Powering the AI data center boom: the infrastructure upgrades behind innovation (www.techradar.com)

As the demand for generative AI and large-scale model training surges, data center operators are rapidly modernizing their infrastructure to support advanced GPU technologies. This transformation involves significant improvements in power, cooling, and connectivity, as traditional systems struggle to meet the escalating requirements of AI workloads. McKinsey forecasts that data center spending could reach $6.7 trillion by 2030, predominantly directed toward AI-focused facilities. However, the growth is tempered by supply chain bottlenecks for key components, design limitations that restrict operational density, and a shortage of skilled engineers to manage these complex builds. In response to this evolving landscape, "neocloud" providers have emerged, specifically designed around high-performance GPU compute without the constraints of traditional cloud demands. Companies like CoreWeave are rapidly scaling up their GPU deployments, significantly enhancing performance and energy efficiency. The shift towards liquid cooling systems and advanced interconnect technologies highlights the increasing complexity of AI infrastructure, necessitating specialized expertise and strong partnerships to navigate engineering and operational challenges. As countries recognize AI infrastructure as vital for long-term competitiveness, those that streamline their data center capabilities will attract investment and talent, reflecting a broader strategic race in the AI sector.