A single power failure at Google Cloud's europe-west2-b availability zone in London on October 20, 2026, knocked Revolut's 80-million-user platform offline for nearly eight hours, disabling card payments, ATM withdrawals, foreign-exchange swaps, and all crypto trading across Europe. Lloyds, HSBC,...
"Customers' money and data remained safe." — Revolut, official statement during service restoration, October 20, 2026
A single power failure at Google Cloud's europe-west2-b availability zone in London on October 20, 2026, knocked Revolut's 80-million-user platform offline for nearly eight hours, disabling card payments, ATM withdrawals, foreign-exchange swaps, and all crypto trading across Europe. Lloyds, HSBC, Bank of Scotland, Halifax, Royal Mail, and the England and Wales court video system went down simultaneously. No blockchain network experienced a liveness failure.
The incident is the second major cloud-induced crypto-access disruption in twelve months. In October 2025, an AWS DNS failure in US-EAST-1 locked Coinbase and Robinhood users out of trading for several hours. Together, the two events underscore a structural fault in the digital-asset stack: the ledger layer is decentralized, but the access layer — exchanges, wallets, custody engines, and trading interfaces — runs overwhelmingly on three cloud providers that control 63% of global infrastructure spending. The UK's new Critical Third Parties regime, which placed AWS, Google Cloud, Microsoft Azure, and Oracle under direct FCA/PRA/Bank of England oversight starting July 13, 2026, received its first live stress test less than 100 days after activation.
At 06:08 UTC on October 20, 2026, Google Cloud reported "multiple services experiencing issues" in its europe-west2 region. The root cause: a power event at the europe-west2-b availability zone — one of three zones in the London region. Seven categories of Google Cloud services failed: Cloud Run, Google App Engine, Google Kubernetes Engine, Cloud Filestore, Google Cloud NetApp Volumes, BigQuery Data Transfer Service, and Managed Service for Apache Kafka. In total, 38 Google Cloud products were affected.
Google mitigated the power event by 11:36 UTC. Full service restoration took until approximately 14:00 UTC — 7 hours and 52 minutes after initial failure.
Affected organizations (confirmed):
| Sector | Organizations | |--------|--------------| | Banking | Revolut, Lloyds, HSBC, Halifax, Bank of Scotland | | Logistics | Royal Mail, DPD | | Retail | Screwfix | | Telecom | BT | | Property | Rightmove | | Justice | England and Wales Cloud Video Platform (court hearings) |
Revolut was the hardest hit. Its status page listed ATM withdrawals, card payments, transfers, identity verification, sign-in, chat support, and stock and crypto trading as fully down. Downdetector recorded nearly 1,500 incident reports for Revolut alone. The outage affected Revolut's 13 million UK customers and extended to users in France and Spain, part of its 80-million-strong global base.
The distinction matters: no blockchain network failed. Bitcoin continued producing blocks. Ethereum finalized slots on schedule. Solana's transaction throughput was uninterrupted. The "outage" was purely at the access layer — the centralized interfaces through which users interact with decentralized networks.
Revolut's 16 million crypto users worldwide lost the ability to buy, sell, or view their crypto holdings. Reports surfaced of portfolios displaying blank balances — a display artifact of the backend database failure, not actual loss of funds. No on-chain assets were at risk; private keys and blockchain state remained intact.
This pattern — access denied while the ledger runs uninterrupted — repeats with increasing frequency. It reveals that for most retail users, "crypto" is not a peer-to-peer system. It is a service mediated by fintech applications hosted on centralized cloud infrastructure. The blockchain's theoretical censorship resistance is irrelevant if the user's only interface to that blockchain is a mobile app running on Google Kubernetes Engine.
Three providers — AWS (28% market share), Microsoft Azure (20%), and Google Cloud (15%) — account for 63% of global cloud infrastructure spending as of Q2 2026. For financial services specifically, concentration is higher. A Bank of England and FCA survey found the top three third-party cloud providers accounted for 73% of all cloud providers named by UK financial firms.
The crypto industry mirrors this pattern. According to reporting by The Defiant and CoinDesk, AWS powers key operations for Coinbase, Binance, Kraken, Crypto.com, and Huobi. The entire Base L2 network — Coinbase's Ethereum rollup — runs on AWS infrastructure. A single region failure can take a major exchange offline while the underlying blockchain continues to process transactions.
The asymmetry is stark: blockchains are designed to have no single point of failure. The access layer that connects users to those blockchains has multiple single points of failure, all concentrated in the same three companies.
The cloud dependency extends beyond exchanges to the network layer itself.
Ethereum: Research by the Cambridge Centre for Alternative Finance found that 31% of Ethereum beacon node activity is concentrated in the United States, heavily clustered on AWS, Hetzner, and OVH. According to Cointelegraph data, three cloud providers account for over two-thirds of Ethereum node hosting at various points. Ethernodes data showed AWS hosting approximately 2,368 execution-layer nodes — roughly 36% of the network.
Solana: As of September 2026, Solana's infrastructure decentralization score stands at 58/100, with approximately 45% of validators running on AWS, according to Web3 Decentralization data. The network's Nakamoto Coefficient is 19. In November 2022, Hetzner's abrupt ban on Solana-related activities removed over 20% of active stake from the network — a live demonstration of single-provider risk. The network survived without liveness failures, but the event reduced its validator set significantly.
This means a sufficiently large cloud failure could, in theory, affect not just user access but the consensus layer itself. The October 20 Google Cloud failure did not reach this threshold — europe-west2 is one region among many — but a simultaneous multi-region failure across AWS and Google Cloud would test this assumption.
The timing is notable. On July 13, 2026 — 99 days before the outage — the UK's Critical Third Parties (CTPP) regime became operational. Under powers granted by the Financial Services and Markets Act 2023, the Bank of England, PRA, and FCA began direct oversight of four designated entities: Amazon Web Services EMEA SARL, Google Cloud EMEA Limited, Microsoft Ireland Operations Ltd, and Oracle Corporation UK Limited.
The regime grants regulators powers to:
The October 20 outage is the first significant incident since the regime activated. The FCA confirmed it was "engaged with affected firms" during the outage. Whether the incident triggers formal action under the new framework remains to be seen, but it provides regulators with a concrete case study of cascading failure — exactly the scenario the CTPP regime was designed to address.
For the crypto sector specifically, the CTPP regime creates an unusual dynamic. Traditional banks using Google Cloud fall under the regime's protection. Crypto-native firms operating in the UK under FCA registration may also benefit, but the regime's primary focus is systemic financial stability, not crypto-specific resilience.
The Google Cloud event follows a nearly identical pattern to the AWS outage of October 20, 2025 — exactly one year prior, to the day. In that incident, a DNS resolution failure in AWS's US-EAST-1 region (Northern Virginia) took down DynamoDB, a core database service. The failure began at 03:11 Eastern Time and lasted approximately three hours.
Coinbase's trading platform and its Base layer-2 network both went offline. Robinhood reported service degradation. ConsenSys' Infura — a critical Ethereum infrastructure provider used by MetaMask and thousands of dApps — experienced disruptions. Binance, Kraken, and OKX, which use different infrastructure configurations, remained operational.
The two events together establish a recurring pattern:
| Factor | AWS Oct 2025 | Google Cloud Oct 2026 | |--------|-------------|----------------------| | Root cause | DNS failure | Power event | | Region | US-EAST-1 | europe-west2 | | Duration | ~3 hours | ~8 hours | | Crypto platforms affected | Coinbase, Robinhood | Revolut | | Blockchain networks affected | None | None | | Users potentially impacted | Millions (US) | 13M+ (UK/Europe) |
The pattern is consistent: centralized access layers fail; decentralized ledgers do not.
Quantifying the direct economic cost of access-layer outages is difficult. Revolut does not disclose hourly crypto trading volume. However, the parameters are measurable:
The indirect cost is more consequential: trust erosion. Each outage reinforces the perception that fintech and crypto platforms are fragile, dependent on infrastructure they do not control. For an industry that positions itself on the premise of user sovereignty and censorship resistance, repeated inability to access funds during cloud failures undermines the core value proposition.
From an economic-value perspective, the outage exposes a hidden cost layer in the crypto stack. Users pay trading fees expecting reliable access. The economic value of that access is dependent on a third-party infrastructure provider whose incentives are not aligned with crypto-specific uptime requirements. Google Cloud's Service Level Agreement offers financial credits for downtime — not protection against missed trades or inaccessible funds.
The October 20, 2026 Google Cloud outage did not break any blockchain. It broke the way most people use blockchains. That distinction matters less than the industry suggests. For Revolut's 16 million crypto users staring at blank portfolios on a Tuesday morning, the theoretical resilience of Ethereum's consensus mechanism was irrelevant. Their access to their assets was mediated by a Google Kubernetes Engine cluster in London that lost power.
The crypto industry has achieved decentralization at the ledger layer but not at the infrastructure layer. Exchanges, wallets, and trading applications remain concentrated on the same three cloud providers that run the rest of the internet. Each outage — AWS in October 2025, Google Cloud in October 2026 — reinforces the same structural weakness.
The UK's CTPP regime represents the first regulatory attempt to address this concentration directly. Whether it can impose meaningful resilience requirements on cloud providers serving crypto platforms — in addition to traditional banks — will determine whether the access layer catches up to the ledger layer's fault tolerance. The data suggests it has not yet done so.