed off. The bankruptcy trustee identified a shortfall of approximately $65M to $96M between what the ledger said end-users owned and what was actually in Evolve's omnibus. End-users — including many at Yotta and Juno — lost access to their money for months; some are still partially recovering as of mid-2026 through trustee distributions.

The FDIC insurance the apps advertised was real — but it covered the bank, not the ledger. When the ledger broke, FDIC didn't pay the end-users. It paid Evolve.

Mercury's exposure and the unwind

Mercury's relationship with Evolve was structurally different from Yotta's or Juno's — Mercury used Evolve as a direct partner bank rather than going through the Synapse middleware for the bulk of its customer accounts. But Mercury's "Mercury Treasury" product, which swept customer cash into money-market funds for yield, did pass through Evolve as one of several rails. And the reputational contagion was real: customers heard "Evolve" in news cycles about Synapse and asked their fund admins whether their cash was at Evolve. Most CFOs do not enjoy that conversation.

Mercury announced in late 2024 that it would migrate customers off Evolve. By early March 2025 the migration completed. The replacement architecture is multi-bank by design:

Choice Financial Group — primary FDIC-insured operating-account partner for new and migrated US customers
Column N.A. — chartered nationally; used for ACH and wire infrastructure
Patriot Bank, N.A. — third FDIC-insured partner for sweep deposits

The result is that a Mercury customer's deposits are now spread across three banks rather than one, with Mercury handling the sub-ledger natively rather than via a third-party middleware. Each of the three banks holds its own slice; FDIC coverage stacks across them, raising effective coverage from the standard $250k-per-bank floor up to $5M+ depending on the customer's "Mercury Vault" sweep configuration. For crypto businesses with seven-figure operating balances, this is materially better than the pre-Synapse single-bank model.

Why this matters for crypto-startup treasurers

Crypto operating accounts have always been a constrained market. Banks willing to service crypto-native businesses are few — Mercury, Brex, Bridgebank (now part of MUFG Union), Anchorage Federal Savings Bank, and a small number of others. After the SVB failure in March 2023, after the Signature wind-down in March 2023, and after the Synapse collapse in 2024, the count of "available banks" for a crypto fund is small enough that single-counterparty risk is now the default condition rather than an edge case.

Three concrete lessons for crypto-startup CFOs:

One — read the partnership disclosures, not the marketing. "FDIC-insured" on a fintech's signup page tells you the partner bank is FDIC-insured. It does not tell you whether your cash is held in an omnibus account intermediated by a BaaS middleware (Synapse-style), or in a sub-account directly at the bank, or — at the simplest end of the spectrum — in a fully-licensed direct-deposit account at a chartered institution. The first model is fragile. The second is workable. The third is what most experienced treasurers default to after 2024.

Two — multi-bank by design. The Mercury-Choice-Column-Patriot architecture is the new floor for any operator above the $250k FDIC threshold. Single-bank concentration was acceptable when banks were boring; it stopped being acceptable in 2023. Spread the operating cash, accept the small operational tax of managing multiple integrations, and price the spread reduction against the risk reduction.

Three — distinguish operating-cash counterparty from custody counterparty. A crypto fund's USD operating cash and its crypto custody are entirely different risk surfaces. Coinbase Prime or Anchorage Digital Bank holds the crypto under either NYDFS BitLicense (Coinbase) or OCC-chartered federal trust banking (Anchorage). Mercury or Brex holds the USD payroll. Conflating them — using a single platform for both — is convenient but introduces correlated failure across two regulated layers.

What Mercury still doesn't do

Mercury is a USD operating bank with crypto-friendly KYC. It is not a crypto bank in the sense that Nexo or Crypto.com are crypto banks. There is no native buy-sell button. There is no in-app yield on stablecoins. Wires to Coinbase Prime, Anchorage, and Kraken are seamless and rapid; that's the integration story. The architecture is "USD bank that does not debank you for being crypto" rather than "bank that holds your crypto." For most crypto-startup CFOs that's exactly what they want — operating cash separate from on-chain capital, both custody-isolated.

Wider read on the BaaS-collapse pattern

Synapse was the first big BaaS collapse but is unlikely to be the last. The structural problem — fintech UI on top of middleware sub-ledger on top of FDIC bank, with no single party fully accountable for the end-user balance — recurs across the BaaS market. The post-Synapse regulatory direction is clear: the OCC and the FDIC issued joint guidance in October 2024 (FIL-43-2024) tightening expectations for banks partnering with BaaS middleware. Several smaller BaaS providers have shut down preemptively. Treasurers should expect more consolidation over 2026-2027.

For crypto-banking specifically, the pattern overlays directly with the lessons we documented in the Juno post-mortem — Juno was a Synapse-era casualty, and its end-users are part of the trustee-recovery process now. The Ziglu special administration in the UK is a parallel-but-different failure mode (regulatory rather than middleware-collapse) but the end-user experience — cash-locked for months — is similar.