The demand for CPUs is set to increase significantly as AI agents are projected to proliferate, with every individual expected to operate multiple agents around the clock. The shift of the GPU-to-CPU ratio from 8:1 to 2:1 over the past year underscores this growing need for CPUs, as they are essential for orchestrating agent functions such as scheduling actions and retrieving data. Additionally, memory serves as a crucial component in maintaining context across the various tasks these agents handle, highlighting the importance of both DRAM and NAND in supporting this expanding AI infrastructure.

AMD: AMD is a major semiconductor company that designs and manufactures CPUs and GPUs for data centers, consumer computing, and enterprise applications. The news identifies AMD as another primary winner in the shift toward CPU-intensive AI agent architectures, given increased demand for processors to handle agent orchestration and coordination.
Intel: Intel is a leading semiconductor manufacturer specializing in processors for data centers, PCs, and enterprise computing. In the context of this news, Intel is positioned as a key beneficiary of growing AI agent infrastructure demand, particularly for CPU capacity needed to orchestrate AI agents and manage parallel task execution.
Deepseek: Deepseek is an AI model known for achieving strong performance with reduced DRAM requirements compared to other large language models. The news cites Deepseek V4 as an example of how even memory-efficient models still demand significant NAND storage capacity, underscoring continued demand for memory infrastructure.

`json
{
“GPU-to-CPU Ratio Shift”: “The ratio of GPU to CPU deployment has changed significantly over the past year, indicating a notable rise in CPU demand relative to GPUs.”,
“Memory as Agent Infrastructure”: “Memory plays a crucial role in maintaining context across simultaneous agent processes, with both types of memory storage being essential to AI infrastructure.”,
“Agent Architecture Requirements”: “AI systems depend heavily on CPUs for orchestrating tasks, implicating an increased demand for CPU resources in extensive agent implementations.”
}
`