Custody Insurance Markets: How Coverage Is Priced and Syndicated
Custody insurance quotes get read as reassurance and should be read as arithmetic. BitGo carries $250 million of primary cover from Lloyd's syndicates against roughly $90 billion in assets under custody. Copper carries $500 million with Canopius leading through Aon, Coinbase Custody a $320 million crime policy placed by Aon, and Marsh assembled an $825 million facility for digital asset custodians. Evertas, an Arch-backed managing general agent, writes up to $420 million per policy after nearly tripling its limits this year. Those are the largest numbers in the market, and against a roughly $3 trillion asset base with an estimated 89 percent of holders uninsured they describe a protection gap above $1 trillion.
What the Policy Is, and What It Is Not
The contract is older than the asset class. Crypto custody cover is written as specie or crime insurance, the same lines that sit behind fine art, bullion and vaults, and it responds to external theft, internal fraud, physical destruction and loss of custodian-held keys.
The exclusion list is where the instrument actually gets defined. Standard policies carve out market losses and price moves after a theft, regulatory freezes and government seizure, smart contract defects, blockchain protocol failures, owner key mismanagement, accounting errors, and war, terrorism and radioactive contamination. Treatment of credential compromise through phishing or social engineering varies by wording, and that variance matters more than the headline limit.
So the policy covers the vault and not the market, the theft and not the consequence. Read the exclusions before the limit, because the limit answers a question the exclusions may already have closed.
How the Tower Gets Built
No single carrier takes these limits. Capacity is syndicated at Lloyd's, where Arch, Atrium, Beazley and Canopius are among the active names, and a broker assembles a layered tower: a primary layer first, then excess layers above it, each priced lower per dollar because each is less likely to be reached. BitGo has run a $700 million tower built as $100 million in its own name plus $600 million of excess.
Pricing reflects the novelty rather than the loss experience. Custodial crypto runs roughly 0.5 to 3 percent of covered value annually and hot wallet or DeFi exposure 2 to 5 percent, against 0.1 to 0.5 percent for fine art and precious metals in storage. That is a five to tenfold premium for an asset that cannot burn, flood or be dropped.
That pricing forces partial coverage, which is the part institutions misread. A $250 million tower at 2 percent costs $5 million a year; proportional cover against $90 billion is arithmetic that does not close. Every published figure is a tower over a slice, and the number that matters is the limit divided by your own position, not the limit in isolation.
Fireblocks carries $30 million through Coincover, and AnchorWatch writes up to $100 million per customer as a Lloyd's coverholder.
The Exclusion Written for the Wrong Architecture
One clause is worth finding in your own policy. Coverage commonly excludes losses where the custodian does not hold all keys exclusively, and whether a policy responds to shared-key arrangements such as threshold signing depends on the exact wording.
Read that against where institutional custody has gone. The industry moved to multi-party computation and threshold schemes precisely because no single person or machine holding a complete key is the stronger design. The wording still contemplates exclusive possession, the model the technology deliberately abandoned.
The result is a possible inversion: the architecture that reduces the probability of loss may weaken the response of the policy bought against it. That is a drafting lag rather than a considered position, resolvable by endorsement but only if someone asks. The question is answerable in an afternoon: does the policy name the signing scheme in use, and if not, has the carrier confirmed in writing that it responds to it?
Why Capacity Is Capped
The constraint is not capital or appetite. It is diversification.
Specie insurance works because the perils are independent. A vault fire in Zurich tells you nothing about a vault in Singapore, so an underwriter can write many risks and rely on the law of large numbers. Crypto custody breaks that at the root. The asset base concentrates on a small number of custody providers, signing implementations and hardware security module families, so a defect in one widely deployed library is a single event across an entire portfolio rather than an isolated claim.
That is why limits sit in the hundreds of millions per risk while the insured base is measured in trillions. The underwriter is not pricing the probability of your custodian failing. It is pricing the probability that everyone's custodian fails at once, and charging for an exposure it cannot spread.
The loss record supports the caution. Chainalysis put 2025 crypto theft at $6.75 billion across roughly 200 incidents, with the February 2025 Bybit incident alone accounting for about $1.5 billion of ether and roughly 69 percent of the total stolen from services that year. One event moved the majority of a year's losses, which is the shape of a correlated book.
What a 10x Would Take
Two paths exist and only one is plausible.
The first is genuine implementation diversity across the custody base, so an underwriter can build an uncorrelated book. That runs against the consolidation the market is undergoing, where a handful of chartered providers hold most institutional assets. Waiting for diversity means waiting for the industry to become less efficient.
The second is to stop treating the risk as diversifiable and move the correlated tail to capital markets, where investors price it explicitly instead of underwriters absorbing it as though independent. Nexus Mutual's staked cover pools, with the reinsurance layer added through Symbiotic in late 2025, are the small early version of that structure.
The direction is encouraging on the margins. Evertas tripling its per-policy limits is real capacity growth, broker facilities have standardized placement, and the global crypto insurance market moved from an estimated $9.49 billion in 2025 to a projected $13.75 billion in 2026. None of that closes a trillion-dollar gap through conventional underwriting. Custody insurance is a useful control on a specific peril at a modest limit, not a substitute for concentration limits, and firms treating a published tower as a reason to hold more with one provider have the causality backwards.
For informational purposes only. Not an offer to buy or sell any security. Available only to accredited investors who meet regulatory requirements.
