OpenAI and NVIDIA announce strategic partnership to deploy 10 gigawatts of NVIDIA systems

OpenAI and Nvidia have locked in a strategic partnership to deploy 10 gigawatts of datacenter capacity, all built on Nvidia's systems. The first phase is targeted for 2026, which in infrastructure terms is basically tomorrow.

To put that number in perspective, 10 gigawatts is roughly the output of several nuclear power plants. That's the kind of energy footprint we used to associate with entire cities, not a single company's compute buildout. OpenAI has been signaling for months that its ambitions outstrip anything currently on the ground, and this deal is the clearest evidence yet of what "unprecedented scale" actually means in dollars, watts, and steel.

The move also cements Nvidia's position as the indispensable supplier at the center of the AI boom. Every major lab needs chips, but few can promise a buyer at OpenAI's scale locking in gigawatt-level commitments years out. For Nvidia, this isn't just another order — it's a long-term anchor tenant for whatever comes after Blackwell and beyond.

What's notably absent from the announcement is much detail on financing, siting, or power sourcing — the parts that tend to make or break megaprojects like this. Permitting alone for gigawatt-scale power draw can take years, and the grid capacity to support it doesn't just appear because two companies shook hands. The 2026 timeline for phase one suggests some groundwork is already underway, but the gap between announcing 10 gigawatts and actually energizing it is where these things usually get complicated.