SuperMicro takes on server leaders as AMD pushes on-premise AI
Market information from analyst IDC has proven that SuperMicro has leapfrogged established server makers Lenovo and HPE because the second-largest PC server maker behind Dell.
SuperMicro skilled progress of just about 134% for the fourth quarter of 2025 with income of $11.7bn, which suggests it accounts for over 9% of the worldwide server market. Dell was forward with 10% market share and income of $12.6bn, whereas Chinese language producer IEIT Methods took the third spot, with income of $5.2bn and a 4% market share forward of Lenovo, which posted income of $5.1bn, and HPE ($3.9bn).
“The race for AI [artificial intelligence] adoption is settling the market tempo, and with corporations ravenous for infrastructure trying not solely at GPUs [graphics processing units], but in addition consuming extra CPUs [central processing units] amongst different elements with the intention to feed their wants, we’re going to see extra worth pressures, and that will influence on market dynamics with much less items however increased common promoting costs going ahead,” stated Juan Seminara, analysis director of Worldwide Enterprise Infrastructure Trackers at IDC.
IDC famous that risky rising costs on sure elements resembling GPUs, dynamic random entry reminiscence (DRAM) and strong state drives (SSDs) has meant that some corporations have been making an attempt to safe costs forward whereas the trade is accommodating to the brand new actuality. It predicted that the influence of this worth volatility could possibly be hitting tougher throughout 2026 as demand retains outpacing service capability within the close to time period.
In addition to Dell, the established server makers appear to be shedding floor within the server market. However they look like a brand new market alternative being pushed by chipmaker AMD, which is the deployment of on-premise PC servers optimised to run agentic AI.
In a bid to entice IT patrons away from cloud-based AI {hardware}, AMD has unveiled what it sees as a brand new class of PC known as Agent Computer systems. In a submit on the AMD web site, the corporate described learn how to run OpenClaw, the open supply AI agent, domestically on AMD Ryzen AI Max+ processors and Radeon GPUs utilizing a Home windows 11 PC with the Home windows Subsystem for Linux (WSL).
AMD stated the PC system configured with 128GB unified reminiscence is able to working “cloud-quality AI agent workloads effectively” utilizing OpenClaw. In accordance with its personal benchmark information, with the Qwen 3.5 35B A3B mannequin, the system delivers round 45 tokens per second and processes 10,000 enter tokens in about 19.5 seconds. AMD stated the configuration helps a most context window of 260,000 tokens, and might run as much as six brokers concurrently, which it stated means it is ready to ship scalable native AI experimentation whereas sustaining sturdy responsiveness on shopper {hardware}.
AMD sees such a system working autonomously relatively just like the pre-cloud period department workplace servers, dealing with duties despatched by customers by way of a browser consumer interface on one other Home windows PC, or by way of Slack or WhatsApp.
PC makers which have “agent-ready” PCs embody HP, Lenovo and Asus. The IDC figures present that income for servers with an embedded GPU within the fourth quarter of 2025 grew 59.1% year-over-year, representing greater than half of the full server market income.
The AMD Ryzen AI Max+ has an built-in GPU, and is presently one of many processor choices for PCs licensed as Copilot+ gadgets. Whereas these gadgets are both laptops or desktop PCs with displays, AMD’s Agent Pc seems to be positioned as extra of a conventional desktop Home windows PC working as a server, and not using a display or keyboard. The setup AMD offers is optimised to run LM Studio. This makes use of Ubuntu on the WSL to supply entry to giant language fashions, which then work with an OpenClaw server working domestically on the identical {hardware}.
