Risks to the AI Economy from Attacks on Undersea Data Cables

1. Why the AI Economy Depends on Undersea Data Cables

Modern AI runs in hyperscale cloud data centers, not on user devices. Inference for LLMs, generative image tools, and recommendation engines is centralized, then delivered via undersea fiber cables that carry over 95% of international data traffic.[1] Subsea networks are the physical backbone of AI services.

AI adoption is intensifying this dependence:

• Large language models are displacing traditional web search, with search traffic down an estimated 15% as users shift to conversational systems.[1]
• This consolidates activity and profits in a few cloud and hardware providers whose services must cross transoceanic cables to reach users.

💡 Key takeaway: As AI centralizes computation and profits, it also centralizes network risk onto a few physical chokepoints.

Enterprise AI is moving into core workflows by 2026, under tight margin pressure.[3] When connectivity fails, firms can lose access simultaneously to:

• Internal automation (agentic copilots, analytics)
• Customer products (AI search, chatbots, personalization)
• Supplier integrations (API‑based AI services)

Dependencies now extend to small firms and freelancers. Designers, illustrators, and translators increasingly rely on LLMs for drafts; even short cloud outages can halt work. Around 26% of illustrators and 36% of translators already report lost work to generative systems, and roughly three‑quarters expect future income reductions.[1]

⚠️ Key point: The same AI tools that displace traditional jobs also create fragile dependencies that can fail catastrophically if the network layer is attacked.

2. Economic and Geopolitical Risks from Attacks on Undersea Cables

Attacks on subsea infrastructure can trigger cascading risks, from technical disruption to economic and geopolitical instability. The diagram below summarizes this pathway.

flowchart LR
    title Path from Undersea Cable Attacks to AI-Economy Impacts
    A[Subsea cable attacks]:::danger --> B[Bandwidth loss]:::warning
    B --> C[AI disruption]:::warning
    C --> D[Workflow failures]:::info
    D --> E[Firm stress]:::danger
    E --> F[Market repricing]:::warning
    E --> G[Geopolitical tension]:::danger

    classDef danger fill:#ef4444,stroke:#ef4444,color:#ffffff;
    classDef warning fill:#f59e0b,stroke:#f59e0b,color:#000000;
    classDef info fill:#3b82f6,stroke:#3b82f6,color:#ffffff;

Physical sabotage, spoofed navigation near landing sites, or cyber‑physical attacks on repeaters and management systems could all degrade access to cloud‑hosted AI services.[2] Even partial bandwidth loss can:

• Slow inference and break latency‑sensitive workflows
• Worsen existing margin pressures from costly AI rollouts[3]

📊 Data point: Cybersecurity spending is already projected to significantly outpace overall IT spending as AI‑driven threats grow.[3]

Labor, productivity, and finance are systemically exposed. CEO expectations for AI‑driven growth remain high, yet:

• Only 1 in 50 AI investments delivers “transformational” value
• Only 1 in 5 shows any measurable ROI[5]

A connectivity shock could rapidly reveal over‑optimism:

• Productivity gains vanish when AI tools go offline
• “Always‑on AI” revenue models fail
• Leaders respond with layoffs or restructuring to protect margins[5]

Financial markets are similarly vulnerable. Much of the perceived AI value is concentrated in a few cloud and semiconductor firms dependent on scalable, always‑available AI services.[1][6] If cable attacks disrupt key regions:

• Usage growth slows or reverses
• Investors reassess valuations
• Volatility spikes in AI‑heavy indices[6]

💼 Market risk: AI stocks already face technological and valuation risks; physical network fragility adds another underpriced factor.[6]

National‑security dynamics raise the stakes further. Semiconductors and compute capacity are now central to state power.[2] The first publicly reported AI‑orchestrated hacking campaign in 2025 showed that agentic systems can shift the offense–defense balance in cyberspace.[2] In crises, states or proxies may see undersea cables as legitimate targets to weaken an adversary’s compute and data flows, turning commercial AI outages into geopolitical flashpoints.

⚠️ Key point: Undersea cables have become strategic assets in a world where compute and connectivity define national leverage.

3. Building Resilience: Policy, Security, and Investment Responses

Advanced chips and compute are increasingly treated as national‑security assets; undersea data cables should receive similar status.[2] Designating them as critical AI infrastructure would support:

• Dedicated funding for cable surveillance, redundancy, and repair
• Joint public‑private threat‑intelligence sharing
• Inclusion of cable resilience in national AI and cyber strategies

Regulation of malicious AI use is still patchy. Analysis of the EU AI Act shows uneven coverage of intentional misuse such as AI‑assisted planning of physical sabotage, with some risks only indirectly addressed.[4] This leaves room for blended operations in which AI‑enabled cyber tools map vulnerabilities and coordinate physical attacks on seabed infrastructure.

💡 Key takeaway: AI security and cable security cannot be managed as separate silos; attackers will combine them.

Rising cybersecurity budgets can be directed toward:

• Hardening software that manages cable traffic and routing
• Deploying anomaly‑detection models to spot AI‑driven intrusions
• Integrating subsea‑network status into enterprise AI resilience plans[3]

Investors should expand AI risk analysis beyond algorithms and margins to physical network resilience.[1][6] Key questions include:

• How diversified are cable routes and landing points?
• Do data centers span multiple geopolitical blocs?
• Has management modeled week‑long regional connectivity shocks?

📊 Data point: As AI reshapes sectors from automotive to enterprise software, dependence on a few cloud platforms means cable‑level disruptions can propagate quickly across portfolios.[1][3][6]

Conclusion: Treating Cables as First‑Order AI‑Economy Risk

The AI economy is concentrating computation, capital, and careers into a few platforms that depend on fragile undersea data cables. At the same time, geopolitical competition over compute, gaps in malicious‑use regulation, and AI‑orchestrated cyber campaigns create incentives to target those cables.[2][4]

For policymakers, business leaders, and investors, the implication is clear: undersea data cables must be treated as a first‑order AI‑economy risk. Building cable resilience into AI strategies, regulatory regimes, and capital allocation—before a major disruption forces change—is now essential for anyone serious about the future of the AI‑driven economy.