A new Cambridge study finds that 72% to 92% of the world's submarine cables would need to fail simultaneously before Bitcoin's network suffers significant global node disconnection. The finding carries important implications for understanding how robust decentralised systems truly are when infrastructure fails.
Yet the research reveals something more unsettling lurking beneath the surface. Whilst random cable faults have historically had negligible impact on Bitcoin, targeted attacks on key submarine chokepoints or major hosting providers could disrupt the network with far less damage. Whilst random cable failures require 72-92% removal to cause damage, a targeted attack on the cables with the highest betweenness centrality, the ones that serve as chokepoints between continents, drops that threshold to 20%. And targeting the top five hosting providers by node count, Hetzner, OVH, Comcast, Amazon, and Google Cloud, requires removing just 5% of routing capacity to achieve the same impact.
Researchers at the Cambridge Centre for Alternative Finance last week published the first longitudinal study of Bitcoin blockchain's resilience to physical infrastructure disruption, analysing 11 years of peer-to-peer network data against 68 verified submarine cable fault events. The scope of the analysis is substantial. Cambridge researchers Wenbin Wu and Alexander Neumueller built a dataset spanning 2014–2025: 8 million Bitcoin node observations, 658 submarine cables, and 385 cable failure reports.
When researchers examined real-world cable failures, the data supported the resilience narrative. The researchers concluded 87% of the 68 historical cable events produced less than a 5% node impact. The largest single event, when seabed disturbances off Côte d'Ivoire damaged 7-8 cables simultaneously in March 2024, knocked out 43% of regional nodes but affected only 5-7 bitcoin nodes globally, roughly 0.03% of the network. The correlation between cable failures and bitcoin's price was essentially zero, at -0.02.
One finding challenged conventional wisdom about network visibility. The growing use of TOR by Bitcoin nodes has unexpectedly increased the network's physical resilience, as relay infrastructure is concentrated in highly connected European countries that are difficult to isolate. As of 2025, 64% of Bitcoin nodes use TOR, making their physical location unobservable. TOR relay infrastructure is heavily concentrated in Germany, France, and the Netherlands, countries with extensive submarine cable and land border connectivity. The four-layer model consistently showed higher resilience than the clearnet-only baseline, with TOR adding between 0.02 and 0.10 to the critical failure threshold.
This pattern reflects an uncoordinated shift across the Bitcoin ecosystem. TOR adoption surged after censorship events like Iran's internet shutdown in 2019, the Myanmar coup in 2021, and the China mining ban. The Bitcoin community shifted toward censorship-resistant infrastructure without any central coordination, and that shift happened to also make the network physically harder to disrupt.
The real vulnerability, however, traces not to geography but to commerce. Bitcoin is vulnerable because a large portion of its network nodes and mining hash rate is concentrated in a small number of commercial data centers run by companies like Amazon, Google, and Hetzner. This concentration emerged for practical economic reasons. Running a Bitcoin node, especially a mining node, requires significant capital expenditure for hardware and substantial operational costs for power and cooling. As a result, participants increasingly turn to large-scale, centralised data centers for efficiency and cost savings.
The study shows resilience declined sharply during Bitcoin's boom. Resilience declined sharply during 2018-2021 as the network grew rapidly but concentrated geographically, hitting its lowest point of 0.72 in 2021 during peak mining concentration in East Asia. The China mining ban in 2021 forced redistribution, and resilience partially recovered to 0.88 in 2022 before settling at 0.78 in 2025.
The paradox is stark. Bitcoin's architecture distributes consensus across thousands of independent nodes yet concentrates those nodes in the hands of a few large infrastructure providers. Random acts of nature pose minimal threat to such a system. Coordinated human action poses far greater danger. Neither outcome flows from a flaw in Bitcoin's protocol design; both reflect the economic incentives that shape infrastructure decisions in the real world. The question for Bitcoin's long-term security is whether this concentration can be reduced without sacrificing the operational efficiency that made it economically rational in the first place.