Optical Fiber Demand Surges as Data Centers Expand

AI infrastructure is shifting toward a new bottleneck as demand for optical fiber accelerates to connect increasingly large GPU clusters. Industry experts warn that while there is no fiber shortage now, rising AI investment, expanding data center capacity, and a lack of specialized talent could make optical connectivity a strategic resource for future AI growth.

In the last years, the race to scale AI has largely centered on GPUs and high-bandwidth memory. Now, another critical piece of infrastructure is emerging as a strategic priority: optical fiber.

As AI models become larger and more computationally demanding, data centers require increasingly dense, high-speed optical networks capable of moving data among tens of thousands of GPUs. Industry leaders say optical connectivity is becoming as essential to AI performance as processing power itself, positioning fiber as the next infrastructure challenge for hyperscalers and cloud providers.

"Behind every AI model and every automated decision there is an invisible infrastructure that makes it possible, which is optical fiber. Without that foundation there is no true digital transformation," says Carlos Zúñiga, Southern Cone Sales Manager, Corning Optical Communications.

The conversation around AI infrastructure has so far focused on semiconductor availability and memory. GPU manufacturers continue to face strong demand, while Samsung, SK Hynix, and Micron have redirected production toward high-bandwidth memory required by AI data centers, increasing costs across multiple consumer electronics segments.

However, processors alone cannot sustain AI workloads. Without high-capacity optical networks, GPUs spend more time waiting for data than performing calculations, reducing system efficiency and increasing operational costs.

As organizations continue expanding AI computing capacity, demand for fiber infrastructure is growing alongside investments in compute resources.

AI's Next Bottleneck Is Not Processing Power, but Connectivity

The trend reflects broader changes across the global data center industry. According to the International Energy Agency, electricity demand from data centers increased 17% in 2025, while consumption from AI-focused facilities surged 50%. The agency projects total data center electricity demand will double by 2030, while AI infrastructure demand will triple over the same period.

At the same time, capital expenditures by five of the world's largest technology companies exceeded US$400 billion in 2025 and are expected to increase another 75% in 2026. Those investments are competing simultaneously for power capacity, semiconductor supply, optical fiber and specialized technical talent.

Hyperscalers Are Racing to Secure Fiber Capacity

For Corning Optical Communications, those market dynamics are already translating into business growth. Zúñiga says the company's optical communications division has become its largest business unit, increasing annual corporate revenue from approximately US$12 billion to nearly US$40 billion.

According to Zúñiga, GPU manufacturers, including Nvidia, alongside hyperscale cloud operators, have begun securing long-term fiber supply agreements much as they previously did with memory components. "The need to secure optical fiber supply responds to the increasing complexity of AI models," he says.

The next phase of infrastructure development extends beyond deploying additional fiber between servers and data centers. The industry is now moving toward bringing optical connectivity directly to processors through photonic technologies designed to accelerate data transfer, reduce energy consumption, and eliminate emerging performance bottlenecks.

Despite rising demand, Zúñiga says the market is not experiencing an actual shortage of optical fiber. "There is no fiber shortage as such today. What exists is pressure and a future challenge," says Zúñiga, while acknowledging that delivery timelines for some projects are beginning to tighten.

Any supply constraints could have implications well beyond AI infrastructure. Optical fiber also supports residential broadband, telecommunications networks, hospitals, and critical public services, making it a foundational component of the broader digital economy.

Latin America's Next AI Challenge Is Building the Right Infrastructure

Latin America's infrastructure landscape reflects uneven development. Mexico, Brazil, and Chile account for the region's three primary data center hubs, with Mexico and Brazil together representing nearly 60% of installed capacity.

However, expanding computing infrastructure is not always matched by the connectivity needed to support it. Zúñiga says many large international operators build facilities on the outskirts of major cities, where land is available but high-capacity fiber infrastructure is often limited.

Another emerging challenge is workforce availability. Unlike the semiconductor industry, where manufacturing capacity defines production limits, optical fiber deployment increasingly depends on technicians capable of splicing next-generation fiber systems.

That expertise became less common as fiber evolved into a standardized commodity during the rapid expansion of fiber-to-the-home networks. AI data centers are replacing traditional 288-fiber backbone cables with ribbon cables containing as many as 2,880 fibers, while some projects in the United States and Latin America are already deploying links with up to 6,912 fibers.

Connecting infrastructure at that scale requires highly specialized precision, reviving technical skills that the industry largely stopped developing when fiber deployment became routine.

The shortage of skilled technicians also affects network latency, limiting applications that require real-time responsiveness. Zúñiga points to robotic surgery as an example, noting that while the underlying medical technology already exists, communications infrastructure still cannot consistently deliver the response times required for remote procedures across countries.

As AI infrastructure expands, network quality is becoming an increasingly important factor in determining where future computing investments will be located. Zúñiga says the availability of next-generation optical networks will play a growing role in site selection for new AI data centers, particularly in Latin America.