Global CPU leader stocks surge aggressively, new AI-driven theme emerges

On April 24, the global semiconductor sector experienced a rare synchronized sharp fluctuation. Before the US stock market opened, Intel released positive signals through its latest earnings report and communication, with pre-market gains approaching 30% at one point; in the A-share market, domestic CPU leader Hygon Information Technology also strengthened, closing up 8.20%.

This is no coincidence. Unlike the compute power trend surrounding GPUs in the past two years, this round of market focus has clearly shifted to leading CPU companies. The market has begun to revisit a key question: Is the growth of AI compute power still just a story of 'more GPUs'?

01 A Significant Shift in AI Compute Logic

For a long time, the investment logic in the AI industry was relatively clear: the larger the model, the more compute power required, and GPUs naturally became the core variable. However, entering 2025, the industry's focus began to shift.

In its recent earnings communication, Intel pointed out that the key issue for AI infrastructure is transitioning from compute supply to system orchestration efficiency. Rather than simply expanding compute scale, enterprises are now paying more attention to how systems are orchestrated, how resources are utilized, and whether the same hardware can support more applications running continuously.

Behind this change lies a shift in the phase of AI applications. Training remains important, but inference, intelligent agents, and industry-specific applications are becoming the new main drivers. These scenarios no longer pursue extreme compute power but instead rely more on stability, efficiency, and cost control.

As a result, the importance of CPUs is being revisited. In a research report published earlier, CITIC Securities explicitly stated that the market had previously 'underestimated' the role of CPUs in the AI era. With the evolution of AI application models, demand for CPUs is being reactivated, driving a reassessment of their value.

02 CPUs Return to Center Stage

In current AI systems, GPUs still handle the primary computing tasks, but the efficiency of system operation increasingly depends on how compute power is organized.

Intel noted that as AI systems grow more complex, the role of CPUs in task scheduling and data flow management is becoming more prominent. This is gradually forming a consensus within the industry. When compute clusters expand to tens of thousands of cards, insufficient CPU performance will directly become a system bottleneck.

CITIC Securities analysts further emphasized that in AI clusters, CPUs not only handle control and scheduling functions but also serve as key providers of shared memory to alleviate VRAM constraints during large model inference. This means that CPUs affect not only whether a system 'can run' but also 'how fully it can run.'

At the same time, the role of the CPU is extending into more practical scenarios, such as the control core in AI clusters, participating in lightweight model inference and cloud resource reuse, and handling motion control and execution logic in embodied intelligence. The common feature of these scenarios is that the CPU's 'position' in the AI system is moving up, returning to the center stage.

03 Valuation Reconstructed by Demand

From the industry perspective, the change in CPU demand does not stem from a single market but rather from changes in how computing power is utilized.

In inference scenarios, some lightweight models can already run on CPUs. Research data from CITIC Securities shows that under single-user conditions, the inference performance of 7B to 14B models on high-end CPUs has become practical. Meanwhile, the long-term low utilization rate of CPUs in cloud computing systems is being addressed by reallocating these resources for AI computations.

In the field of embodied intelligence, it has become an industry consensus that CPUs and GPUs form a division of labor structure. Multiple robotics manufacturers adopt a 'CPU+GPU' combination, where the CPU handles real-time control and communication functions, forming the foundation for stable system operation.

These overlapping changes have transformed the CPU from merely a 'supporting component' into a fundamental unit directly involved in the operation of AI systems.

04 Ecosystem Determines Certainty

Although CPU demand is recovering, the benefits to the industry will not be evenly distributed. A key limiting factor is that enterprise-level computing systems are highly dependent on software ecosystems. CITIC Securities' research report points out that while architectures like ARM have increased their share in the server market, x86 still occupies a significant position in AI systems due to its ecosystem compatibility advantages.

Industry insiders indicate that this will directly affect vendors' actual profitability. CPU solutions capable of integrating with existing software systems and reducing migration costs are more likely to enter real business systems. Among current domestic vendors, Hygon Information stands out as a reliable choice.

Relevant materials show that based on C86 ecosystem compatibility, Hygon Information has achieved continuous penetration in key industries such as finance, telecommunications, and energy.

The aforementioned individuals believe that after the structure of AI computing power changes, the importance of CPUs within systems increases, which also expands the potential benefits for such manufacturers. From an industry evolution perspective, this round of change is more akin to a redistribution of infrastructure rather than simple substitution. Throughout this process, demand will concentrate on infrastructure vendors with ecosystem承接 capabilities.

Overall, the strengthening of the CPU sector this time appears to be an inevitable result of the changing logic in AI computing power. As the industry shifts from 'increasing computing power' to 'enhancing the efficiency of computing power usage,' CPUs are returning to a central position within systems, and their value will be reassessed. This will not diminish the importance of GPUs but will alter the value distribution across the entire industrial chain.

In this process, the true beneficiaries will not only be chip-making companies but also infrastructure vendors capable of entering core system pathways and addressing genuine business needs. And this is the aspect of CPU value reevaluation under the current AI computing power framework that deserves closer attention.

The cover image and illustrations of this article are owned by the copyright holder. If the copyright owner believes their work is unsuitable for public viewing or should not be used free of charge, please contact us promptly, and we will make immediate corrections.

15K Views