Meta's multi-billion dollar deal to rent Amazon's Graviton 5 CPUs, not just GPUs, signals a potential architectural shift for AI. This move suggests that CPU architecture could be more efficient or cost-effective for agentic workloads, challenging the conventional wisdom that GPUs are the only viable hardware for scaling AI applications.

While GPUs dominate AI hardware discussions, the proliferation of AI agents is causing a significant, often overlooked, CPU shortage. Agents rely on CPUs for web queries, data processing, and other tasks needed to feed GPUs, straining existing infrastructure and driving new demand for companies like Arm and Intel.

AI's evolution from training-heavy (GPU-dominant) to inference- and agent-heavy (CPU-intensive) workflows could invert the traditional data center chip ratio. This represents a seismic shift, creating a massive tailwind for CPU manufacturers like Intel.

While GPUs train models, CPUs are essential for two key workloads: running reinforcement learning environments and executing the code generated by AI. This has created a massive, often overlooked demand spike, making CPUs a critical, sold-out component in the AI infrastructure stack and a hidden bottleneck.

The focus on GPUs for AI overlooks a critical bottleneck: CPU shortages. AI agents require massive CPU power for non-GPU tasks like web queries and data prep. This demand is straining existing infrastructure and creating new market opportunities for CPU makers like ARM.

The focus on GPUs for AI overlooks a critical bottleneck: a growing CPU shortage. AI agents rely heavily on CPUs for orchestration tasks like tool calls, database queries, and web searches. This hidden demand is causing hyperscalers to lock in multi-year CPU supply contracts.

The AI narrative has focused on GPUs for training, but the proliferation of AI agents for task execution is creating a massive, overlooked demand for CPUs. This shift to inference and orchestration is reversing Intel's recent decline.

SiFive's Krste Asanović highlights that while GPUs are the focus of the AI boom, the CPUs that feed them data are a critical bottleneck. As AI accelerates tasks like coding by 30x, the corresponding CPU-bound tasks like compiling also need a 30x speedup, driving demand for specialized CPU IP.

After the current memory crunch, the next AI infrastructure bottleneck will be CPU and networking. The complex orchestration required for emerging agentic AI systems will strain these resources, a trend already visible in companies like Fastly seeing demand spikes just for workload orchestration.