What banks fear most isn’t a falling coin price—it’s having their keys stolen. This time, Google has gone and split the keys up.
On October 28, at its inaugural Digital Assets Symposium in New York, Google Cloud unveiled Google Multi-Party Compute (MPC), designed for banks and other financial institutions to custody and transfer crypto and tokenized assets. The product is currently in private preview and is not yet generally available. Commercial launch is planned for early 2027, and pricing has not been announced.
The idea is straightforward: the complete private key is never kept in one place. Instead, it’s split into multiple parts and distributed across Google Cloud, a bank’s own hardware, and other clouds. Before a transaction can be signed, it has to pass through a policy engine: spending limits, destination restrictions, and approvals from multiple departments all have to check out. Customers still retain their own key material. The product also supports key recovery, and the architecture can be upgraded to resist quantum attacks in the future.
The numbers help explain why banks are discussing this first. Chainalysis estimates that 43.8% of crypto assets stolen in 2024 were lost due to private key compromises. According to DeFiLlama, as of June this year, blockchain projects had lost a cumulative $16.69 billion to hacks, with around 40% linked to private key compromises. Google’s parent company, Alphabet, has already put the system to use for its own digital assets; the related stock token is $GOOGL .
My take: what has held banks back from entering the space was never whether blockchains exist, but whether they’d trust a single company with the keys that can move their money. Splitting the keys and checking transactions before signing builds “don’t just take my word for it” right into the architecture—and that’s more convincing than slogans. But a private preview is not a full launch. The real test comes after commercial availability in early 2027: how many banks will actually entrust it with large assets, and who takes responsibility if policies are misconfigured or approvals get stuck? The industry will first look at the design partners and the initial real-world rollout before deciding whether this is an industry standard.
On October 28, at its inaugural Digital Assets Symposium in New York, Google Cloud unveiled Google Multi-Party Compute (MPC), designed for banks and other financial institutions to custody and transfer crypto and tokenized assets. The product is currently in private preview and is not yet generally available. Commercial launch is planned for early 2027, and pricing has not been announced.
The idea is straightforward: the complete private key is never kept in one place. Instead, it’s split into multiple parts and distributed across Google Cloud, a bank’s own hardware, and other clouds. Before a transaction can be signed, it has to pass through a policy engine: spending limits, destination restrictions, and approvals from multiple departments all have to check out. Customers still retain their own key material. The product also supports key recovery, and the architecture can be upgraded to resist quantum attacks in the future.
The numbers help explain why banks are discussing this first. Chainalysis estimates that 43.8% of crypto assets stolen in 2024 were lost due to private key compromises. According to DeFiLlama, as of June this year, blockchain projects had lost a cumulative $16.69 billion to hacks, with around 40% linked to private key compromises. Google’s parent company, Alphabet, has already put the system to use for its own digital assets; the related stock token is $GOOGL .
My take: what has held banks back from entering the space was never whether blockchains exist, but whether they’d trust a single company with the keys that can move their money. Splitting the keys and checking transactions before signing builds “don’t just take my word for it” right into the architecture—and that’s more convincing than slogans. But a private preview is not a full launch. The real test comes after commercial availability in early 2027: how many banks will actually entrust it with large assets, and who takes responsibility if policies are misconfigured or approvals get stuck? The industry will first look at the design partners and the initial real-world rollout before deciding whether this is an industry standard.
