Stealth no more: Thinking Machines launches Inkling open model

Cerebras CEO Andrew Feldman predicts that personal AI assistants will become as common as household appliances, operating through a hybrid model that combines local device processing with cloud computing. The statement reflects industry expectations about AI's trajectory toward consumer adoption. The prediction suggests a shift toward distributed AI architecture rather than purely cloud-based or on-device solutions.

Three enterprise leaders discuss how agentic AI is forcing fundamental changes to company structure and operations in 2026. Rather than layering AI onto existing processes, Block, Cloudflare, and Klaviyo are redesigning organizational models, eliminating whole functions, and rebuilding teams around shared capabilities. The shift prioritizes human judgment and customer outcomes while using AI agents to handle complex coordination and code generation across systems.

Applied Materials and Lam Research are positioned as overlooked opportunities in chip equipment manufacturing despite broad gains across the AI chip sector. Both companies supply critical machinery for chip etching and deposition used in memory chips, which are essential components for AI systems. As memory chip makers rapidly expand manufacturing capacity, demand for their equipment has surged, yet these suppliers have not received the same investor attention as chip designers and manufacturers.

Prime Intellect, a San Francisco-based inference provider valued at $1 billion, is acquiring Moondream, a startup with fewer than 10 employees known for open-source vision-language models. The deal is primarily motivated by Moondream's engineering expertise in optimizing model efficiency rather than its existing product. Prime Intellect aims to differentiate itself in the crowded inference market by making open-source models run faster and cheaper on Nvidia servers.