Worldwide spending on AI infrastructure reached $89.7 billion in Q1 2026, up 33.1 percent from Q1 2025, and for the first time Arm-based AI server hosts out-earned their X86 counterparts, according to IDC's latest forecast, which now projects $1.21 trillion in annual AI infrastructure spending by 2030.

IDC has raised its forecast for 2026 and beyond, citing both price inflation and increased demand as drivers of spending levels. The report covers AI accelerated servers and storage — machines powered by GPUs and other kinds of XPUs.

How did Arm overtake X86 in AI hosts?

In the AI accelerated segment, X86 AI hosts — machines based on Intel and AMD processors — took in $34.6 billion in revenues in Q1 2026. Arm-based systems surpassed them, capturing a 60.5 percent share of AI hosts versus 39.5 percent for X86.

The driver is Nvidia. Wide adoption of its "Grace" CG100 server processors in rackscale NVL72 platforms pairing "Hopper" and "Blackwell" GPUs pushed Arm revenue past X86. Google is also now using its Axion Arm processors in its TPU clusters, and Amazon Web Services has used its Graviton chips as AI host CPUs for a number of years.

One caveat: these figures measure the value of a server node or rackscale system with all GPUs installed, and processors represent a relatively small portion of overall system cost. High-speed networking and denser racks now claim higher shares of total cost, even as GPUs get more expensive.

What does the total server market look like?

Total server spending in Q1 2026 came to $122.6 billion. The breakdown:

• X86 machines: $63.9 billion (52.1 percent share)
• Arm machines: $56.4 billion (46 percent share)
• IBM Power and z systems: an estimated $2.3 billion (1.9 percent share)

On these numbers, Arm will pass X86 in overall server revenue soon, regardless of what Intel and AMD do. Hyperscalers, cloud builders and some AI model builders are shifting to cheaper homegrown CPUs, and Arm is how they got there — a move some may eventually repeat with RISC-V for even lower costs.

Non-AI spending still favors X86: $29.3 billion for non-AI X86 systems versus about $3.4 billion for Arm servers that were not AI hosts.

IDC's shipment data carries a methodological caveat. IDC counts chassis by CPU type rather than distinct NUMA domains, so a chassis holding eight Arm server sleds — each a standalone server sharing DPUs for network access — counts as one shipment, not eight. That explains why Arm shows only 13 percent shipment share in IDC data. Counted by distinct compute NUMA domains, Arm has slightly more shipments than X86 across the world's datacenters, thanks to hyperscaler and cloud builder purchasing.

Where is the money being spent geographically?

The United States dominates AI infrastructure purchases but is no longer the fastest-growing region:

• United States: $67.9 billion, up 30.3 percent
• China: $7.8 billion, up 9.3 percent
• Asia Pacific/Japan: $5.8 billion, up 62 percent — twice the US growth rate
• Western Europe: $5.1 billion, up 20.4 percent
• Middle East and Africa: just over $1 billion, a 3.3x increase from a small base

Is IDC's forecast conservative?

The compound annual growth rate for AI infrastructure spending between 2023 and 2030 comes to 49.2 percent — a figure similar to what both Nvidia and AMD have independently signaled. Spending goes above $1 trillion in 2029 and grows more moderately into 2030.

Comparing IDC's numbers with the AI accelerator forecast (compute engines plus HBM memory) that AMD CEO Lisa Su presented at the recent Advancing AI 2026 event in Silicon Valley shows the two track closely — until they diverge in 2029 and 2030.

Factoring in the server and switch spending those accelerators drive, and applying a multiplier that grows over time as networking, storage, racks, CPU hosts, DRAM and flash claim larger portions of system cost, produces a substantially higher trajectory. The multiplier rises because even as GPU and XPU costs climb, the costs of the other elements of AI systems are rising faster.

The result: a big gap between the $1.21 trillion in 2030 AI infrastructure spending that IDC expects and the $2.11 trillion implied by AMD's AI accelerator TAM figures. The only way to accurately predict the future is to live it.