Investment in subsea cable projects is expected to reach $13B between 2025 and 2027, almost 2x the amount invested between 2022 and 2024, driven by AI buildout
Over 95% of international data and voice call traffic travels through nearly a million miles of underwater communication cables.
Context & Ripple Effects
Subsea cables are becoming a more explicit part of AI infrastructure planning, alongside compute and data centers. Meta's planned five-continent Waterworth system and AWS's Fastnet cable project between Maryland and Ireland show large cloud platforms pursuing dedicated long-haul capacity.
The spending outlook also arrives as ownership and route choices have become strategic: US Big Tech's growing share of used subsea capacity has made the cable layer less of a neutral telecom utility and more of a platform-controlled asset.
First-order effects
- A projected $13 billion of investment through 2027 expands the near-term project pipeline for cable developers, marine installation firms, and equipment suppliers serving international routes.
- Cloud and AI operators gain a clearer path to add dedicated cross-border capacity, reducing reliance on a fixed pool of existing cable capacity as their traffic needs grow.
Second-order effects
- More private cable construction reinforces hyperscalers' ability to control the network links between their regions, increasing pressure on telecom operators to partner, co-invest, or differentiate on routes and services.
- New capacity commitments can intensify competition over cable routes, landing points, and ownership structures, especially where network resilience and geopolitical alignment affect deployment choices.
Third-order effects
- If the buildout persists, subsea connectivity will increasingly be financed and designed as a core AI asset rather than as background telecom infrastructure, concentrating more network influence with the largest compute buyers.
- The result could be a more fragmented global cable map: greater capacity and redundancy on priority routes, but more scrutiny of who owns, operates, and can access the infrastructure.
The trend: AI's infrastructure capital cycle is extending from chips and data centers into the long-distance networks that connect distributed compute regions.