The Future of AI Computing: Insights from Data Center World 2025
During the recent Data Center World 2025 event, Manoj Sukumaran from Omdia provided valuable insights into the ongoing AI computing revolution. Drawing parallels to the evolution of mobile networks, Sukumaran likened the current state of AI technology to the 2G era, indicating that we are just scratching the surface of what is possible.
One key point Sukumaran emphasized was the projected $1 trillion global data center capex by 2030. He highlighted the need for a complete infrastructure overhaul to accommodate the demands of large language models, as current computing systems are not optimized for this purpose. Sukumaran also pointed out innovative solutions being explored in the industry, including the potential adaptation of ship engines for data center use.
Another critical challenge discussed by Sukumaran was the issue of data movement, which currently accounts for 60% of energy consumption in AI systems. To address this, next-generation chips like Nvidia’s upcoming Rubin Ultra are integrating compute and memory elements in single packages, aiming to improve efficiency and performance.
Despite the significant challenges posed by power consumption, cooling, and infrastructure, Sukumaran remains optimistic about the future of AI computing. He believes that a combination of innovation and resources will drive solutions forward as the industry scales from today’s 100,000-GPU clusters to anticipated million-scale deployments.
For those interested in learning more about Sukumaran’s insights, the full interview can be watched here.
