The Second Act of AI Is Physical: Inside the Great Data-Center Buildout

For the operations and capital-projects leaders who read this publication, the most important story in artificial intelligence is no longer about the models. It is about the concrete, steel, silicon, and electricity required to run them. The AI race has become one of the largest infrastructure undertakings in modern history, and 2026 is the year it moved from announcement to construction site.

The scale of the buildout

The clearest symbol is Stargate, a data-center venture backed by OpenAI, SoftBank, Oracle, and Abu Dhabi’s MGX. Announced in early 2025 with a plan to invest up to 500 billion dollars in roughly ten gigawatts of computing capacity in the United States, it moved quickly from concept to steel. Its first campus came online in Abilene, Texas, and by late 2025 newly announced sites had pushed planned capacity toward seven gigawatts and committed investment past 400 billion dollars.

Running in parallel, OpenAI and NVIDIA announced a partnership to deploy at least ten gigawatts of NVIDIA systems, representing millions of GPUs, with NVIDIA intending to invest up to 100 billion dollars progressively as each gigawatt comes online. The first phase is targeted to go live in the second half of 2026 on NVIDIA’s Vera Rubin platform. As OpenAI chief executive Sam Altman put it, everything starts with compute.

Why this is an operations story

A gigawatt is power-plant-scale energy. Multiply that by ten, or by the tens of gigawatts now planned across the industry, and the constraints stop being about algorithms and start being about the fundamentals of large capital projects: land, power generation and grid interconnection, water for cooling, long-lead equipment like transformers and switchgear, and a skilled construction workforce that is already stretched thin. The companies leading this effort, from NVIDIA’s Jensen Huang to SoftBank’s Masayoshi Son and Oracle’s Larry Ellison, are effectively running some of the largest coordinated construction and procurement programs on the planet.

That makes the buildout a case study in the disciplines this publication covers. Securing chips, power, and cooling at this scale is a supply-chain resilience problem of the first order, and the capital commitments involved demand exactly the kind of rigorous business impact analysis that underpins any major project. As these facilities fill with AI systems, securing them becomes its own discipline, which we cover in our look at the Open Secure AI Alliance.

More coverage is in our Operations and Project Delivery section. The NVIDIA and OpenAI partnership is detailed in NVIDIA’s announcement.

Frequently asked questions

What is Stargate?

Stargate is a large AI data-center venture backed by OpenAI, SoftBank, Oracle, and MGX, announced in 2025 with plans to invest up to 500 billion dollars in about ten gigawatts of U.S. computing capacity. Its first campus opened in Abilene, Texas.

How much is being invested in AI data centers?

Commitments run into the hundreds of billions of dollars. Stargate alone has planned investment past 400 billion dollars, and NVIDIA intends to invest up to 100 billion dollars in its systems partnership with OpenAI as capacity is deployed.

Why is AI now considered an infrastructure story?

Running large AI models requires enormous computing power, which in turn requires vast amounts of land, electricity, cooling, and specialized equipment. Building that capacity is a major construction, procurement, and supply-chain challenge, not just a software one.