Cambridge Study: Undersea Cable Outages Do Not Affect Bitcoin; Centralization of Cloud Services Is the Real Risk

Author: CryptoSlate

Translation: Deep Tide TechFlow

Deep Tide Introduction: A research team at Cambridge University used 11 years of data and 68 verified submarine cable failure events to demonstrate that cutting underwater cables has almost no impact on the Bitcoin network. However, they also discovered a real vulnerability—not underwater, but within cloud service providers like Hetzner, AWS, and Google Cloud. This study’s conclusion strongly counters the “Bitcoin fragility theory” and provides a quantitative framework for actual infrastructure risks.

Full text below:

In March 2024, submarine disturbances near Ivory Coast severed seven underwater cables, with the Internet Disruption Impact Score (IODA) exceeding 11,000.

For Bitcoin, the global impact was minimal. Only about five nodes were affected, accounting for roughly 0.03% of the network, with an impact of -2.5%, within normal fluctuation ranges.

No price volatility, no consensus disruptions.

A new Cambridge study analyzed 11 years of Bitcoin network data and 68 verified submarine cable failure events, concluding that submarine cable failures have historically caused very limited disruption to the Bitcoin network.

In comparison, targeted attacks on a few hosting networks are more effective at disrupting visible nodes than random infrastructure failures, with effects differing by an order of magnitude.

Notably, China’s mining regulation and the global spread of censorship-resistant infrastructure may have inadvertently led Bitcoin toward a more robust network topology.

Tor has long been regarded as a privacy tool, now serving as a structural resilience layer. Most Bitcoin nodes run on Tor.

Empirical data contradicts concerns

Cambridge researchers Wenbin Wu and Alexander Neumueller built a dataset covering 2014 to 2025: 8 million Bitcoin node observations, 658 submarine cables, and 385 cable failure events, cross-referenced with interruption characteristics.

Out of 385 reports, 68 matched verifiable outages, with 87% of verified cable events resulting in less than 5% node changes. The average impact was -1.5%, with a median of -0.4%.

Node disruptions show almost zero correlation with Bitcoin price (r = -0.02). Cable failures making headlines in affected regions usually leave no trace in Bitcoin’s distributed network.

The study models Bitcoin as a multi-layer network: physical connections via 354 submarine cables linking 225 countries, routing infrastructure (autonomous systems), and the peer-to-peer Bitcoin overlay.

In scenarios of random cable removal, the critical threshold for over 10% node disconnection ranges from 0.72 to 0.92. Before meaningful network fragmentation occurs, most inter-country cables would need to fail completely.

Where is the real vulnerability?

Targeted attacks operate very differently. Removing 72% to 92% of cables is needed to reach a 10% node disconnection threshold in random failures; for high betweenness centrality cables, this drops to 20%.

The most effective strategy is targeting top autonomous systems (ASNs) by node count, removing only 5% of routing capacity to reach the threshold.

The authors characterize this ASN-targeted scenario as “hosting provider shutdowns or coordinated regulatory actions, rather than physical cable cuts.” The model identifies top networks including Hetzner, OVHcloud, Comcast, Amazon Web Services (AWS), and Google Cloud.

A March 2026 Bitnodes snapshot confirms this pattern: out of 23,150 accessible nodes, Hetzner hosts 869, Comcast and OVH each host 348, Amazon 336, and Google 313.

This does not mean “five providers can destroy Bitcoin.”

Even if the public network is fully removed, most nodes will still operate because Tor carries the majority of the network. But this finding reveals where coordinated actions could cause connection disruptions and propagation failures that random cable failures do not.

Recent cloud outages illustrate this risk. Amazon attributed a March 2026 outage to software deployment failures; other reports describe outages in AWS Middle East following attacks on data centers.

These events did not cause significant impact on Bitcoin but proved that provider-related failures are real, not just theoretical.

Tor as a structural resilience layer

The composition of the Bitcoin network has changed significantly.

Tor adoption grew from near zero in 2014 to 2,478 nodes (23%) in 2021, then to 7,617 nodes (52%) in 2022. By March 2026, out of 23,150 accessible nodes, 14,602 are Tor nodes, accounting for 63%. This growth coincides with multiple censorship events: Iran’s internet shutdown in 2019, Myanmar’s coup in 2021, and China’s mining ban in 2021.

Node operators have shifted toward censorship-resistant infrastructure without coordination, indicating the network’s self-adaptive, self-organizing capacity.

Tor introduces a challenge: most Bitcoin nodes are now unobservable in terms of location.

The authors address this by constructing a four-layer model, treating Tor relay infrastructure as an independent network layer. Tor relays are physical servers with known locations.

Using consensus weight data from 9,793 relays, they modeled how cable failures that disconnect countries can simultaneously take relays offline.

Surprisingly, the four-layer model consistently yields higher critical failure thresholds than models considering only the public network, with increases between 0.02 and 0.10.

Most Tor relays’ consensus weights are concentrated in Germany, France, and the Netherlands—countries with extensive cable connections. Cutting cables that disconnect peripheral countries does not weaken relays in well-connected countries.

Attackers would need to remove more infrastructure to simultaneously disrupt both public routing and Tor circuits.

The China factor

Bitcoin’s resilience hit its lowest point of 0.72 in 2021, coinciding with peak hash rate concentration.

Cambridge data shows that in 2019, 74% of hash power was in East Asia. Geographic concentration of nodes reduced the network’s resilience from its peak between 2018 and 2021 by 22%.

The rebound in 2022 was strong. After China’s mining ban, infrastructure decentralization increased the threshold to 0.88, and Tor adoption accelerated.

While the authors avoid single-causal explanations, regulatory pressure drove geographic redistribution and adoption of censorship-resistant infrastructure—both enhancing network robustness.

The apparent centralization is partly due to measurement bias. As Tor adoption increased, the public network sample concentrated in fewer locations, with the Herfindahl-Hirschman Index rising from 166 to 4,163, but Hetzner’s actual share decreased from 10% to 3.6%. This centralization reflects sample composition changes, not true centralization.

Cloud services are the real risk

Concerns about submarine cable security will continue to rise. Baltic Sea investigations, EU security toolbox, and reports on Russian infrastructure all point to ongoing geopolitical anxieties.

For Bitcoin, historical data shows most cable incidents are noise.

The real infrastructure concern is whether policy coordination, cloud outages, or hosting restrictions can cause connection shocks at the autonomous system level.

The operational threshold for ASN-targeted scenarios is 5% of routing capacity, representing a critical point where visible network nodes experience significant disruption, not a consensus failure.

Most of Tor’s share provides a safety net in extreme scenarios. Protocol layers not included in this study—such as relay networks, compact block relay, and Blockstream satellites—add additional resilience layers, making the estimates conservative.

Bitcoin is not as fragile as critics claim, but it is not completely decoupled from infrastructure either.

The network demonstrates graceful degradation under stress rather than catastrophic collapse. Censorship pressures have driven infrastructure adoption, which in turn enhances resistance to coordinated risks.

Threat models focusing solely on submarine cable cuts overlook a closer choke point: a few networks where coordinated actions can cause temporary disruptions without dramatic submarine operations or warfare.