Overview
- NVIDIA said its next‑generation Vera Rubin AI platform and Spectrum‑X silicon‑photonic Ethernet are in full production, positioning the company to supply large, factory‑scale AI deployments and claim major efficiency gains for data‑center networks.
- Intel’s CEO reported an uptick in customer calls seeking more CPUs and acknowledged supply limits for processors that are being pressed back into central roles by agent and model workloads.
- TSMC told shareholders it expects chip capacity to struggle to keep pace with AI demand for years even as it raises revenue guidance and boosts capital spending to expand output.
- Large capital moves are unfolding: DeepSeek is reported to be closing a roughly RMB 50 billion private financing with founder and strategic investors, and SpaceX is negotiating a record‑scale, all‑new‑share IPO that targets a very large valuation and unusually low underwriting fees.
- The near‑term squeeze broadens beyond chips to power, sites and networks, which could slow some AI projects, raise costs for data‑center buildouts, and push customers to secure supply, financing, and colocated energy capacity sooner than planned.