According to reports, Google will launch Google Multi-Party Compute today at its first digital assets summit in New York—a cryptographic custody system that splits signing keys across multiple systems. The product is currently undergoing private previews with several financial institutions. According to CoinDesk, Alphabet has already begun using this system to secure its own assets.
Google plans to achieve commercial availability in early 2027, but has not yet disclosed pricing.
The system applies multi-party computation to signature keys. The key is split into multiple shards, each held by different parties or machines, and a transaction requires a predetermined number of shards to complete the signature.
The full key is never assembled in any single location. The bank can store one share on its own hardware, another on Google Cloud, and a third with another cloud provider. Institutions can set thresholds such as 2-of-3 or 5-of-7. Before any signature is executed, the policy engine running in Google Confidential Space verifies that the transfer complies with limits, destination restrictions, and approval rules.
Google stated, according to CoinDesk, that it cannot access any key shares in unencrypted form.
Cryptographic security issues
The selling point of this product is built upon a long-standing issue. Chainalysis estimates that 43.8% of stolen cryptocurrencies in 2024 were linked to private key compromises. Data from DeFiLlama shows that, as of June, approximately 40% of the $16.69 billion lost through hacks resulted from compromised keys. In June, an attacker stole seven keys from a compromised developer’s computer at Humanity Protocol and emptied wallets on two networks.
Such incidents demonstrate that excessive centralization of signing authority creates a single point of failure. The industry therefore recommends adopting multi-signature solutions from multiple vendors to distribute control across different hardware and software providers.
Google has also linked this product to AI-enabled fraud risks. Chainalysis found that in 2025, scams associated with AI tool providers resulted in an average loss of $3.2 million per incident, compared to an average loss of $719,000 for other scams. The report merely correlates these two figures and does not prove that AI caused the difference.
Ethereum researcher Justin Drake has warned that AI may eventually find mathematical shortcuts that undermine wallet signatures. However, no AI system has yet demonstrated this capability. Google’s MPC design addresses compromised credentials and unauthorized transfers, not the breaking of underlying cryptography.
Quantum computing presents another, more long-term risk. Google researchers estimate that a sufficiently powerful future machine could derive a Bitcoin private key from an exposed public key in approximately nine minutes. Around 6.9 million bitcoins have exposed public keys. Google states that its architecture could later adopt post-quantum signatures without requiring institutions to rebuild their systems, but has not provided a timeline. The background on how quantum computers threaten Bitcoin illustrates which addresses face the highest risk.
Google is entering a market led by Fireblocks, Fordefi, and BitGo. Google told CoinDesk that its banking relationships and security expertise will give it an advantage. The company is also positioning itself as a neutral provider for those unwilling to rely on banks that are competitors.
Banks are already using specialized platforms, such as the custody services offered by BNP Paribas in collaboration with Metaco, and multi-institutional vaults built on BitGo.