The global artificial intelligence boom is reshaping one of the semiconductor sector’s most important battlegrounds: central processing units (CPUs).
The market for CPUs has long been dominated by Intel and AMD, but now a wave of challengers is pouring in, including Nvidia, Qualcomm, MediaTek, Arm, Google, and Amazon, as well as Chinese players.
They are all pursuing the same growing opportunity. The market for CPUs used in next-generation AI data centers is projected to grow more than 40% annually over the next five years, surpassing USD 220 billion by 2030, according to multiple industry executives and research firms.
CPUs, the “brains” of most computers and servers, have long been seen as a mature, stable business with limited growth potential. Initially, the artificial intelligence boom appeared to sideline them even further as graphics processing units (GPUs) and other AI accelerators took on the vital role of training AI models.
In this early, training-focused stage of the AI computing boom, the ratio of CPUs to GPUs in running computing workloads was typically one-to-eight. That ratio has narrowed to roughly one CPU for every four GPUs, and industry executives expect it to tighten further to one-to-two, or even one-to-one, as reasoning, inference and agentic AI workloads place greater demands on CPUs.
AI agents are in particular focus, according to Bank of America analyst Vivek Arya. These systems, which plan, retrieve information and execute code, introduce so-called orchestration tasks to manage workflows. Because the tasks must be carried out in sequence and are sensitive to latency, they are better suited to being run on CPUs than on accelerator chips, Arya said.
Bernstein Research estimates the data center CPU market will expand from just USD 37 billion in 2025 to USD 223 billion by 2030. Bank of America has a more conservative outlook, but still expects the market to reach more than USD 170 billion by 2030, a nearly fivefold increase from current levels and a 37% annual growth rate through 2030.
The surge in demand for CPUs has tightened supplies across the electronics sector, including for PC makers, Nikkei Asia previously reported. This supply crunch, in turn, has attracted a wave of new chipmakers into the market.
“CPU attach rates are soaring through the roof. You can’t find CPUs anywhere, they’re already been bought up,” Tony Pialis, who runs the new data center business at Qualcomm, told investors in New York last month. The mobile chip leader has been among the most aggressive in tapping the growing CPU market.
Nvidia is also joining the fray. The GPU giant begun offering its Vera CPUs as a standalone product, expanding beyond its end-to-end AI computing platform strategy. CEO Jensen Huang said the CPU business alone could generate as much as USD 20 billion in annual revenue for the company this year.
Intel, the longstanding market leader in data center CPUs, is seeing a similar rise in demand, according to CEO Lip-Bu Tan.
“I got several calls over the past few weeks asking for more CPUs,” Tan said on the sidelines of the Computex trade event in Taipei last month. Intel’s market share for data center CPUs remains above 50%.
AMD CEO Lisa Su, meanwhile, projects that the CPU market will grow at least 35% in the next five years, unlike the 3–4% in the past.
SoftBank-backed Arm is also targeting the CPU renaissance, unveiling its first physical chip, a CPU for data center servers, this year. Arm said its CPU business could generate up to USD 15 billion in annual revenue by 2030.
Cloud computing giants including Google, Amazon and Microsoft are developing custom Arm-based CPUs for their own data centers. Including both its chip and licensing businesses, Arm told Nikkei Asia its technology now powers more than 50% of CPUs deployed in the world’s leading hyperscale data centers.
This influx of competition is poised to reshape the CPU market landscape.
Longtime leaders Intel and AMD, whose x86 processors set the standard for high-performance computing, have long led the market. By contrast, many of today’s challengers, including Nvidia, Google and Amazon, are betting on Arm-based infrastructure, which is more commonly used in smartphones and mobile processors because of its energy efficiency.
“The CPUs of Nvidia and Arm are Arm-based architecture, which gives them an edge in power saving, especially as GPUs already consume so much power,” said Brady Wang, associate director with Counterpoint Research. “The CPUs of Intel and AMD, on the other hand, have always been good at conducting complicated tasks.”
According to TechInsights, Arm-based data center CPUs are set to gain global market share over the next few years from about 20% in 2026, though the overall market is expanding as well.
Despite this architectural shift, the leading players remain concentrated in the Western Hemisphere.
“For data centers, I think there’s a lot of opportunity in this market,” said James Sanders, an analyst at TechInsights. “The relative share of x86 is forecast to decline year over year over the next few years, and part of that story is the growing adoption of custom silicon based on [Arm architecture.] But overall, it’s really a story of strong market growth. I think there’s room for everyone in this market.”
China has long hoped to develop domestic alternatives to cut reliance on foreign CPUs, and that push has gained momentum in the past few years.
China’s key domestic CPU developer Hygon Information Technology, which uses x86 architecture but is on a US trade blacklist, is gaining local market share as Chinese cloud service providers accelerate domestic sourcing, sources told Nikkei Asia. Hygon’s revenue has grown significantly in the past few years.
“Hygon will benefit from strong x86 CPU demand and gain share in China,” said David Dai, an analyst with Bernstein Research. This growth is underpinned by continued product improvements and better interoperability with domestic AI accelerators, Dai said, which should make Hygon platforms increasingly viable for cloud service providers beyond its traditional customer base of the government and state-owned enterprises.
Bernstein estimates Hygon’s share of China’s server CPU market will increase from about 20% this year to roughly 35% by 2028, although foreign suppliers are still expected to account for the majority of the market.
Hygon is not alone. Shanghai Zhaoxin Semiconductor, another domestic CPU developer focused on PCs and servers, is preparing for a listing on Shanghai’s Star Market, China’s equivalent of the Nasdaq.
Other homegrown CPU makers, including Loongson—once viewed as China’s best hope of building a homegrown rival to Intel—are also expanding. But despite steady revenue growth in recent years, both Loongson and Zhaoxin remain loss-making.
Meanwhile, Huawei has continued to develop its Kunpeng server CPUs as part of its broader effort to build a self-sufficient computing ecosystem.
Because of their comparatively lower computing power, CPUs are subject to less restrictive US export controls than GPUs. As a result, the CPU market is one of the few segments where American chipmakers can still compete in China, even as they face growing competition from the country’s local players.
Both Nvidia’s Huang and AMD’s Su have acknowledged the importance and competitiveness of the Chinese market. Qualcomm CEO Cristiano Amon told Nikkei Asia that his company plans to specially design chips for Chinese data center customers that are in compliance with US export controls. Arm, meanwhile, announced that ByteDance is an important first adopter of its AGI CPUs.
The PC market is another arena where the CPU race is heating up thanks to the rise of agentic AI, even though it is already crowded with existing players like Intel, AMD, and Qualcomm.
MediaTek, the world’s top smartphone chip developer by shipments, has made inroads into designing CPUs for laptops for the first time, joining hands with Nvidia to co-develop the RTX Spark superchip for Windows PCs. The chip comes with a MediaTek customized CPU integrated with Nvidia’s Blackwell GPU.
The two companies have high hopes for RTX Spark, with MediaTek CEO Rick Tsai saying the product has the “right” architecture for the agentic AI era and will give the PC “new life after 40 years.”
An executive with a supplier to Nvidia, Apple, HP, and Google told Nikkei Asia that the market should not take Nvidia’s standalone Vera CPU and RTX Spark chip lightly.
“Considering Nvidia’s dominance in AI infrastructure, we think its CPUs would put a certain pressure on Intel and AMD,” the industry veteran said. “However, we should also not underestimate Intel’s strong tech capability, manufacturing, and its market position in the end devices like the PC industry. The race is just beginning.”
This article first appeared on Nikkei Asia. It has been republished here as part of 36Kr’s ongoing partnership with Nikkei.