But AI has now become an entrenched part of the sector-wide narrative.
Recently launched projects provide some evidence that cable operators, hyperscalers, providers, and more are building networks that adapt to the data-intensive AI workloads expected to dominate the next decade.
However, is AI demand genuinely behind the new investments into subsea infrastructure, or have existing projects been repackaged with a trendy AI label?
On examination, some case studies from September show that the answer may lie somewhere between the lines.
Factor One: Hyperscalers take the lead
Some of the strongest evidence comes from hyperscalers.
On 21 September, Meta announced Petal: a 7,000km transatlantic cable connecting the US and France due to enter service in 2029. The project will be the world’s first subsea cable at petabit capacity – delivering 1 petabit per second (Pbps) – and the first to deploy multi-core fibre technology at scale.
Meta has connected their subsea cable projects to the AI rush before. When announcing Project Waterworth in 2025, Meta explained that demand for AI and emerging technologies makes investing in developing infrastructure crucial.
Otherwise, future AI growth would put more pressure on pre-existing cables – taxing capacity and undermining their resilience.
Meta’s ongoing investment in subsea cables follows a wider trend. Hyperscalers are moving beyond buying international bandwidth and are becoming direct owners and developers of the global infrastructure that supports their products.
Factor Two: More routes mean more resilience
Earlier this month, Capacity reported on Exa Infrastructure’s new 6,552km Exa Meridian cable linking New Jersey, USA, and the UK. The cable will deliver up to 500Tbps and become the first single-operator transatlantic cable built in more than a decade.
Despite being carrier-neutral themselves, the company is specifically citing demand from hyperscalers and cloud providers as a justification for the system. However, these clients want to pave the way for AI-powered bandwidth growth by buying as much capacity as possible to meet AI’s exponential demands.
Meanwhile, Australian developer SUBCO announced three new Indo-Pacific cable systems spanning more than 24,000km, explicitly targeting the surge in AI-driven bandwidth demand across the region.
“AI factories are dead without massive subsea connectivity to carry tokens from factory to customer,” Bevan Slattery, founder and co-CEO of SUBCO, commented at the time.
Taken together, these announcements show the degree to which AI is becoming a meaningful factor in long-term capacity planning. Operators increasingly expect future AI workloads to generate large data flows between different continents and are willing to invest in building cables at length to do it.
Factor Three: Converging the technologies
Other September developments highlighted how cable infrastructure and data centre infrastructure are becoming increasingly intertwined.
This month, Telxius opened a new carrier-neutral, AI-ready landing gateway and edge data centre in Cancun. The facility is designed as the landing point for the upcoming Tikal cable while also housing a data centre.
Mónica Martínez, CMO at Telxius, described the project as a convergence of subsea connectivity, terrestrial networks, and edge infrastructure.
Martínez explained the benefits of merging the technologies, saying: “By bringing connectivity and computing resources closer to where demand is growing, we enable lower latency, greater resilience, and faster access to cloud, content, and AI-driven services for customers across the region.”
Similarly, Digital Realty announced plans to build a new cable landing station at its LAX12 facility in Los Angeles.
Rather than treating cable landings as standalone assets, the project will interconnect with terrestrial networks while also offering subsea cable operators a full landing and powering service for their systems.
The goal behind this ecosystem is to strengthen connectivity between Asia-Pacific and North America while creating a more scalable environment for international traffic – traffic which is becoming increasingly busy in the AI era.
These projects reflect a developing industry trend. As AI workloads become distributed across multiple facilities, it becomes valuable to optimise international connectivity through convergence rather than just focusing on landing cables.
Factor Four: Global connectivity
It would be misleading to argue that the entirety of new submarine cable investment is purely concerned with AI. For example, EllaLink’s deployment of two submarine branching units for future connections into Brazil doesn’t seem driven by the AI race.
The system will eventually provide dedicated connectivity to Pará and Maranhão in the Amazon region, as part of the ‘Mais Conectado’ project delivered under the EU Global Gateway strategy.
Pará and Maranhão will benefit from a combined initial capacity of 17Tbps, improving connectivity for underserved communities that have historically relied on long terrestrial routes.
Investing in robust connectivity is also a humanitarian effort for the two states, which currently have some of the lowest scores on the UN’s Human Development Index (HDI) in the country. Better connectivity will result in more resilient public digital services in sectors like healthcare, education, business development, and e-government.
Likewise, upgrades such as the Africa Coast to Europe (ACE) consortium cable’s move to more than 30Tbps has positive consequences for both connectivity and quality of life.
Philippe Recco, President of the ACE Management Committee, said: “International routes drive economic growth, digital inclusion, and cloud adoption across Africa,” noting that the system connects 500 million people from more than 20 countries.
While more cables are being installed in the region, resilience is still a concern. March 2024 saw faults on four cables in West Africa, which caused outages in several African countries.
Therefore, the ACE upgrade is aimed at serving broad demand growth across international networks, even if AI contributes to that growth.
The business case in these two examples is largely rooted in regional development. While future AI services will likely benefit from upgraded infrastructure, connectivity and economic development are the core pillars of the projects.
What’s the verdict?
AI is likely accelerating demand rather than creating it. Demand for international connectivity was already growing due to cloud computing, streaming, enterprise digitalisation, and internet usage.
This demand also grows stronger as underserved and isolated regions work to become connected. The importance of the internet as a catalyst for economic and cultural development has become clear in recent years; the UN itself said “Digital connectivity should be a human right” in 2020.
However, AI is increasingly powering business investment decisions and influencing where capacity is deployed. Demonstrating this is the increasing volume of projects being designed specifically with AI-era requirements in mind, rather than the number of overall projects being constructed worldwide.
AI is becoming an infrastructure planning standard rather than a traffic trend. As a result, the subsea sector is no longer just responding to current bandwidth demand; the sector is working to become a pivotal backbone in the infrastructure AI will require.
RELEVANT STORIES
Capacity Europe 2026
The 25th anniversary edition of Capacity Europe 2025 will bring together 3,500+ decision-makers from the global connectivity and digital infrastructure community.
