ARM-based servers are growing 70 percent this year in terms of units shipped, but have not yet reached the market share the chip designer envisions.
Research firm IDC predicts that servers with ARM processors will account for 21.1 percent of all worldwide servers shipped in 2025. That is significantly more than last year, but still far from Arm’s goal. They aim to capture half of the data center CPU market by the end of this year.
This success is primarily thanks to large-scale rack configurations like Nvidia’s DGX GB200 NVL72, intended for AI workloads. These are often supported by ARM processors, such as Nvidia’s Grace Hopper combo.
The use of servers with one or more GPUs, often called AI servers, is rising by 46.7 percent and represents almost half of the total market value this year. Hyperscale customers and cloud providers are further driving demand. IDC expects the total server market to triple in three years.
Server Market Reaches Record High
In the first quarter of 2025, the global server market reached record revenue of 95.2 billion dollars, a year-on-year increase of 134.1 percent. For the full year 2025, IDC estimates total revenue at 366 billion dollars, an increase of 44.6 percent compared to 2024.
x86 servers, still the industry standard, are growing by 39.9 percent to 283.9 billion dollars. Non-x86 systems, including ARM, are growing faster at 63.7 percent and collectively reach 82 billion dollars.
read also
Sea cooling in a closed circuit: the future of data centers?
Regionally, the United States stands out with an expected growth of 59.7 percent, accounting for 62 percent of global server revenue by the end of 2025. China is growing by 39.5 percent and accounts for over 21 percent. EMEA and Latin America are growing modestly at 7 and 0.7 percent respectively, while Canada shrinks by 9.6 percent this year after a large project in 2024.
Stargate project
IDC expects the demand for computing power to continue to rise due to the development of AI applications. The relaunch of the Stargate project, which aims to invest up to 500 billion dollars in AI infrastructure, also plays a role. Additionally, the introduction of DeepSeek‘s R1 reasoning model creates additional demand for infrastructure.
According to IDC, the step from simple chatbots to models that can reason independently is a significant driver behind the increasing demand for processing capacity. The analyst firm expects that more efficient AI models will require less computing power, allowing companies to scale their systems better.
Nevertheless, the need for powerful servers with GPUs remains high because new AI applications such as agentic AI demand significantly more computing power, especially for inferencing. As a result, both x86 and ARM servers continue to benefit from the AI wave.