Compute Corridor keynote series

Resilience in digital infrastructure is often discussed as an engineering attribute. It is described through route diversity, landing-station hardening, repair readiness, power redundancy and network architecture. Those things matter. But they are not the whole story. For investors and operators, resilience has become a commercial question as well: a question of whether an asset can sustain demand, preserve pricing power and retain strategic relevance through disruption and change.

The next decade will test that proposition. Demand for international capacity is rising, but it is rising unevenly. Cloud platforms, AI workloads, sovereign requirements, content distribution, enterprise digitisation and new data-centre clusters are changing where traffic originates, where it terminates and how customers value performance. At the same time, geopolitical risk, permitting complexity, supply-chain concentration and climate exposure are making the old assumption of frictionless connectivity less credible.

The investment horizon therefore needs to widen. The question is not simply whether a new cable or fibre route can be financed and built. The question is whether it can become part of a resilient commercial system: one that gives customers options, gives partners confidence and gives capital a credible path to durable returns.

A route map is not a market strategy. A cable can connect two points and still fail to create a compelling proposition. The strongest systems are designed around the commercial ecosystem that surrounds them: landing stations, terrestrial backhaul, metro access, data-centre interconnection, cloud on-ramps, local partners and the products that make capacity usable for customers.

That is why the boundary between subsea and terrestrial fibre is becoming less meaningful from the customer’s perspective. Customers do not buy a wet segment in isolation. They buy a service outcome: reach, latency, availability, scalability, operational assurance and a clear commercial interface. The asset owner that can orchestrate those elements has a stronger position than the owner that simply offers a point-to-point route.

For investors, this changes diligence. It is no longer enough to test construction cost, projected utilisation and a headline demand forecast. The commercial questions are more granular. Which customer segments value the route? What alternatives do they have? How does the system connect into the data-centre and cloud ecosystems where demand is concentrated? Who owns the customer relationship? And what product architecture turns raw capacity into recurring, defensible revenue?

Resilience is often treated as a cost centre: more diverse routes, more spares, more operational capability. In the right market, it is also a source of value. Customers with critical workloads do not only seek the lowest unit price. They seek confidence that their services will remain available when a route is cut, a landing station is constrained or a market is disrupted.

That confidence has to be translated into products. It can appear in protected wavelengths, diverse path commitments, managed restoration, multi-landing options, service-level structures and transparent operational reporting. It can also appear in partnership models that make it easier for customers to buy a resilient end-to-end service across multiple asset owners.

The commercial discipline is to distinguish between resilience that is visible and valued by the customer, and resilience that is expensive but invisible. The former can support differentiation and pricing. The latter may still be necessary, but it should be understood as a strategic cost of participation. The difference is made through product design, sales enablement and the evidence an operator can provide about how the network performs under stress.

Landing stations are no longer passive endpoints. They are strategic interfaces between international systems and domestic digital economies. Their value depends on more than physical security. It depends on the quality of terrestrial connectivity, the density of interconnection, the proximity of data centres, the availability of power and the ability to support multiple commercial models.

A landing station connected to a narrow terrestrial corridor has a different commercial profile from one embedded in a dense, diverse metro and intercity fibre ecosystem. The first may deliver a route. The second can become a platform. It can attract carriers, cloud providers, content platforms, enterprises and local partners because it reduces the friction of reaching customers and applications.

This is where public policy and private capital need to align. Permitting, rights of way, landing access and cross-border coordination can determine whether a system is merely built or genuinely useful. The most effective policy frameworks do not attempt to predict every technology outcome. They create conditions for diversity, interconnection and investment to compound over time.

Traditional infrastructure underwriting often seeks certainty: contracted revenue, visible utilisation and a stable demand curve. Those remain important. But digital infrastructure also rewards optionality. A system that can serve multiple markets, connect into several data-centre clusters, support different capacity products and accommodate new demand patterns may be more valuable than one optimised for a single forecast.

Optionality is not an excuse for vague planning. It must be grounded in real commercial pathways. It means identifying adjacent markets, credible partners, expansion rights, product upgrades and operational capabilities before they are urgently needed. It means building governance that can make decisions quickly when demand shifts. And it means avoiding contractual structures that lock an asset into a narrow role just as the market broadens.

For boards and investment committees, the practical question is simple: what choices will this asset preserve five years from now? The answer should cover routes, customers, products, partners and capital. A resilient investment is one that retains the ability to respond without having to start again.

The digital economy will not become less dependent on connectivity. It will become more dependent on connectivity that is reliable, adaptable and commercially accessible. AI, cloud and data-centre growth will increase the consequences of weak links between international capacity, terrestrial networks and local digital ecosystems.

The opportunity is therefore not only to build more infrastructure. It is to build better commercial systems around infrastructure. That means treating resilience as a product, landing stations as platforms, terrestrial fibre as a strategic extension of subsea investment and partnerships as a core part of the operating model.

The investment horizon is not defined by the next cable landing. It is defined by the value that can be created across the life of the asset. The operators and investors who design for that horizon will be better placed to withstand disruption, capture demand and build infrastructure that remains relevant long after the construction phase is complete.

Hybrid cable systems are compelling because they promise more from the same corridor. They suggest that a single infrastructure investment can carry communications, support sensing, enable environmental observation or provide other strategic functions. In a world of constrained capital, complex permitting and rising demand for data, that proposition deserves serious attention.

But the phrase double the function equals half the price should be treated as a question, not a conclusion. Shared infrastructure can reduce duplication. It can also introduce new interfaces, new governance requirements and new failure modes. The commercial case is not created by technical possibility alone. It is created when the functions, customers, liabilities and operating responsibilities can be designed into a coherent business model.

The right ambition is not to force every cable into a hybrid role. It is to identify where hybridisation creates genuine system value: where the same route, landing point, maintenance capability or data stream can serve more than one strategic need without compromising the economics or reliability of the core communications service.

For most submarine systems, the primary service remains communications. Customers buy capacity because it supports cloud access, content delivery, enterprise operations, mobile networks, financial services and national digital economies. Any hybrid function must be assessed first through its impact on that core proposition.

That means asking difficult questions early. Does the additional function affect route selection, cable design, landing-station requirements, repair processes or service availability? Does it introduce regulatory obligations that change the operating model? Does it create data rights or security concerns that customers will need to understand? And if there is a conflict between functions, who has priority?

A hybrid system succeeds when the answers are explicit. The communications service should not become an unpriced subsidy for adjacent functions. Nor should the additional function be treated as an afterthought. The architecture needs clear separation where separation is required, and clear integration where integration creates value.

The economics of hybrid systems are strongest where there is a real overlap in infrastructure needs. A route that already requires marine survey, permitting, landing access, power, monitoring and maintenance may offer a lower-cost platform for additional sensing or observation than a separate dedicated deployment. The value comes from shared fixed costs and shared operational presence.

However, shared cost is not the same as shared value. The additional function needs a customer, a budget and a decision-maker. It may be a government agency, a research institution, an energy operator, a port authority, an environmental body or a consortium of users. Each has different procurement cycles, risk tolerances and expectations around data access.

The commercial model must therefore be designed around demand, not around the novelty of the technology. Is the additional service sold as a subscription, a data product, a public-service capability, a co-funded asset or a strategic option? Can it be contracted independently? Can it be scaled? And can the revenue support the incremental complexity over the life of the system?

The most difficult part of a hybrid system may not be the cable. It may be governance. Multiple functions often mean multiple stakeholders: cable owners, operators, governments, research bodies, equipment suppliers, data users and maintenance providers. Without a clear governance model, the system can become slow to operate and difficult to commercialise.

Governance should define decision rights, data rights, security responsibilities, cost allocation, maintenance priorities and escalation paths. It should also define what happens when the system needs to change. Hybrid infrastructure will operate across long asset lives, while the technologies and institutions around it will evolve. Contracts need enough precision to protect the parties and enough flexibility to accommodate new uses.

This is not administrative detail. It is part of the investment case. A technically elegant system with unclear governance can struggle to attract capital because the risks cannot be priced. Conversely, a well-governed model can make a complex system investable by showing how responsibilities and returns are allocated.

Hybrid systems often create the prospect of new data streams. But data becomes a product only when it is reliable, interpretable, accessible and connected to a user decision. Raw data without a route to use is not a revenue model.

The commercial opportunity may sit in processing, alerts, analytics, integration into operational workflows or long-term datasets rather than in the data feed itself. That requires product management: defining the user, the use case, the service level, the interface and the price. It also requires a realistic view of who will own those capabilities. A cable operator may not be the best organisation to commercialise every data product alone.

Partnerships can be decisive here. The right partner may bring domain expertise, distribution, credibility with end users or an existing platform. The cable system provides a unique input; the partner turns that input into a service customers can adopt. The commercial architecture should reward both roles without obscuring accountability.

There is a risk in assuming that every function should be integrated because the infrastructure is shared. Sometimes the best design is modular: a communications system with defined interfaces that allow additional capabilities to be added, upgraded or operated separately. Modularity can preserve optionality and reduce the risk that one function constrains another.

This matters especially when demand is uncertain. A system can be designed to be hybrid-ready without committing all capital and operational complexity on day one. That approach allows the core communications business to proceed while preserving a credible path for additional functions when customers, regulation and partners are ready.

The test is whether integration reduces total system cost and improves customer value, not whether it produces the most impressive technical diagram. Commercial discipline means being willing to leave value on the table today if doing so protects a larger and more durable opportunity tomorrow.

Hybrid cable systems can be an important part of the next generation of digital and maritime infrastructure. They can help make corridors more useful, create new forms of public and private value and improve the economics of strategic investment. But they will not succeed through engineering alone.

They require a business model that protects the primary service, identifies real demand for the additional function, allocates governance clearly and turns data into usable products. They require investors and operators to distinguish between shared cost and shared value. And they require partnerships that are designed for the life of the asset, not just the construction phase.

Double the function does not automatically mean half the price. In the right circumstances, it can mean a more valuable, more resilient and more strategically relevant system. The work is to make that value visible, contractible and investable.

AI is often discussed as a compute story. It is a compute story, but it is also a network story. Training, inference, data movement, model distribution, cloud access and enterprise adoption are changing the volume, direction and performance requirements of traffic across Asia. The result is not simply more demand. It is a different shape of demand.

That distinction matters. Capacity planning built on yesterday’s internet patterns can miss where value will be created next. AI workloads are concentrated in some places, distributed in others and increasingly connected to data-centre clusters, cloud regions, enterprise campuses and edge environments. They create new dependencies between power, compute, storage and connectivity.

The question for the industry is therefore not whether AI will increase bandwidth demand. It will. The more useful question is where capacity will be needed, what form it must take and which participants will capture the commercial value of providing it.

Asia’s opportunity is large precisely because it is diverse. The region contains mature digital hubs, fast-growing consumer markets, manufacturing centres, financial centres, emerging cloud regions and countries with very different regulatory, energy and connectivity profiles. AI will not create a single regional capacity map. It will create a set of interconnected but distinct demand corridors.

Some markets will concentrate frontier training and large-scale cloud infrastructure. Others will become inference and application hubs, closer to enterprise users, data sources and local language markets. Some will benefit from their role as transit, landing or interconnection locations. Others will need stronger domestic fibre and data-centre ecosystems before they can capture more of the value chain.

For operators and investors, the implication is clear: regional narratives are not enough. Capacity decisions need to be grounded in specific demand clusters, customer segments, power availability, data-centre roadmaps, cloud strategies and cross-border routes. The opportunity is in the connections between these elements.

The data-centre market has rightly focused on power. AI increases the importance of power density, grid access and energy strategy. But megawatts alone do not define the quality of an AI infrastructure opportunity. Network adjacency matters just as much.

A data-centre campus with strong access to multiple carriers, cloud on-ramps, internet exchanges, terrestrial routes and international capacity can support a broader range of customers and services than one with power but limited connectivity. The same is true in reverse: a network corridor that reaches a growing compute cluster can become more valuable when it is integrated into the data-centre and cloud ecosystem around it.

This is why the next capacity map should be drawn as an ecosystem map. It should show power, land, data centres, cloud regions, subsea landings, terrestrial fibre, exchanges, enterprise demand and regulatory boundaries. The commercial question is where these assets reinforce one another and where a missing link constrains the whole system.

AI workloads are not uniform. Training can require large movements of data and high-capacity links between compute clusters, storage environments and research or platform locations. Inference can be more distributed, especially where applications need low latency, local data handling or proximity to users. Model updates, data replication and operational telemetry add further layers of traffic.

That means capacity products will need to evolve. Some customers will require very high-capacity, predictable inter-data-centre connectivity. Others will value flexible, scalable access to cloud and exchange ecosystems. Some will need protected routes and strong service assurance because the workload is business critical. Others will prioritise rapid provisioning and the ability to scale with uncertain demand.

The winners will not simply sell more bandwidth. They will understand the workload, the customer decision and the operational consequence of performance. They will package capacity in ways that make it easier for customers to deploy AI services across multiple locations without having to assemble every network component themselves.

AI will increase the strategic importance of cross-border capacity in Asia. Data, models, applications and cloud services will move across national boundaries even as data sovereignty, security and regulatory requirements become more complex. The result will be a premium on routes that are diverse, well-governed and connected into the right digital ecosystems.

This does not mean every workload will move freely. In many cases, data will need to remain local while models, software, metadata or aggregated outputs move across borders. That creates a more nuanced demand profile. Networks will need to support both regional scale and local control.

For subsea and terrestrial operators, the opportunity is to provide more than a route. It is to provide confidence: confidence in availability, compliance, operational visibility and the ability to connect into the cloud, data-centre and partner ecosystems customers already use. That confidence can become a differentiator when capacity is abundant in theory but difficult to use in practice.

The next capacity opportunities are likely to emerge at the intersections of three forces. First, where compute and data-centre development is accelerating. Second, where enterprise and consumer demand is large enough to support local AI applications. Third, where connectivity is currently a constraint on the ability to link those markets to regional and global platforms.

That points toward a portfolio of opportunities rather than a single destination. It includes strengthening established hubs with more diverse international and metro connectivity. It includes building resilient corridors into emerging cloud and data-centre markets. It includes improving domestic fibre and interconnection so that local demand can reach AI infrastructure efficiently. And it includes creating new partnerships between network operators, data-centre providers, cloud platforms and enterprises.

The discipline is to avoid chasing headlines. AI demand forecasts can be impressive, but capacity investments need a commercial pathway. Who will buy? What product will they buy? What alternatives do they have? How quickly can demand be converted into contracted or repeatable revenue? And what does the asset need to connect to in order to become part of the customer’s operating environment?

AI is rewiring Asia’s networks because it is rewiring the relationship between compute, data, applications and users. The infrastructure response cannot be limited to adding capacity where traffic has historically been highest. It must identify where new digital ecosystems are forming and what connectivity they will require to function.

That requires a commercial map before a capacity map. The commercial map identifies customers, workloads, partners, products, regulatory constraints and routes to market. The capacity map then follows: subsea systems, terrestrial fibre, interconnection, data-centre connectivity and cloud access designed around real demand.

Asia has the scale, diversity and digital momentum to create many of the world’s most important AI infrastructure corridors. The opportunity for the industry is to make those corridors investable and usable. The question is not simply where to put more capacity. It is where capacity can become part of a durable commercial system that helps the region’s next generation of digital services grow.