Arm based servers, global artificial intelligence infrastructure expenditures It surpassed x86 systems for the first time in history. According to IDC data, companies will implement Arm-based accelerated servers in the first quarter of 2026. 53 billion dollars spent. On the other hand, the spending amount of x86-based systems decreased in the same three-month period. $34.6 billion remained at the level. In the server market, which Intel and AMD have dominated for many years, the balance of spending has shifted in favor of Arm for the first time.
However, IDC globally artificial intelligence infrastructure expenditures increased by 33.1 percent annually to $89.7 billion He announced that he was leaving. Server systems with $87.6 billion While it constituted 97.6 percent of the total expenditure, storage products received a share of $2.2 billion. The research company also made its market forecast for the entire year 2026. to 497 billion dollars raised it. IDC predicts that total spending will exceed $1.08 trillion in 2029 and $1.21 trillion in 2030.
On the other hand, the rise on the Arm side is not only explained by classic CPU sales. Nvidia’s products, which combine Arm-based processors with GPUs in the same rack-scale systems, play an important role in the increase in spending. Major cloud providers and tech companies are no longer just increasing the number of GPUs doing model training. These companies are also renewing the CPU layer, which handles tasks such as data preparation, memory management, network traffic, workload planning and accelerator management.
Besides that artificial intelligence agentsexpands the tasks handled by server processors. Rather than responding to a request in a single step, continuously running agents access data sources, summon vehicles, and schedule tasks. It then re-evaluates the results and restarts the process chain when necessary. AMD states that this process requires more CPU resources for GPU management, memory access, data preparation and control flow.
Arm processors push the boundaries of power and rack density
Data center companies no longer grow server clusters solely on processor count and raw performance. Electrical capacity, cooling requirement, and power limit per rack artificial intelligence servers directly determines its scale. That’s why Arm emphasizes performance per watt and high core density in its Neoverse-based designs. Lower power consumption makes it possible to run more processing cores and accelerators within the same electricity budget.
Additionally, Arm introduced in March 2026 Arm AGI CPU introduced its own data center processor to the market for the first time. The company previously licensed processor architecture and core design to its customers and did not sell server processors directly. Arm AGI CPU, 136 Neoverse V3 cores, 300W power consumption and offers 6 GB memory bandwidth per second per core. Arm states that liquid-cooled systems can reach more than 45 thousand cores in a single rack.
However, the company claims that AGI CPU-based systems offer more than twice the performance per rack compared to x86 platforms. Meta worked with Arm during the development of the processor. OpenAI, Cloudflare, SAP, Cerebras and SK Telecom are also among the companies supporting the platform. These collaborations show that Arm processors are gaining use not only in experimental servers but also in large-scale commercial data centers.
Spending leadership doesn’t mean Arm is beating x86 across all server workloads, however. The figures shared by IDC are especially for GPU-accelerated and rack-scale artificial intelligence servers It covers. General-purpose enterprise applications, virtualization, databases and existing software infrastructures continue to use x86 processors extensively. Intel and AMD maintain their strong market shares in these areas thanks to the software ecosystem they have supported for many years.
On the other hand, AMD emphasizes that companies can run their existing applications on x86 systems without recompiling or managing different code bases. This compatibility reduces the migration costs large companies face when moving existing data center software to new servers. AMD also positions its EPYC processors with high core count, large memory capacity and standard x86 compatibility. EPYC 9005 series on high-end models to 192 cores It turns out up to .
In addition, AMD explains in its own test results that the EPYC 9965 processor provides up to 1.7 times higher end-to-end performance than the Intel Xeon 6980P in certain artificial intelligence workloads. The company also argues that in some tests, EPYC systems offer higher performance per watt against Nvidia Grace-based Arm servers. However, AMD obtains these figures from tests conducted under its own hardware and software conditions. Therefore, results may vary depending on the data center software and accelerator used.
IDC data also reveals that spending on the x86 side has decreased significantly in a short time. According to the research firm, spending on x86-based accelerated servers will increase by about From 52 billion dollars to 35 billion dollars landed. By contrast, spending on Arm-based systems has nearly doubled over the same period. New generation Intel Xeon and AMD EPYC processors may change the spending distribution again in the following quarters.
On the other hand, the $53 billion level reached by Arm clearly shows the criteria by which data center companies change their processor preferences. Companies are looking for higher energy efficiency on the CPU side to increase the power budget allocated to GPUs. Additionally, designs that increase the number of cores per rack enable placing more servers in the same data center area. While Arm-based processors meet these needs, the x86 side continues to compete with software compatibility and enterprise application support.
IDC’s 2026 forecast shows that this competition will continue with a much larger spending volume. 497 billion dollars The annual forecast directly concerns processor, GPU, network equipment, storage system and cooling infrastructure manufacturers. The competition between Arm, Intel and AMD is not limited to processor sales figures. Electricity consumption, rack density, software compatibility, and processing requirements of artificial intelligence agents all play a role in purchasing decisions.
Join Channel