Chipmaker stocks have seen significant gains recently, with Micron rising 80%, SanDisk increasing by 52%, and Intel climbing 85%. This surge is attributed to a shift in AI systems architecture towards a model known as "orchestration," which distributes workloads across multiple processing channels instead of relying solely on larger, centralized units.
This change is expected to increase the demand for traditional central processing units (CPUs) relative to graphics processing units (GPUs), which had previously driven Nvidia's growth during the initial phase of AI development.
Analysts from Morgan Stanley, including Shawn Kim, suggest that as AI software becomes more capable of handling generalized instructions—termed "agentic" AI—the ratio of CPUs to GPUs in AI systems will rise, leading to greater complexity and a shift in infrastructure spending towards CPUs, networking, and memory.
Tech companies like Meta and AMD are also emphasizing the importance of CPUs in their AI strategies, with Meta announcing a significant deal to purchase chips from AMD while highlighting the evolving compute requirements for agentic AI.
The orchestration approach has already demonstrated its effectiveness in enhancing AI capabilities cost-effectively, as evidenced by research organizations replicating results from Anthropic's AI model using less advanced public models. This has led to a broader understanding that AI does not exclusively rely on GPUs, as noted by industry experts.
Additionally, sectors connected to data center operations, such as electronic design automation and memory systems, are benefiting from this pivot towards orchestration, with companies like KLA Corp and Samsung being highlighted as beneficiaries of this trend