Artificial intelligence is changing the digital world in a way we can measure in physical constraints. Not in abstract “cloud” terms, but in megawatts, fibre pairs, landing stations, cooling systems, and latency budgets.
As leaders met at PTC’26 to explore the convergence of digital infrastructure, telecommunications, and ICT, the conversation was increasingly about fundamentals: resilience, sustainability, and where the next generation of infrastructure should be built - not only what it should do.
Because the intelligence era isn’t just about smarter software. It’s about building the right places - and the right interconnections - to run it.
From “people-to-content” to “compute-to-compute”
The internet era optimized networks to connect people to content. The intelligence era is redesigning networks to connect compute to compute.
AI training and inference drive new traffic patterns:
Enormous east–west flows between GPU clusters, storage, and distributed data lakes.
Latency sensitivity, because AI systems increasingly operate in real time.
Demand for resilient paths, because AI-enabled services are becoming mission-critical.
At the same time, submarine cables remain the backbone of global connectivity. The subsea layer already carries the vast majority of international traffic, and AI only amplifies the need for capacity, route diversity, and operational resilience.
This is why subsea networks, internet resilience, security concerns, sustainability, and “emerging technologies” like AI and big data are not separate conversations: they are one system-level conversation.
The AI constraint nobody can hand-wave away: power
AI pushes infrastructure into a new reality: power and cooling are no longer “inputs” you can assume will be available in the markets you’ve always used.
Many traditional Tier 1 hub markets are constrained by grid availability, expansion headroom, permitting timelines, and cost. In parallel, hyperscale and AI deployments are increasing density and energy requirements.
This shifts how operators think about where workloads should live:
Place power-hungry training and large batch workloads where energy is scalable and efficient.
Place latency-sensitive inference and user-facing services closer to users and interconnection points.
Connect these layers with high-capacity, diverse routes to keep the system coherent and resilient.
It’s the same principle hyperscalers have long used across market tiers, but now applied to AI at scale, with a much bigger energy footprint.
Iberia as a new European exchange zone and why Sines is central to that story
A visible change is underway in Europe’s connectivity map. For decades, the center of gravity sat in the familiar north western arc. Today, the Iberian Peninsula, Portugal’s Atlantic coast paired with Spain’s major metros, is emerging as a new exchange zone where subsea systems, terrestrial backbones, and large-scale compute intersect.
Sines illustrates the strategic logic.
Measured in milliseconds, Sines delivers consistently low and ultra-low latency into Europe’s main consumption markets, turning “Atlantic edge” geography into a performance advantage rather than a periphery position. And from a wider South Atlantic perspective, Sines also offers competitive paths into Latin America, North America, and West Africa.
This matters because intercontinental AI traffic increasingly needs optionality: multiple routes, multiple carriers, multiple on-ramps, all of this without being forced into congested corridors.
Several strategic cable systems reinforce this positioning, including direct connectivity into north-eastern Brazil and new transatlantic paths that strengthen route diversity. Combined with expanding terrestrial connectivity, the result is a credible platform for a South Atlantic Corridor linking the Americas, Europe, and Africa.
The new model: from landing stations to digital ecosystems
Historically, cable landing stations were controlled handoff points. Compute, cloud, and peering tended to live far away, requiring long backhaul routes and duplicated infrastructure. That model is changing.
AI-era infrastructure benefits when subsea landings, carrier-neutral interconnection, internet exchanges, and hyperscale datacenters are co-located, because proximity reduces latency, simplifies architecture, and can improve energy efficiency per bit.
It also creates an ecosystem flywheel:
Each new subsea system increases route diversity and attractiveness.
Each new carrier or IX increases optionality for everyone.
Each new cloud or AI deployment increases traffic gravity, attracting more participants.
This “ecosystem beats island” logic is how other global hubs have scaled successfully. The Atlantic now has an opportunity to build comparable plug-and-play ecosystems, designed from the start for AI-era requirements.
What “AI-ready” really means in practice
“AI-ready” isn’t a marketing label. It is an engineering outcome. AI workloads demand:
• High-capacity intra-campus fiber and optical paths for east–west traffic.
• Ultra-low-latency interconnect between halls and buildings.
• Multiple diverse on-ramps to subsea and terrestrial networks.
• Long-term power and cooling headroom.
• Sustainable design choices that can keep scaling without hitting regulatory or environmental limits.
This is why Start Campus is developing SINES DC as more thana hyperscale site: as a carrier-neutral platform designed to evolve into an interconnection and AI ecosystem.
The design principles reflect the requirements of the intelligence era:
• Neutral meet-me rooms, enabling carriers, subsea systems, and tenants to interconnect without bias.
• A campus dark-fiber fabric engineered for scale(including high fiber counts from day one to connect buildings as the campus grows).
• Infrastructure designed for the dominance of east–west traffic in GPU cluster architectures.
• A sustainability trajectory aligned with renewable power and efficient coastal cooling approaches, including seawater-assisted solutions.
When those layers converge - subsea, interconnection, compute, and sustainable power - you don’t just get a bigger data center: You get a strategic node in the global intelligence fabric.
Subsea networks are evolving too: toward meshed resilience and sustainability
The subsea layer itself is being reshaped by AI demand. The direction is clear:
Move beyond purely linear topologies toward more meshed architectures.
Increase route diversity and “self-healing” capabilities.
Co-locate AI-ready compute nearer to landing points to reduce latency and backhaul costs.
Integrate sustainability into design and operations, not as an afterthought.
There is also growing interest in “dual-use” innovation: using cable infrastructure to support environmental monitoring through new sensing approaches deployed on operational systems.
Taken together, these trends point toward a future where the Atlantic corridor is not only faster and higher capacity, but also more resilient, more observable, and more sustainable.
What PTC’26 is bringing into focus
PTC’26 frames the moment well because its agenda spans the full system:
• Subsea networks: resilience, security concerns, future trends.
• Emerging technologies: AI, big data, quantum, IoT and more.
• Networking technologies: the convergence of data centers, wireless, satellite, and subsea.
• Sustainability: energy efficiency, green technologies, and alignment with global goals.
• Policy and regulation: privacy, cybersecurity, standards, and competition.
• Developing regions and emerging markets: the digital divide and funding connectivity.
That breadth matters. The intelligence era is not a single-industry story. It’s an infrastructure story that requires alignment across cable owners, carriers, hyperscalers, internet exchanges, energy partners, regulators, and local communities.
A practical conclusion: build corridors, not chokepoints
The next decade will reward infrastructure strategies that create optionality:
• Optionality in routes.
• Optionality in power.
• Optionality in interconnection.
• Optionality in where compute can scale sustainably.
The South Atlantic corridor through Sines is compelling because it brings these ingredients into one geography: subsea gateways, EU stability, renewable potential, and an ecosystem model designed for openness.
The opportunity now is execution: building the partnerships, the interconnection density, and the resilient mesh that turns geography into a lasting advantage.
From internet to intelligence, the map is being redrawn.
The question is: who builds the nodes that make it work?
________________________________________
Fernando Azevedo, Head of Connectivity, joined from Amazon Web Services in Dublin where he worked as Network Development Manager for the AWS global backbone. He previously worked for leading connectivity players, including Angola Cables.
Get in touch with Fernando Borges Azevedo