AI Server Growth in 2025 Faces Uncertainties, DeepSeek Effect to Boost AI Inference Share

Estimated reading time: 1 minute

TrendForce’s latest investigations reveal that global AI server shipments grew by 46% in 2024, driven by strong demand from CSPs and OEMs. However, multiple factors, including US chip restrictions, the DeepSeek effect, and supply chain readiness for GB200/GB300 racks, could impact AI server shipments in 2025. TrendForce has outlined the following three scenarios based on these uncertainties.

The base case—considered the most probable—factors in Microsoft, Meta, Amazon, and Google’s recent announcements of increased capital expenditures in cloud and AI infrastructure, with expected annual growth exceeding 30%. This investment is expected to sustain momentum for AI server demand.

Currently, the supply chain outlook for shipments of AI servers remains stable. Under this scenario, global shipments of AI servers in 2025 are projected to grow by nearly 28% year-over-year. However, if GB200/GB300 rack shipments face scaling challenges, CSPs may prioritize HGX, MGX, or their own ASIC-based infrastructure instead.

The worst case scenario, with the second-highest probability, estimates the growth of AI server shipments to slow to 20-25% based on three key factors: First, stricter US regulations on AI chip exports to China create greater market uncertainty. Second, delayed shipments of NVIDIA GB racks as their complex design pushes large-scale deployment to the second half of 2025. The GB300, set for release around the same time, features higher specifications, which could extend testing and validation timelines for server rack systems. Lastly, the DeepSeek effect, which may lead customers to reduce dependence on high-end AI solutions.

The most optimistic scenario (bullish case) assumes proactive AI policies from both the US and China, such as the US’ Stargate Initiative, which could accelerate AI server deployment.

While DeepSeek may impact high-end GPU adoption, US-based CSPs are procuring proprietary ASIC AI servers faster than expected. Additionally, DeepSeek is expected to expand AI applications, driving edge AI server growth. Based on these factors, AI server shipments could increase by nearly 35% in 2025.

TrendForce highlights that DeepSeek’s influence will drive CSPs toward lower-cost proprietary ASIC solutions, shifting focus from AI training to AI inference. This shift is expected to gradually increase the share of AI inference servers to nearly 50%.

While GPUs will remain essential for developing AI models, the market will become more segmented. CSPs focusing on large-scale AI algorithms will continue investing in high-performance solutions, while cost-sensitive enterprises at the edge will prioritize more economical alternatives.