Picture a “cloudless” day—not in the meteorological sense, but one in which digital operations come to a sudden, catastrophic halt. Financial transactions stall. AI applications lose access to data. Supply chains slow, and customer experiences suffer.

It may sound improbable, yet recent events suggest otherwise. In 2024, subsea cable damage caused widespread internet outages that disrupted banks, fintech platforms and cloud-based businesses across Africa; in 2025, Red Sea cable cuts severely disrupted access across Asia and the Middle East. Internet traffic was rerouted through longer paths, increasing latency for cloud services and adding costs for businesses reliant on real-time digital operations.

These incidents are a reminder that what we call “the cloud” is neither virtual nor invincible. We often imagine it floating somewhere above us. In reality, it lies at the bottom of the ocean: More than 99% of all international data is carried not by satellites but by a physical backbone of subsea cables.

The regional digital economy runs uninterrupted only because companies made continuous, heavy investment in critical transit hubs like India and Singapore.

AI is placing unprecedented demands on the cloud, the foundation of today’s digital economy. AI workloads and hyperscale data centres need ever-greater bandwidth—with 400 gigabits (gbps) per second now the baseline and 800 gbps already emerging.

As countries and businesses race to scale AI, much of the focus has been on compute and data centres. But the networks connecting those systems need to keep pace as well. That raises a critical question for Asia: is the region overlooking a key AI infrastructure bottleneck?

Without sufficient investment in subsea cables and end-to-end networks, businesses could spend heavily on AI infrastructure only to see the returns eroded by disruption, higher operating costs and unreliable access to data.

For the C-suite, that presents a looming crisis: while AI investment is a board-level strategic priority for 77% of enterprise leaders, 65% of enterprises continue to operate on legacy or still-developing infrastructure, according to a Tata Communications study.

With AI workloads making global traffic patterns increasingly unpredictable, enterprises need dynamic, flexible data-centre-to-data-centre connectivity that can scale rapidly with demand to protect the bottom line and operational stability. Meeting this need will require continued investment from network providers in the infrastructure and capabilities needed to deliver scalable, resilient connectivity.

Most discussions of Singapore’s AI ambitions revolve around compute capacity, talent and regulation. Those are all important. But they overlook what happens when the data that AI systems depend on cannot be transmitted quickly, reliably or securely.

Singapore and India are emerging as two of the region’s most important data transit hubs, underpinned by their strategic locations and growing digital infrastructure.

Singapore serves as a gateway to Southeast Asia, with one of the world’s densest concentrations of data centres. Meanwhile, India acts as a link between Asia, the Middle East and Europe, supported by major cable landing points and expanding data centre capacity in cities such as Mumbai and Chennai. Together, these strengths make both countries important hubs for international data traffic, underscoring the need for resilient and diverse connectivity routes to keep data moving efficiently between major financial centres and commercial workloads.

The industry needs to rethink how it approaches network architecture. For decades, infrastructure was built to support predictable, linear traffic. Today, infrastructure must now accommodate increasingly volatile AI-driven data demands.

This requires better integration between subsea cables, terrestrial fibre and individual data centres, as a bottleneck at a coastal landing station can immediately negate the speed of a transoceanic cable.

However, keeping these cables secure and running requires far more than laying fiber in the ocean. True resilience demands a coordinated focus across four dimensions: capacity planning, network diversity, protection and repair readiness, and disciplined lifecycle operations.

This behind-the-scenes management—including robust monitoring, strategic spares management, tight integration of subsea and terrestrial networks, and clearly defined long-term governance—is what ensures a single cable cut does not become a regional blackout.

Crucially, this industry-led push is now being reinforced by public policy. The recent adoption of the Enhanced ASEAN Guidelines for Strengthening Resilience and Repair of Submarine Cables—which recommends capping repair processing times at 10 days—is a vital step forward. Regional policymakers are making rapid repair readiness of economic stability.

For businesses, the practical consequence is a change in what gets funded. Compute and data centres often take priority in investment decisions because their capacity and returns are visible. Meanwhile, the cables, landing stations and alternative routes that keep them connected are less visible, and can be harder to justify as investments in their own right. Yet, the trade-off is clear: underinvestment in connectivity risks leaving expensive AI infrastructure vulnerable to disruption, congestion and higher operating costs. Connectivity must be treated as an essential and strategic asset not simply an operational cost hidden in the IT budget. The cloud begins beneath the ocean. As AI adoption accelerates, Singapore’s competitiveness as a global digital hub will depend on the strength of that invisible lifeline beneath the sea.

Kapil Kumar Jain is vice president of global network services at Tata Communications.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.