Keysight has launched the “Beyond the Bottleneck: AI Cluster Networking Report 2025”
As artificial intelligence accelerates across industries, the demands on data centre infrastructure are rising fast
Keysight Technologies, Inc. in collaboration with Heavy Reading, has released the Beyond the Bottleneck: AI Cluster Networking Report 2025, revealing that artificial intelligence (AI) adoption is accelerating faster than infrastructure can keep up. The global study highlights the urgent need for telecom and cloud providers to shift from expansion to optimisation in order to support next-generation AI workloads.
As artificial intelligence accelerates across industries, the demands on data centre infrastructure are rising fast. Building out infrastructure is no longer enough on its own; 62% of respondents said they aim to get more out of their infrastructure without new investment. Operators are turning to performance optimisation strategies, such as real-world AI workload emulation, not only to validate performance but also to enhance efficiency and accelerate the deployment of next-gen AI clusters.
According to the report, nearly 89% of respondents plan to expand or maintain AI infrastructure investments in the coming year. The top drivers of this growth include cloud integration (51%), faster GPUs (49%), and high-speed network upgrades (45%).
Key Findings from the Report Include:
- Optimisation First Approach: While investment continues, 62% of operators say they aim to extract more value from existing infrastructure without new capital expenditures.
- Emulation Becomes Essential: A resounding 95% of respondents said real-world workload emulation is critical, yet many lack the tools to simulate production-scale AI environments effectively.
- Rising Infrastructure Pressures: Over 50% of operators report budget constraints (59%), infrastructure limitations (55%), and talent shortages (51%) as their biggest barriers to scaling AI.
- High-Speed Networking on the Rise: Adoption of advanced networking technologies is gaining momentum, with 34% exploring 800G, 22% trialling 1.6T, and 58% evaluating Ultra Ethernet as a high-performance networking option.
- Network Bottlenecks Are the Next Frontier: With 55% of operators deploying 400G interconnects and a growing interest in 1.6T, network capacity is emerging as a decisive factor in AI scalability.
The study illustrates a major industry shift: infrastructure evolution is no longer just about capacity—it’s about efficiency, performance, and reliability. As AI models grow more sophisticated, tools like real-world AI workload emulation are critical for maximising infrastructure potential while controlling costs.
“AI data centres are reaching a tipping point where performance and scale alone are not enough. Operators need deeper insight, tighter validation, and smarter infrastructure choices,” said Ram Periakaruppan, vice president and general manager, Network Applications & Security Group at Keysight. “This research confirms what we see in the field: success in the AI era hinges on optimising every layer of the network. Keysight is proud to support this shift with solutions that enable providers to emulate, validate, and future-proof their AI infrastructure at scale.”
About the Survey
The AI Cluster Networking Report 2025 is based on a global survey conducted by Heavy Reading from March to April 2025. Respondents included professionals from leading telecom and cloud service providers across the Americas, Asia Pacific, and Europe.
For more news visit: News – CIE
