Bernstein has forecasted significant growth in the global server market, predicting it will reach a value of $1 trillion by 2028. They project global server and GPU AI server shipments will experience compound annual growth rates of 15% and 22%, respectively, with 8-GPU-equivalent server shipments expected to increase by 48% in 2026. This growth is driven by planned and under-construction data center investments totaling approximately $1.2 trillion, as major cloud providers are enhancing their capacities to meet increasing AI workloads. Investment in custom ASICs is also rising, as hyperscalers adopt these technologies alongside merchant GPUs to optimize AI performance.
AMD: AMD develops processors, GPUs, and accelerators optimized for servers and AI applications. Bernstein’s analysis includes AMD’s Helios racks among the products slated for Q4 shipments, positioning the company within the expanding AI infrastructure ecosystem.
Rubin: Rubin is NVIDIA’s next-generation GPU architecture designed for advanced AI training and inference. According to the Bernstein report, Rubin-based racks are projected to begin shipping in Q4 as part of broader AI server growth.
Helios: Helios is AMD’s AI-optimized server rack platform built around its accelerator technology. The report anticipates Helios racks entering shipments in the fourth quarter alongside competing NVIDIA offerings.
NVIDIA: NVIDIA designs GPUs and AI accelerators that power data center and high-performance computing workloads. The report highlights the upcoming Rubin architecture racks as a key driver expected to enter volume shipments in the fourth quarter.
Bernstein: Bernstein is a global investment research firm specializing in technology, semiconductors, and infrastructure analysis. In this news, the firm updated its long-term forecasts for server and AI server shipments, emphasizing accelerating demand for GPU-based systems through 2028.
Google TPU: Google TPU is Google’s family of custom ASICs purpose-built for machine learning acceleration. Bernstein projects robust growth in TPU shipments as custom silicon captures a larger share of CoWoS-based AI chip deployments.
Custom Chip Adoption: Hyperscalers and cloud providers are increasingly turning to custom ASICs alongside merchant GPUs to optimize AI performance and efficiency.
Server Shipment Trends: GPU-dense server configurations, including 8-GPU equivalents, are expected to see accelerated deployment as rack-level solutions mature.
AI Infrastructure Outlook: Major technology firms are scaling data center capacity to support rising AI workloads, with planned and under-construction investments continuing to accumulate.