At this year’s International Telecoms Week (ITW), one message came through clearly during the keynote “More fiber? Better fiber? Meeting growing demand cost-effectively & securely” panel: the industry is no longer simply building for bandwidth growth. It is building for an entirely new era of infrastructure demand.
The panel brought together leaders from across the subsea, connectivity, and infrastructure ecosystem, including Allen Meeks, CEO of MOX Networks, alongside executives from C3NTRO Telecom, Alcatel Submarine Networks, and Arcadian.
The discussion explored how AI, hyperscale infrastructure, geopolitical uncertainty, and security concerns are fundamentally changing how networks are designed, financed, and deployed.
AI Demand Is Accelerating Faster Than the Industry Has Seen Before
A recurring theme throughout the discussion was the extraordinary pace of AI-driven infrastructure demand.
As one speaker noted, the industry is now operating on “accelerating timescales, because everybody wants it tomorrow”. What previously took years to plan is now being demanded almost immediately, even as supply chains, permitting, construction timelines, and subsea deployment remain inherently long-term challenges.
For Meeks, this is not theoretical. MOX Networks was effectively built around high-volume AI data movement long before the current AI boom.
During the panel, Meeks explained how MOX originated more than a decade ago from the healthcare and life sciences sector, where the challenge of moving massive genomic and proteomic datasets required infrastructure capable of handling AI-scale workloads well before the wider market caught up.
Rather than reacting to today’s AI surge, MOX was architected around these scale challenges years ago. “We were trying to future-proof against scale issues that we were running into 13 years ago,” Meeks explained during the session.
That background gives MOX a unique perspective on what modern AI infrastructure actually requires. According to Meeks, the challenge is no longer just about adding more bandwidth. It is about ownership, scale, seamless interconnection, and removing operational complexity from increasingly global infrastructure environments.
Over the past several years, MOX has expanded beyond its original healthcare roots, investing heavily into long-haul fiber infrastructure across North America and strategic subsea connectivity into Asia and South America. The goal, as Meeks described, is to create seamless, large-scale connectivity without relying on fragmented chains of providers.
Diversity No Longer Means What It Used To
Another major theme from the panel was the evolving definition of network diversity.
Historically, diversity primarily meant geographical redundancy. Today, the discussion has become far broader. Panelists discussed the growing need for jurisdictional diversity, equipment diversity, manufacturing resilience, and open network architectures that avoid dependence on any single vendor, supplier, or geopolitical region.
This reflects a wider industry shift. As AI infrastructure becomes increasingly strategic, operators are being forced to think beyond traditional fiber routes and toward long-term operational resilience.
Speakers highlighted how hyperscalers and AI infrastructure providers are now demanding entirely new approaches to network design. The challenge is compounded by supply chain pressures, including 12-to-14-month lead times for transport systems and significant increases in fiber demand globally.
Several panelists echoed Meeks’ view that AI infrastructure cannot simply be treated as an extension of the cloud era. One particularly important point raised during the session was that many current networks are still not truly optimized for inference workloads. As Meeks bluntly noted during the discussion: “Nothing is at this point.”
That distinction matters. Training AI models has already driven huge infrastructure investment. However, inference, the real-time deployment and usage of AI applications at scale, could create even larger and more distributed networking demands over the coming years.
The industry, as several speakers acknowledged, still does not fully understand what the long-term capacity requirements of inference will ultimately look like. As a result, operators are building aggressively now to avoid future constraints later.
Security and National Resilience Are Moving Center Stage
Beyond AI, the panel also highlighted how national security and geopolitical concerns are increasingly influencing network infrastructure strategies.
Discussions ranged from supply chain independence and manufacturing resilience to concerns around protecting critical infrastructure from emerging threats.
One speaker noted that many national security networks still rely heavily on commercial infrastructure originally designed decades ago for very different threat environments. Another described the broader move away from globalization toward more localized and strategically independent infrastructure ecosystems.
For operators, this creates additional complexity around where infrastructure is built, how routes are diversified, and how critical systems are protected.
It also creates new opportunities. As industrial reshoring, advanced manufacturing, and AI facilities move into more remote regions, entirely new connectivity routes will be required. While infrastructure investment must still link major cities or hyperscale campuses, it is increasingly about enabling the next generation of digital and industrial infrastructure wherever it emerges.
Building the Infrastructure Behind the AI Economy
One of the clearest conclusions from the session was that the telecom and digital infrastructure industries are entering a new phase of evolution.
The conversation is not limited to capacity growth. It is focused on creating infrastructure ecosystems capable of supporting AI-scale compute, distributed inference, geopolitical resilience, and increasingly complex operational environments simultaneously.
For MOX Networks, this moment aligns closely with the company’s long-term strategy.
While many organizations are only now beginning to confront the realities of AI-driven infrastructure demand, MOX has spent years building networks capable of transporting enormous volumes of mission-critical data across healthcare, research, cloud, and enterprise environments.
As the panel demonstrated, the next era of infrastructure is not limited to fiber. It will require smarter architectures, resilient supply chains, diversified routes, and networks designed from the ground up for the demands of the AI economy.
