AI Ignites a Boom in Optical Networking as Data Centers Race to Meet Surging Demands

AI is Making Networking ‘Sexy’ Again

Optical Networking Innovation Accelerates

Artificial intelligence is significantly impacting networks, prompting data center operators and network owners to rethink their current technologies. The expected surge in AI-driven networking demands is still somewhat unclear but widely acknowledged to be substantial, requiring considerable investment.

Most of the early AI-related traffic originates within data center fabrics—the interconnected links that connect thousands of compute units handling AI workloads. This spike in demand is fueling investments in new data centers and upgrades to existing facilities.

Industry experts highlight AI as the main driver behind increasing network capacity across hyperscalers, service providers, and enterprise networks alike. For example, global data center capital expenditures rose 51% in 2024, reaching $455 billion, with hyperscalers like AWS, Microsoft, and Google Cloud investing heavily in custom accelerators to support AI training.

Currently, AI workloads account for about 15% of total data center activity, but projections suggest this could rise to 40% by 2030. This growth is expected to double sales for Ethernet optical transceivers used in AI clusters.

AI’s Influence on Inter-Data Center Networking

Beyond intra-data center traffic, there is growing demand to enhance inter-data center optical connections. These links, which connect large data centers across regions or edge locations closer to users, are crucial for AI data inferencing.

Improving these inter-data center connections involves complex equipment and less controlled environments. Market leaders such as Huawei, Ciena, and Nokia are advancing higher-speed fiber optic and transceiver technologies capable of speeds up to 1.6 Tb/s.

The optical transport market showed strong growth at the end of 2024, signaling recovery from previous revenue contractions. Industry leaders see this as a sign that supply and demand in optical transport equipment are balancing out.

Networking is regaining attention as a critical part of AI infrastructure. Hyperscalers are focusing not only on the fastest GPUs and TPUs but also on building sophisticated networking ecosystems within data centers. Optical innovations are especially vibrant, with daily advancements pushing the technology forward.

Balancing Innovation with Practical Challenges

While AI compute generates vast amounts of data, this data must be efficiently transferred from hyperscale data centers to enterprises and consumers. The cost of AI training is high, but inference—the use of trained models for real-world applications—is where revenue is expected to grow.

Data center operators face challenges such as power constraints and physical space limits, leading to the distribution of data centers across multiple locations. This creates a need for new optical technologies that provide low-cost, low-latency interconnections over distances up to 20 kilometers.

Reliability is also crucial. Hyperscalers prioritize stable, resilient network connections that minimize failures, supporting the demanding requirements of generative AI workloads.

Networking Opportunities for Operators

Network operators are rethinking their business models to capitalize on AI-driven demand for higher throughput and lower latency. For instance, Lumen Technologies is investing heavily in building AI-specific networking fabrics to connect major data centers optimized for AI workloads.

Lumen has already secured $5 billion in sales from hyperscalers and large tech companies, with expectations for an additional $7 billion pipeline. The next wave of demand is anticipated from enterprises adopting AI inference for business transformation in sectors like finance, healthcare, and retail.

Hybrid architectures combining on-premises data with cloud-based AI models will require massive bandwidth and tight latency controls. Specialized connectivity fabrics will be essential to meet these needs.

Telecom operators also have a unique advantage given their geographic reach. They can support AI inferencing closer to where data is generated, including on-premises at large enterprises or at the network edge near customers.

Partnerships like Verizon’s deal with Nvidia demonstrate how carriers are integrating AI to enhance enterprise services using 5G private networks and mobile edge computing. Similarly, T-Mobile US is transforming cell sites into edge data centers to better handle AI workloads.

These developments point to a significant expansion in internet capacity and AI-driven networking demands, requiring continuous investment and innovation.

Conclusion

AI is reshaping networking priorities, pushing optical technologies and network infrastructures into the spotlight. Data centers and network operators must adapt to handle growing AI workloads both within data centers and across wider networks. This shift presents both challenges and opportunities for those building the future of connectivity.

For those interested in expanding their AI skills and understanding its impact on networking, exploring specialized AI courses can provide valuable insights.