Telcom operators and cloud providers are struggling to keep pace with the infrastructure demands from the rapid adoption of AI, and they should shift from simply expanding to optimising their setups, a new report by Keysight Technologies suggests.
In a study of more than 100 industry leaders in the sector, it found that 89 per cent of respondents intend to expand or maintain AI infrastructure investments in the coming year. This is driven by factors such as cloud integration (51 per cent), faster graphics processors (49 per cent), and high-speed network upgrades (45 per cent).
In addition, 62 percent of respondents plan to maximise current infrastructure without new capital spending. This would involve tools such as real-world AI workload emulation, which can help with performance validation, eke efficiency gains and enable faster deployment of next-generation AI clusters.
With the rising infrastructure pressures, 50 per cent of operators surveyed say their biggest challenges when it comes to scaling AI include: budget constraints (59 per cent), infrastructure limitations (55 per cent), and talent shortages (51 per cent).
Another factor impacting AI scalability is network capacity, with 55 per cent of operators deploying 400G technology in their networks. Yet, operators are also showing a growing interest in 1.6T networking, which is still emerging but promises much higher speeds.
At the same time, high-speed networking is on the rise, as the adoption of advanced networking technology is accelerating, with 34 per cent of respondents investigating 800G, 22 per cent testing 1.6T, and 58 per cent considering Ultra Ethernet as a high-performance networking choice.
Keysight, a testing and measurements company, had conducted the survey of the industry leaders globally with research outfit Heavy Reading earlier this year.
Another finding is that real-world workload emulation is critical, according to 95 per cent of respondents. However, many do not have the tools to effectively simulate production-scale AI environments.
AI success depends on optimising every layer of the network, said Ram Periakaruppan, vice-president and general manager for network applications and security at Keysight.
“AI data centres are reaching a tipping point where performance and scale alone are not enough,” he noted. “Operators need deeper insight, tighter validation, and smarter infrastructure choices.”
