DIGITAL TRANSFORMATION
Inside the data centre, those demands are accelerating quickly, with optical interfaces projected to grow from about 59 million ports per year in 2025 to 325 million by 2030, according to market research firm LightCounting.
To help with this, Nokia is developing the data centre switching, routing and optical technologies needed to build those fabrics, enabling intelligence to move securely, efficiently and at the scale the AI era demands.
“ AI workloads place very different demands on the network than traditional cloud applications,” David adds.“ Training requires massive east-west bandwidth, low latency and lossless performance, as well as the ability to scale without adding unnecessary cost, power or complexity.”
Why AI needs advanced networks This architectural shift has sparked what industry insiders call the“ AI supercycle”, an investment loop where hyperscalers like Amazon, Google Cloud and Microsoft build gigawatt-scale data centres that require unprecedented physical and optical capacity to function as unified systems.
“ AI cannot scale without advanced networks,” David says.“ Compute may create intelligence, but the network determines how quickly, efficiently and reliably that intelligence can move between compute resources, across data centres and out to users and AI-enabled machines. As reasoning capability compounds quickly, the real question becomes how to connect enough compute, memory and power to keep up.
“THE BIGGEST OPERATIONAL RISK FOR DIGITAL INFRASTRUCTURE PROVIDERS IS LACK OF SUPPLY”
Ciarán Delaney Chief Operating Officer EXA Infrastructure
“ There is no one-size-fits-all architecture, as the network must be optimised for the application, reach, power envelope and operating model. For many customers, the next phase is moving beyond the limits of any single site and connecting distributed capacity as one larger AI system.”
This is why Nokia works closely with AI and cloud providers on a co-creation model that brings together switching, IP, optical, silicon, systems and automation to help them scale AI infrastructure with the performance, efficiency and operational simplicity they need.
“ As a highly vertically integrated supplier with in-house design and semiconductor fabrication capabilities across a broad range of networking solutions, Nokia is in a unique position to co-design, develop and manufacture solutions to meet specific hyperscaler needs,” David says.
90 October 2026