A new study from the Cambridge Centre for Alternative Finance delivers the first long-term, data-driven look at how resilient Bitcoin is to physical infrastructure damage — and the picture is both reassuring and sobering.
Using 11 years of peer-to-peer network data and 68 verified submarine cable fault events, researchers simulated thousands of failure scenarios to see what it would take to significantly fragment the Bitcoin network. They ran 1,000 Monte Carlo simulations per scenario and compared real-world outages to modeled attacks and random failures.
The headline: Bitcoin is extremely robust to random infrastructure failures. Between 72% and 92% of inter-country submarine cables would need to be cut simultaneously before the network sees major node disconnection. In the dataset of 68 real cable incidents, more than 87% caused under 5% node impact. The single largest event — seabed disturbances off Côte d’Ivoire in March 2024 that damaged 7–8 cables — disrupted 43% of regional connectivity but only knocked offline about 5–7 Bitcoin nodes globally, roughly 0.03% of the network. Correlation between cable outages and Bitcoin price was effectively zero (-0.02), meaning infrastructure hits are invisible against normal market noise.
But resilience depends on the adversary. The study draws a crucial distinction between random failures (storms, accidents) and targeted actions (coordinated state attacks or regulatory shutdowns). If an attacker focused on the cables with the highest betweenness centrality — the chokepoints that fiber traffic must pass through — the threshold for damage falls to about 20% cable removal. Worse, targeting the five hosting providers that host the most Bitcoin nodes — Hetzner, OVH, Comcast, Amazon, and Google Cloud — requires removing just 5% of routing capacity to produce the same disruptive effect.
That asymmetry reframes the threat model: natural disruptions are unlikely to knock Bitcoin offline, but a concentrated campaign against specific undersea routes or a few dominant hosting providers could materially degrade the network.
The paper also traces how resilience has evolved. Bitcoin was most geographically distributed and robust in 2014–2017 (critical failure threshold 0.90–0.92). Rapid growth and geographic concentration during 2018–2021 drove resilience down, hitting a low of 0.72 in 2021 amid mining concentration in East Asia. The 2021 China mining ban forced redistribution and resilience rebounded to 0.88 in 2022, settling at about 0.78 in 2025.
One counterintuitive finding involves TOR. In 2025, 64% of Bitcoin nodes used TOR, making their physical locations unobservable. Previous assumptions held that hidden TOR nodes could mask geographic concentration and thus unseen fragility. Cambridge’s four-layer model found the opposite: TOR relays are heavily concentrated in Germany, France and the Netherlands — jurisdictions with dense submarine and land connectivity — making them harder to isolate. Adding TOR to the model increased the critical failure threshold by 0.02–0.10 compared with a clearnet-only baseline. The researchers describe this dynamic as “adaptive self-organization”: censorship and shutdown events (Iran’s 2019 internet blackout, Myanmar’s 2021 coup, and China’s 2021 mining ban) spurred TOR adoption, which inadvertently strengthened the network’s physical resilience without central coordination.
Why this matters now: geopolitical hotspots like the Strait of Hormuz and regional conflicts raise the real possibility of undersea infrastructure disruption. The study’s takeaway is that Bitcoin would likely shrug off random cable failures — but coordinated, targeted attacks on chokepoints or a handful of major hosts remain a credible risk.
Implications for the ecosystem are clear: continued geographic distribution of nodes, diversification of hosting providers and paths, and attention to chokepoints in the physical internet all matter for long-term resilience. The network has proven adaptive, but the study maps the limits of that adaptability and highlights where policymakers and operators should focus to prevent single points of failure.
Read more AI-generated news on: undefined/news
