Nvidia and SK Group unveil $500B AI factory and HBM pact

This deal is a statement about where the bottleneck in the race to AGI now lies. A 2‑gigawatt AI factory and a multiyear HBM pact between Nvidia and SK essentially lock in a gigantic slice of future compute and memory capacity for high‑end AI workloads. Rather than a one‑off capex announcement, this looks like the hardware backbone for the next generation of frontier models, including agentic systems and physical AI.

Strategically, it deepens Nvidia’s role as the central co‑architect of global AI infrastructure and cements Korea as a sovereign compute hub, not just a memory supplier. For model labs, this raises the bar for what it means to be “serious” about scaling: access to multi‑gigawatt, purpose‑built AI factories, not just cloud credits. It also increases pressure on other countries to secure comparable supply chains or risk permanent dependence on a handful of foreign GPU and HBM vendors. In practice, whoever can most effectively translate this firehose of compute into robust, efficient training and deployment pipelines will gain a durable advantage in pushing toward AGI‑class systems.