Fanning the readiness for a data centre boom, global server spending is set to skyrocket in 2025, driven by GPU-powered AI demands and cloud expansion.
The worldwide server market is set to surge to $366.0 billion in 2025, marking a 44.6% increase from 2024, according to a recent data from the International Data Corporation (IDC). This growth is being driven largely by graphics processing units (GPU)-accelerated servers and increasing demand from hyperscalers and cloud providers.
In the first quarter of 2025, the server market achieved a record value of $95.2 billion, up 134.1% year-on-year (YoY). The report highlighted it to be the highest quarterly growth ever recorded.
Servers equipped with embedded GPUs are expected to represent nearly half of the entire market this year, growing 46.7% from 2024. The x86 server segment is forecast to climb by 39.9% to $283.9 billion. Non-x86 servers are projected to rise even more steeply, increasing 63.7% YoY to $82.0 billion.
“GPU-based servers continue to dominate and push the market forward,” said Juan Seminara, research manager at IDC. “Organisations are increasingly relying on accelerated computing to power workloads related to artificial intelligence, machine learning, and complex data analytics.”
ARM-based servers are also seeing rapid adoption, thanks to large-scale rack deployments. IDC expects this segment to grow by 70.0% in 2025, making up 21.1% of total server shipments.
Regionally, the United States leads with 59.7% growth, contributing nearly 62% of global server revenue in 2025. China follows with a 39.5% increase, accounting for over 21% of worldwide revenue. Japan and APeJC are projected to grow 33.9% and 10.8% respectively. EMEA and Latin America are showing modest gains at 7.0% and 0.7%, while Canada is set to decline by 9.6% due to an unusually large deal in 2024.
“The industry’s move from basic chatbot AI to advanced reasoning and agentic models is demanding several orders of magnitude more computing power,” said Kuba Stolarski, research vice president of Worldwide Infrastructure at IDC.
“Even as models become more efficient, their requirements for real-time inferencing and scalability are growing, especially in multi-user environments,” he noted.
