Hardware

GPT-5.6 Sol runs at 750 tokens per second on Cerebras wafer-scale silicon — roughly 15x a GPU. When frontier inference gets an order of magnitude faster, latency becomes the competitive axis.

NVIDIA's first consumer SoC and Microsoft's agentic Windows move part of the AI economy onto the device and redraw the data and security boundary that cloud AI had centralized.

While Nvidia dominates headlines, a massive infrastructure buildout is creating opportunities across optical networking, memory chips, storage, and data center components as companies project $380 billion in spending