Data centers struggle to maximize GPU ROI

Infrastructure managers are underutilizing costly GPUs due to inefficient resource allocation and a lack of automation tools, hindering their AI return on investment (ROI), reports Fierce Networks. According to Rafay Systems CEO Haseeb Budhani, many organizations rely on manual processes that leave GPUs idle, with nearly one-third of enterprises utilizing less than fifteen percent of their GPU capacity. Solutions like GPU self-service platforms could take years to implement, exacerbating delays in AI deployment.

Beyond AI, GPUs are also driving ROI in high-performance computing (HPC) and big data applications. By optimizing infrastructure, organizations can use AI to connect and make sense of unstructured data, previously a major challenge, notes Hammerspace CMO Molly Presley. Such data harmonization and analytics are central to several of the newer CXO titles, including CMO and CSO.

Telcos are entering the GPU market with services like GPU clouds to compete with hyperscalers, offering enhanced flexibility for AI and non-AI workloads. IBM and AMD’s upcoming GPU-as-a-service collaboration on IBM Cloud exemplifies efforts to deliver efficient, scalable GPU solutions.