OpenAI's Stargate Project Scales to 10 Gigawatts as Samsung and SK Join the Build

Where will the infrastructure be built?

Five new datacenter sites have been announced across the United States, though specific locations were not disclosed in the initial announcements. These facilities are part of what OpenAI describes as a reindustrialization effort, with the company emphasizing job creation alongside compute expansion.

The Samsung and SK partnership extends Stargate's footprint to Korea, where the companies will scale advanced memory chip production and develop next-generation datacenters. This geographic diversification matters because it reduces reliance on any single region for critical AI infrastructure components.

Why does this scale matter for AI?

Training frontier AI models requires enormous amounts of compute. Each generation of models tends to demand significantly more processing power than the last. By building out 10 gigawatts of capacity, OpenAI is essentially betting that the demand for AI compute will continue to grow exponentially.

The infrastructure also serves a strategic purpose. OpenAI has framed Stargate as foundational to its long-term vision of delivering AI benefits broadly. More compute capacity means the ability to train more capable models, serve more users, and potentially run the kinds of reasoning-intensive systems the company has been developing.

What comes next?

The partnership announcements suggest Stargate is moving from planning into active construction. The Oracle agreement alone represents a significant acceleration, and the addition of Samsung and SK addresses one of the key bottlenecks in AI scaling: the availability of high-bandwidth memory chips.

OpenAI has positioned this buildout as part of advancing U.S. AI leadership, a framing that resonates with ongoing policy discussions about compute infrastructure and national competitiveness. The tens of thousands of jobs the company projects will accompany construction add a workforce development dimension to what is fundamentally a technology infrastructure play.

Whether 10 gigawatts proves sufficient for the AI systems of the next decade remains to be seen. But the scale of commitment suggests OpenAI and its partners are planning for a future where compute demand far exceeds anything available today.

James Chen · 2 hours ago · 3 min