The initial deployment is expected to provide 4.25 gigawatts of AI factory capacity and use Nvidia’s full-stack DSX platform, including GPUs, CPUs, networking and infrastructure software.
Nvidia chief executive officer Jensen Huang said each generation of systems deployed at the site could represent around 1.5 million Nvidia GPUs and generate an estimated $150 billion to $200 billion in Nvidia revenue.
The company could eventually secure another 3.75 GW at the site.
OpenAI has committed to substantial deployments of Nvidia infrastructure through 2030, with existing and planned commitments representing about 12 GW of Nvidia compute. That could rise to about 16 GW if Nvidia expands its PORTS-Pike arrangement.
At that scale, Nvidia estimates the opportunity at roughly $600 billion in compute through 2030.
Nvidia said it will support around 4 GW of PORTS-Pike infrastructure over a 20-year term, with its backing limited to defined portions of lease and power payments and a specified residual-value commitment.
The guarantees are expected to take effect in phases as data centers enter service from 2028 through 2030.
Huang argued that land, power and shell capacity is becoming as strategically important to AI development as advanced chips, memory, packaging and networking.
For frontier AI companies, he said, growth is increasingly constrained not by algorithms or demand, but by access to enough compute.