Riyadh, KSA: A global survey commissioned by Ciena sheds light on the increasing demands AI workloads will place on data center infrastructure. With more than 1,300 data center decision-makers across 13 countries participating, the survey reveals that the next five years will see at least a 6X increase in interconnect bandwidth needs.
AI workloads are expected to drive the highest demand for data center interconnect (DCI) infrastructure, surpassing cloud computing and big data analytics. According to the findings, 43% of new data center facilities will be dedicated to AI workloads, as AI model training and inference require massive amounts of data movement. Fiber optic capacity is a key factor, with 87% of respondents indicating they will need 800 Gb/s or higher per wavelength to meet these demands.
Jürgen Hatheier, Chief Technology Officer at Ciena, emphasized that AI is poised to accelerate network traffic growth significantly, pushing operators to rethink their network architectures. “AI is not just about compute; it’s about connectivity,” said Hatheier, stressing the need for robust infrastructure to support AI’s potential.
Sustainability is also a growing priority, with 98% of respondents highlighting the importance of pluggable optics to reduce power consumption and network footprint. As AI compute requirements continue to increase, 81% of participants predict the training of large AI models (LLMs) will be distributed across multiple data centers, necessitating interconnected DCI solutions.
The survey also found that 63% of experts prioritize AI resource utilization, followed by reducing latency by placing inference compute closer to users (56%), and data sovereignty requirements (54%). A majority (67%) of respondents favor Managed Optical Fiber Networks (MOFN) for long-haul data center connectivity, indicating a shift from traditional dark fiber deployments.
As the AI revolution continues, the survey underscores the critical role of advanced networking solutions to unlock AI’s full potential. Ciena’s research shows that data center operators must adapt their infrastructure to meet the rapidly growing bandwidth demands driven by AI.
