AI boosts CPU temperatures, Intel hits a new record! Domestic CPU manufacturers are expected to see a "performance realization" inflection point

Driven by the explosive growth in AI computing power, U.S. semiconductor-related company stock prices continue to rise sharply. On the evening of May 11 Beijing time, the Philadelphia Semiconductor Index in the U.S. stock market once rose over 2%, with companies like Nvidia, Micron Technology, Qualcomm, Intel, and AMD all hitting record highs during trading.

As high-end chips, CPUs have long been one of the weaknesses in China’s integrated circuit industry. With the current industry boom driven by AI inference demand, domestic CPU manufacturers are expected to achieve a dual breakthrough in technological capability and product shipments more quickly, ushering in an “earnings realization” inflection point.

AI Boosts CPU Popularity

On the evening of May 11 Beijing time, U.S. stock market opened, and chip stocks surged significantly. The Philadelphia Semiconductor Index once rose over 2%, with Nvidia, Micron Technology, Qualcomm, Intel, and AMD all reaching record highs during trading. Qualcomm once surged over 13%, Intel rose over 6%, with its stock price increasing more than fivefold in a year.

On May 10 local time, Nvidia CEO Jensen Huang delivered a keynote speech at Carnegie Mellon University’s 2026 graduation ceremony and was awarded an honorary Doctor of Science and Technology degree, with the degree certificate presented by Intel CEO Chen Liwu.

Chen Liwu later posted on X (Twitter) that Intel and Nvidia are collaborating on developing exciting new products.

Currently, both companies have not officially confirmed specific product details.

Nvidia and Intel had previously announced cooperation across multiple product areas. Nvidia invested $5 billion in Intel for joint development of data center and consumer platforms, involving custom processors, advanced packaging, and foundry manufacturing in several core fields.

Behind Intel’s sharp stock price increase is a structural shift in AI computing demand from “training-centric” to “inference and intelligent agent-led.” CPUs are transitioning from supporting roles alongside GPUs to becoming the core hub of AI systems, returning to industry prominence. Rapid market demand growth has also driven Intel and AMD’s stock prices higher.

In the era of AI applications and intelligent agents, the significant increase in CPU usage stems from two factors:

First, as large models move from labs to large-scale deployment, inference demand surpasses training. Especially after AI agents (Agentic AI) become mainstream, complex system-level tasks such as dynamic interaction, logical judgment, state management, and cross-model collaboration—over 90% of which are handled by CPUs—directly boost CPU demand. Second, in the context of rapid expansion of the AI inference market, ASIC chips with higher energy efficiency are growing faster than GPUs, but ASICs can only perform “computing” tasks and cannot handle the interaction and scheduling functions undertaken by CPUs. Therefore, more CPUs are needed in AI servers and data center systems.

Currently, the typical computing ratio is 1 CPU to 8 GPUs (1:8). Industry organizations like TrendForce believe that over time, this ratio may level off or even reverse.

With demand rising rapidly and capacity constraints, the global CPU market is experiencing supply shortages, triggering a price increase wave.

Overall, since March 2026, consumer-grade CPU prices have increased by 5% to 10%, server CPU prices by 10% to 20%, and high-end AI-specific CPUs by over 30%. In channels, shortages have spread from high-end server CPUs to all categories, with longer delivery times for server-grade CPUs and some models doubling in spot prices.

Institutions forecast that the global CPU shortage will be difficult to alleviate between 2026 and 2027. Supply chain news indicates that Intel and AMD are planning a new round of price hikes in the third quarter of this year.

Domestic CPUs Achieve New Breakthroughs

“Overall, this year’s market demand is good, and domestic manufacturers have not yet raised prices,” a person from a domestic CPU manufacturer told reporters. Due to memory price increases, sales of domestically produced CPUs compatible with DDR4 old models have significantly risen, and demand for server-grade products is also rapidly growing.

Domestic CPUs are moving from performance parity with international mainstream products to a stage of “good usability.”

“Around now, companies like D, Lenovo Kaitan, and Zhiwei Intelligent, as well as ODMs, have initiated and entered testing phases for the KH-50000 validation for server-grade CPUs. Most of these clients are expected to achieve mass production of server systems by 2026,” disclosed Zhaoxin in their recent response to the second round of IPO inquiry on the STAR Market.

Zhaoxin disclosed that in Q3 2025, the company launched the new generation server CPU product KH-50000, with 500 units sold and a total revenue of 9.9105 million yuan. Leading domestic server OEM D has placed an order for 1,000 units of KH-50000 processors through China Electronics International.

Public data shows that current domestic CPU architectures include three main technical routes: x86, ARM, and自主指令集 (independent instruction set). Products generally cover all scenarios and meet domestic market needs. On the x86 route, companies include Hygon Information and Zhaoxin; on the ARM route, Huawei Kunpeng and Tianjin Feiteng; and companies developing自主指令集 include Loongson Zhongke and Shenwei.

Which domestic CPU is the strongest? Zhaoxin cited data from Yiou Think Tank in its inquiry response: in key industry server markets, Hygon Information and Huawei Kunpeng together held over 90% of the market share in 2024, while the other four domestic CPU manufacturers (Zhaoxin, Loongson, Feiteng, Shenwei) combined held less than 10%.

In the key desktop PC processor market, Zhaoxin held about 40% of the market share in 2024, ranking first.

Hygon Information and Loongson Zhongke also recently disclosed new business developments.

Loongson Zhongke revealed during recent institutional research that many applications of Loongson server CPUs are in intelligent computing scenarios. For example, the newly built autonomous intelligent computing center at Zhengzhou Port has over 500 servers all using Loongson 3C6000/D server chips, with multiple implementation cases; they also hope to make breakthroughs in storage server chips in the second half of this year.

Hygon Information disclosed that the current explosive demand for AI computing power presents three major opportunities:

First, the explosion of AI computing demand, accelerated large model and intelligent computing center construction, continuously drives the demand for DCU and other intelligent computing products; second, the dual drive of domestic acceleration and commercial deployment has gained broad recognition from customers, forming large-scale sales and supporting the expansion of CPU and DCU markets; third, digital economy empowerment and the advancement of Digital China construction promote digital transformation in government, finance, energy, and other industries, boosting demand for general and intelligent computing.

Public data shows that in 2024, the sales revenue of domestically produced CPUs was about 250 billion yuan, a year-on-year increase of 41.3%. From 2025 to 2027, the compound annual growth rate of domestic CPU market is expected to reach 13.3%, with the market size surpassing 360 billion yuan by 2027.

In the AI server field, global computing power competition is intensifying, driving demand for CPUs and AI chips. IDC predicts that in 2025, China’s AI server shipments will grow by 56.3% year-on-year to 985k units; by 2029, shipments are expected to reach 2.72 million units, with the Chinese market exceeding 350 billion yuan.

It is worth noting that in the high-end processor chip field, x86 architecture processors have an early start and a relatively mature ecosystem, giving them a clear advantage over other architectures.

(Source: Shanghai Securities News)