Moderated by Julian Rawle (Cambridge Management Consulting), the session featured Eduardo Mateo (NEC), Budi Satria Dharma Purba (Telin), and Matthew Goh (KKR). What followed was one of the conference’s most candid and technically grounded discussions.

Demand keeps outrunning supply

The panel opened with a pointed question from Rawle: is the industry over-building, or is demand still accelerating?

For Telin’s Budi Satria Dharma Purba, the answer was unequivocal. With Asia–Europe capacity under pressure and new builds hitting geopolitical roadblocks, he argued the industry is far from meeting future needs. “Even without AI, diversity and redundancy are driving massive subsea demand today,” he said.

Purba highlighted how key routes remain congested or vulnerable. The Red Sea situation has delayed major systems, while intra-Asia links such as SJC1 were saturated before upgrades arrived. Even with new systems like SJC2 and ADC coming online, he warned that “there will be a lot of challenge now in terms of how the supply can meet the demand.”

He cited research showing intra-Asia requirements potentially exceeding supply by 2–3x by 2028, with transpacific demand possibly reaching 4x.

From the investment side, Matthew Goh of KKR echoed these concerns: “In the mid-term, there’s still plenty of demand for cables, supply has to catch up.” Resilience, national security, and new data-intensive digital services are driving countries to rethink their subsea strategies. For Goh, the trend is clear: networks need more paths, more redundancy, and more protection against chokepoints.

Technology isn’t the constraint, delivery is

While demand surges, submarine cable suppliers face intense pressure to deliver. NEC’s Eduardo Mateo, the panel’s technologist, emphasised that the bottleneck isn’t optical innovation. “Modern transponders operate at fundamental limits, the fibres you lay today will remain relevant for the next decade,” he explained. SDM-based systems now enable petabit-scale cables, and suppliers are expanding manufacturing capacity through automation and standardisation.

But the barriers to delivery lie elsewhere.

“The challenge isn’t technology, it’s delivery: vessels, permitting, and human expertise,” Mateo said. Vessel availability is especially tight, with an aging global fleet and limited weather windows for long transpacific operations. Permitting delays can freeze projects for years.

Purba reinforced this with a real-world example: when Telin sought supplier support for 25,000 km of new builds, he was told production lines were already committed. The answer was blunt: wait up to three years, or consider direct appointments. With more than 100 new systems globally in the pipeline, he described supply-chain congestion as “a big concern” across the GLF carrier community.

Private capital rethinks the vessel equation

Goh noted that KKR is actively responding to these constraints through its investments in Optical Marine Services (OMS). While he deferred operational details to OMS itself, he highlighted broader dynamics: the industry previously avoided vessel overbuilds due to painful lessons from the early 2000s, but the next decade’s demand profile is different. Long-term redundancy needs, driven by cloud, geopolitics, and AI-adjacent data flows,  justify new tonnage and expanded operational capacity.

The AI wildcard: A driver or not (yet)?

AI’s role in subsea demand, possibly the hottest question in telecoms, drew nuanced answers from all three panellists.

Mateo captured the uncertainty best: “AI’s impact on subsea cables? Depends on business models, regulations, and how globally distributed networks will evolve.” The location of future AI training clusters, regulatory limits on cross-border data movement, and the architecture of hyperscaler inference nodes will all determine long-term capacity patterns.

Purba agreed with the ambiguity, but stressed that some hyperscalers are already shifting toward globally distributed AI architectures, a trend that increases the need for inter-regional capacity. Demand from these players, he noted, is “real now,” with customers requesting multiple terabit-class wavelengths across diverse routes.

Goh, however, urged grounding in present realities. Today’s subsea boom, he argued, remains driven by classic cloud, content distribution, and cyber-resilience needs, not yet by AI training workloads. “I don’t think subsea is even necessarily driven entirely by AI today… it’s just by diversity, just regular stuff we do.” That may change, he acknowledged, as global inference networks scale.

Planning for a moving target

One recurring theme was the mismatch between investment timelines and market evolution. As Purba put it: “Building a cable today takes 4–5 years by the time it’s ready, market dynamics may have shifted.” The industry has repeatedly underestimated demand, leading to buyer’s remorse for systems that seemed oversized on day one but obsolete in capacity terms within one or two years.

Mateo offered reassurance that technological obsolescence is less of a threat today: optical physics has approached fundamental limits, meaning new transponders will remain compatible with today’s fibre for many years.

RELATED STORIES

How Asia’s subsea cables are changing connectivity logic

ITW Asia 2025: Singtel’s CTO on rewiring Asia’s digital infrastructure for the AI era