Technology briefing · 26 Jul 2026

AI's next leap is built in data centres — but control may spread out

NVIDIA and SK Group announced an AI infrastructure plan worth more than $500 billion, while Microsoft published a broad letter supporting open-weight models. Together, the stories show AI advancing through both enormous compute factories and models that can be controlled closer to the user.

A generic AI data centre connects to an open model network and local computers

1. NVIDIA and SK Group plan a $500-billion-plus AI build-out

On 24–25 July, the companies described a plan covering an SK Telecom NVIDIA Vera Rubin AI factory of up to 2 gigawatts and a long-term SK hynix partnership for next-generation HBM memory. The first AI factory is planned for 2027. These are letters of intent and company plans, not an operating consumer service.

2. Microsoft letter backs downloadable and modifiable AI models

Microsoft published its Open Weights and American AI Leadership letter on 24 July. Signatories include Google, Meta, NVIDIA, OpenAI, GitHub and the Linux Foundation. The letter defines open-weight models as models that can be downloaded, inspected, modified and run on an organisation's own infrastructure. It also acknowledges that released weights are difficult to trace or reverse once modified.

Sources