Nvidia reportedly has more than $170 billion on the line for AI cloud expansion · Digg

Google and Meta open-sourced TorchTPU, enabling PyTorch to run on Google's TPUs. Tests show TPUs are cost-effective for batch tasks but slower in latency-sensitive applications. Meta uses TorchTPU in production and has invested in both Google TPUs and Nvidia GPUs. Nvidia still leads in latency and FP4 support.

Wayve, Stellantis, and NVIDIA partnered with the French Government to advance autonomous vehicle deployment in France and the EU. The collaboration integrates Stellantis' L4-Ready platforms, Wayve's AI Driver, and NVIDIA's Halos OS on Hyperion architecture, covering the full value chain from chips to finished vehicles.

Nvidia Corp. is considering a $2,000 annual software fee per GPU for AI agents, according to CEO Jensen Huang. This could generate $10B in annual revenue if 5M GPUs are licensed. Nvidia already offers AI Enterprise at $4,500 per GPU, but it's unclear if the new fee is a replacement. The move aligns with Nvidia's shift towards software and recurring revenue streams.

Nvidia CEO Jensen Huang proposed a $2,000 annual GPU software fee for AI agents, aiming to expand its software business. The fee's scope and timing remain unclear. Nvidia's existing AI Enterprise software costs $4,500 per GPU annually. The company also launched OpenShell Reference Architecture for secure AI agent management.

Meta reported a 26% revenue increase to $64.6B, but negative free cash flow of -$2.1B. The company raised its 2026 capital spending forecast to $135B-$145B and announced new non-voting shares. Executives from Broadcom, BlackRock, and Nvidia expressed concerns about capital scarcity for AI development. AI-linked stocks like Nvidia, CoreWeave, and Lam Research fell, while broader market indices showed mixed performance. Treasury yields also declined.

Amkor Technology (AMKR) stock rose after securing a $1.3 billion deal with NVIDIA (NVDA) to expand semiconductor packaging capacity in South Korea, bringing total agreements to $3 billion. The deal supports NVIDIA's AI supply chain, with Amkor planning to increase interposer wafer capacity by over 300% by 2028.