The global server market consumed $122.62 billion in the first quarter of 2026, up 30.4 percent year on year but down only 2.1 percent sequentially, according to IDC. Rising prices for CPUs, GPUs, DRAM and flash storage counterbalanced the supply-demand imbalance enough to hold the sequential decline to a near-negligible slip.

Supply currently exceeds — or rather, demand exceeds supply — by an estimated 25 percent to 30 percent, based on hints from the large OEMs. That gap, combined with component price inflation, has given server makers room to pass costs through and, where contracts have not locked in machine prices, to add opportunistic increases. The result: a market that declined only slightly against a historical pattern of much steeper Q4-to-Q1 drops.

What does the new spending pattern look like?

Before the GenAI boom, the server market followed a sawtooth rhythm for decades. IT shops spent aggressively in Q4 to preserve next year's budgets, so Q4 always beat Q3, and Q2 usually beat Q1. The 2.1 percent sequential decline in Q1 2026 is small by those historical standards.

For the past several quarters, a new normal level of server spending has taken hold despite the usual puts and takes. The GenAI wave — visible as an exponential explosion in IDC's quarterly revenue data reaching back to 1999 — now coincides with a massive, global refresh cycle for non-AI systems. That refresh lands exactly when component prices are rising sharply because demand outruns supply.

How big is the accelerated segment?

GPU-accelerated servers generated $68.9 billion in Q1 2026, up 24.8 percent year on year and down 2.5 percent sequentially. Price increases nearly compensated for shipment declines caused by component shortages. Nobody views this as a demand problem.

GPU systems took 56.2 percent of overall server revenue. The figure becomes less surprising considering that a datacenter GPU accelerator with its HBM memory costs around $50,000 today, and next-generation multi-chip devices will approach or exceed $100,000.

IDC also broke out, for the first time in its publicly available data, revenues for other XPU-accelerated machines:

• $17.1 billion in XPU system sales in Q1 2026
• 13.9 percent of server revenue, up from 8.2 percent a year ago
• Built mostly on Google TPUs and AWS Trainium, plus a smattering from Microsoft, Cerebras Systems, Groq and a handful of others

Combined, GPU and XPU systems brought in $86 billion, or 70.2 percent of all server revenue — up from 62.2 percent in the year-ago quarter.

Is x86 dominance ending?

By revenue, the answer is nearly yes. X86 hosts drove $63.9 billion in sales, down 2.9 percent year on year and down 8.5 percent sequentially. Without component shortages and price inflation pushing buyers away, the x86 server business would likely have grown dramatically in the quarter.

Non-X86 machines — dominated by hyperscalers' homegrown Arm processors and Nvidia systems built on the "Grace" Arm server CPU — accounted for $58.7 billion, up 2.1X year on year and up 5.8 percent sequentially.

Revenue share in Q1 2026:

• X86 systems: 52.1 percent
• Non-X86 machines: 47.1 percent

IBM's Power Systems RISC machines and System z mainframes account for a slice of the non-X86 figure, so it is not all Arm gear. RISC-V machines are also expected to rise before long.

Which vendors are winning?

Dell, anointed as a chosen OEM by Nvidia, benefits from that status through enterprise customers as well as large sovereign, neocloud and model-builder deals. Recent financial coverage of Dell and HPE shows Dell's AI server business is an order of magnitude larger than HPE's, making Dell's overall server business 5.5X bigger than HPE's.

The ODMs serving the hyperscalers and cloud builders are running at full tilt, constrained only by component supplies. Their business would be considerably larger if chips were easier to obtain. For once, the server makers — who work with the most demanding, penny-pinching customers in the world — appear to be earning some margin on the vig.