Every custody provider says the same thing in a press release: your keys are safe, replicated, redundant. The interesting question for an IP reader is never whether redundancy exists — it is how the redundant copy is moved without ever assembling the secret it protects. US12,615,147, granted to Nasdaq, Inc. on April 28, 2026, is a clean answer to that question, and claim 1 is unusually legible about it.

The setting is multi-party computation (MPC) custody. Instead of one private key sitting in one vault, the key is split into shares held by separate nodes; a transaction is signed by the nodes computing together, so the whole key is never materialized in one place. The patent does not claim MPC custody itself — that ship sailed years ago. What it claims is the replication step: copying those shares from a primary asset-custody subsystem in one cloud data center to a backup subsystem in a different data center, for disaster recovery, without the key ever being reconstructed end to end.

The backup MPC client sends an export public key from each backup MPC node to the primary MPC client, where each export public key is part of a corresponding export public key-export private key pair.

That sentence is the hinge of the whole grant. Read the direction of travel carefully: the backup side originates the export public keys and pushes them to the primary side. The primary MPC client then hands each export public key to its matching primary node, and in response each primary node returns its private-key share already encrypted under that node-specific export public key. The encrypted shares travel to the backup client, which routes each one to its corresponding backup node, and only that backup node — holding the matching export private key — can decrypt and store its share.

The limitation that matters is the pairwise binding of export keys to nodes. Because each share is encrypted under a public key whose private half lives only at the destination node, no intermediary — not the primary client, not the backup client, not the transport — can read a plaintext share. And because the shares are per-node, decrypting any single one does not reconstruct the signing key. The claim therefore captures a specific topology: backup-originated key material, node-by-node encryption, destination-only decryption.

What it conspicuously does not capture is just as important for scope. The independent claim is not a monopoly on “backing up an MPC wallet.” A scheme that ships shares re-encrypted under a single shared transport key, or one where the primary subsystem generates the wrapping keys, or one that reshares via fresh secret-sharing rather than this export-pair handoff, reads outside the recited mechanism. The patent fences a method, not the goal.

The CPC classification confirms where Nasdaq is building. The grant sits in H04L 9/0894 (key distribution involving secret sharing), H04L 9/0825 (public-key encryption), H04L 9/3247 (digital signatures), and — notably — H04L 9/50, the blockchain-specific cryptography class. That last code is the tell that this is custody infrastructure aimed squarely at on-chain assets, not a generic enterprise key-management filing.

A practical reason to dwell on the directionality is that it inverts the intuition most engineers bring to backup. The reflex is to imagine the primary system pushing copies outward — encrypting under its own keys and shipping the bundle to a standby site. This claim does the opposite: the standby site reaches in and supplies the wrapping keys, so trust flows from destination to source. That choice has a real security consequence. If the primary subsystem is the one that is more likely to be attacked while online and handling live signing traffic, you do not want it minting the keys that protect the backup copies; you want the colder, quieter backup to hold the private halves. The claim encodes that threat model into its structure, which is part of why it reads as engineering rather than boilerplate.

The recovery semantics also bound the claim usefully. The decrypted backup share is stored so it can be used if the backup subsystem takes over operation — the patent is explicit that the point is failover, not routine multi-site signing. That matters for an infringement read: a system that keeps two live MPC clusters co-signing in real time, rather than holding one as a recoverable standby, is operating in a different mode than the one the claim describes. The grant is about resilience and disaster recovery, and its limitations are written for that purpose.

It is worth noting who is filing it. Nasdaq is not a crypto-native exchange; it is market-structure incumbent that has been quietly assembling a digital-asset custody estate. The named inventors, Shankar Raju and Jiayue Chen, also appear on a companion grant, US12,445,274, covering dynamic provisioning of MPC node clusters across multiple computing environments. Read together, the two patents sketch a deliberate portfolio: one grant on standing up the MPC clusters, another on replicating their key shares for recovery. That is the shape of an operator fencing its own custody plumbing rather than a defensive one-off.

For a reader trying to map the custody landscape, the takeaway is narrow and precise. This is not a claim that should be cited as “Nasdaq patented MPC.” It is a claim over a particular disaster-recovery handoff: destination nodes mint export key pairs, source nodes encrypt their shares under those keys, and only the destination can decrypt. Whether a competing custody architecture infringes turns entirely on whether it performs that specific export-pair exchange — and most of the field has other ways to move a backup. The grant is real, the mechanism is concrete, and the scope is exactly as wide as claim 1 says it is and no wider. To read the claim language yourself, the canonical record is deep-linked above.