TAIPEI, TAIWAN – Global AI server shipments are forecast to grow more than 28% year over year in 2026, driven by continued investment from North American cloud service providers and a rising share of ASIC-based systems, according to new market research from TrendForce.

TrendForce projects total global server shipments, including AI servers, to increase 12.8% year over year in 2026, accelerating from 2025 levels as demand shifts from large language model training toward inference workloads. The growth of AI agents, Copilot upgrades, and LLaMA-based applications has pushed CSPs to expand both dedicated AI infrastructure and general-purpose servers that support pre- and post-inference processing.

The combined capital expenditures of the top five North American CSPs – Google, AWS, Meta, Microsoft and Oracle – are expected to rise 40% year over year in 2026. TrendForce said a portion of that spending will be directed toward replacing general-purpose servers deployed during the 2019–2021 cloud investment cycle, alongside new capacity additions.

Google and Microsoft are expected to lead procurement of general-purpose servers to support rising inference traffic from services such as Gemini and Copilot. While GPUs will remain the dominant AI server platform in 2026, accounting for an estimated 69.7% of shipments, ASIC-based AI servers are projected to reach 27.8% of total shipments, marking their highest share since 2023.

TrendForce said shipment growth for ASIC-based AI servers is expected to outpace GPU-based systems, led by expanded in-house silicon efforts from companies such as Google and Meta. Google’s TPU platform, originally developed for internal use, is increasingly being offered to external customers, including AI model developers.

Systems based on NVIDIA’s GB300 platform are expected to drive the majority of GPU-based AI server shipments in 2026, while VR200-based platforms are projected to gain momentum in the second half of the year.

• Next