Summary of the Core Content
This news article highlights a significant, often overlooked transformation in the AI industry: Over the past three years, during the boom of generative AI, everyone has been scrambling for NVIDIA GPUs, with CPUs being relegated to a secondary role. Intel’s market value even fell significantly behind that of NVIDIA. However, with the rapid adoption of AI agents capable of performing tasks independently, the demand for CPUs has skyrocketed exponentially. Leading cloud providers such as Amazon, Microsoft, Google, and ByteDance’s火山引擎 have all abandoned the x86 architecture, which has been standard in servers for decades, in favor of self-developed CPUs based on the Arm architecture. This represents the fastest technological shift in the history of data centers, and the landscape of the server chip market is set to be completely rewritten in the next five years.
Detailed Explanation
Why Did GPUs Dominate Before, and Now Are CPUs Suddenly Popular?
In the past, AI relied heavily on GPUs, which were used to process large amounts of data quickly for simple “question-and-answer” tasks. A single GPU, with its numerous cores, could handle the entire process, while the CPU’s role was limited to transmitting data. In extreme cases, eight GPUs paired with one CPU were sufficient. However, as AI evolved to more advanced “agents,” the need for CPUs increased dramatically. For example, an AI agent tasked with organizing a team building event would need to check colleagues’ schedules, compare flight prices, select hotels within a budget, send out notifications, and handle financial reimbursement processes—all involving numerous small-scale operations. In this context, using a GPU would be like using a 10-ton truck to buy just two pounds of vegetables, where the fuel cost would outweigh the value of the vegetables; CPUs, which are better at handling multi-threaded tasks, are much more efficient.
Why Are Giants Switching to Arm Architecture?
For decades, the server CPU market has been dominated by Intel and AMD’s x86 architecture. Now, the giants are turning to Arm for several key reasons:
- Control: They can customize their CPU designs to meet the specific needs of AI applications, without being constrained by chip manufacturers.
- Cost Savings: AI data centers consume enormous amounts of electricity. The Arm architecture is more energy-efficient, saving millions of dollars annually for large-scale deployments. California, for instance, has implemented policies to limit new data center construction due to energy concerns.
- Market Competition: The market has become more competitive, with multiple players vying for a share. NVIDIA has entered the CPU market with its Vera CPU; Arm and Qualcomm offer pre-built CPUs to smaller cloud providers; leading companies like Amazon, Microsoft, and Google are developing their own CPUs using the Arm architecture.
Who Is Most Affected by This Change?
The x86 architecture still holds a significant advantage due to its mature ecosystem, but the shift to Arm is already underway. By 2029, 90% of AI servers are expected to use Arm CPUs, and the market size could grow sixfold by 2030. This will lead to lower costs for AI services and faster adoption of AI technologies, making them more accessible to everyone. Additionally, it will enhance the security of China’s domestic computing supply chain by reducing reliance on imported chips.
What Will This Mean for Us?
- Lower Prices: Reduced hardware and energy costs will make AI services more affordable.
- Faster Adoption: AI technologies will become more widespread, automating tasks and improving daily experiences.
- Enhanced Security: Domestic companies will have more control over their computing infrastructure, reducing dependence on foreign suppliers.
In summary, the transition to Arm-based CPUs is transforming the server chip market and will have a significant impact on the cost and availability of AI services, as well as on the daily lives of individuals and businesses.