Google Cloud’s private-preview service splits signing keys across separate environments, with commercial availability targeted for early 2027.


Google Cloud is aiming at one of institutional crypto’s most persistent security problems: a private key that, if stolen, can give an attacker control of assets. Its new Google Multi-Party Compute service splits signing keys across separate environments, so no single system holds the complete key. That changes the security question from whether one machine can be protected perfectly to whether an attacker can compromise enough separate pieces to authorize a transaction.
That distinction matters because digital asset systems rely on cryptographic signatures to approve transfers. In a conventional setup, whoever gains access to the relevant key may be able to act as its owner. Splitting the key can reduce the damage caused by a breach in one place, though it adds coordination among the parties that hold the shares. The approach doesn't remove the need for sound access controls; it changes how control is distributed.
How shared signing is supposed to work
Unveiled Wednesday at Google’s first symposium dedicated to digital assets, the service lets banks keep one key share on their own hardware, another in Google Cloud and a third with a competing cloud provider. The parties can sign transactions together without rebuilding the full key. Institutions choose the signing threshold, such as two of three shares or five of seven; compromising fewer shares than required won’t allow an attacker to forge a signature.
Here's the thing: the threshold is a policy choice with operational consequences. A lower threshold can make signing easier when one participant is unavailable, but it also means fewer compromised shares may be enough to authorize activity. A higher threshold can demand broader agreement, while making the process more dependent on every required participant being reachable. Google says customers can set that balance themselves, but the service's practical value will depend on how institutions configure it and manage exceptions.
Distributing shares across an institution, Google Cloud and another provider also spreads reliance across different systems. That could help limit the effect of a single compromised environment. It can introduce new dependencies, too: institutions will need to coordinate their own hardware and cloud arrangements, and understand how signing proceeds when a participant or service is unavailable. Similar systems typically have to account for both security and continuity. A design that blocks unauthorized transfers but regularly interrupts authorized ones won't be attractive to financial operations teams.
Controls and availability
Google says it cannot read customers’ key shares in plaintext, including shares stored on its servers. A policy engine in Google Confidential Space checks proposed transactions against institutional rules, such as spending caps and approved destinations. That review is advisory: enough authorized signers can still override a denial.
The advisory limit is significant. Transaction rules may provide a useful check before a transfer is signed, but they aren't an absolute barrier if the required signers decide to proceed. That leaves institutions responsible for deciding who can authorize an override, under what circumstances, and how those decisions are monitored. The distinction between a policy warning and a hard technical restriction can be easy to miss in product descriptions, yet it affects how much protection the control provides.
The platform combines threshold signing with transaction policies and key recovery. It’s in private preview, with unnamed banks and partners testing it. Alphabet is also using the technology to protect its Ethereum and Solana holdings. Google is targeting commercial availability for early 2027 and hasn’t disclosed pricing.
Private preview gives Google a chance to learn how the service behaves in institutional workflows before wider availability, but the announcement doesn't identify the participating banks or describe their testing results. Alphabet's use of the technology shows that Google is applying it to its own holdings as well. It doesn't, by itself, establish how the service will perform for customers with different systems, approval processes or recovery needs.
Network support and longer-term threats
At launch, the service will support ECDSA, used by Bitcoin and Ethereum, and Ed25519, used by Solana. Google’s documentation also discusses longer-term AI and quantum risks, while focusing on immediate threats such as stolen keys, outages and lost key material. Post-quantum cryptography could be added later.
That focus is sensible. Stolen credentials, service interruptions and misplaced key material are present operational problems for any organization managing access to valuable digital assets. Quantum computing is a longer-term concern in Google's materials, and the mention of possible future support shouldn't be read as a claim that the current product addresses it. For now, customers will need to judge the service against the networks and signing methods it supports at launch.
Recovery deserves particular attention. Losing key material can create a different kind of crisis from theft: an institution may be unable to authorize a legitimate transaction even though no attacker has taken control. A recovery feature may help address that risk, but the announcement doesn't spell out the recovery process or the safeguards around it. Those details will matter to a bank assessing whether the service fits its existing controls.
What banks will need to decide
Rich Widmann, Google Cloud’s director of digital assets strategy, said regulated financial institutions adopting blockchains and tokenization need “scalable, proven infrastructure to secure their clients’ assets.” The product puts Google in competition with custody and security providers including Fireblocks, BitGo and Fordefi. Its test will be whether banks see distributing control across providers as a practical way to limit the damage from a compromised system.
This is more significant than it looks: Google is proposing that cloud providers become part of the control structure for institutional crypto assets, not merely a place to run supporting software. That may appeal to organizations seeking to divide authority, but it also asks them to trust a setup spanning providers and their own systems. If you're working in this space, the useful question isn't just how many key shares exist. It's who operates each environment, how exceptions are handled, and whether teams can keep the signing process working during an outage.
And yet, the product's promise is still prospective. It's in private preview, pricing hasn't been disclosed, and commercial availability is targeted for early 2027. The service will have to show that its security model can fit real institutional procedures without making routine signing and recovery too cumbersome. More control distributed across systems. That’s the pitch; the operational details will decide whether banks buy it.
Discussion
Sign in to join the discussion.