OpenAI has quietly stepped back from building its own massive data centers, opting instead to rent computing capacity from Microsoft as it prepares for a potential initial public offering. The move marks a significant pivot from the company's earlier "Project Stargate" announcement, a $500 billion joint venture unveiled in January 2025 alongside President Trump at the White House. The decision reveals how even the world's most prominent artificial intelligence company must balance its computing ambitions with financial realities.

What Happened to Project Stargate?

When OpenAI announced Project Stargate in early 2025, the initiative promised to build a network of data centers across multiple continents to power the next generation of AI models like GPT-5 and beyond. One of the flagship projects was a massive facility in Narvik, Norway, initially branded as "Stargate Norway," which would have been the northernmost data center in the world, located above the Arctic Circle.

However, by April 2026, OpenAI had withdrawn from the Norwegian project entirely. The company also pulled out of a planned expansion in the United Kingdom and canceled plans to expand its flagship facility in Abilene, Texas. Instead of building these facilities, OpenAI now rents capacity from Microsoft, which has stepped in to become the primary customer for data centers like the one in Norway.

"OpenAI had dragged its feet on signing a formal contract with Nscale even after announcing the project publicly," according to reporting on the negotiations.

Source reporting on OpenAI-Nscale negotiations

An OpenAI spokesperson explained that the company would still use the Norwegian facility as a customer of Microsoft, operating under existing contracts that made more financial sense than direct ownership. This arrangement allows OpenAI to access the computing power it needs without the massive upfront capital investment required to build and operate data centers.

Why Is OpenAI Showing Financial Discipline Now?

The timing of OpenAI's retreat from data center construction is significant. With the company expected to file for an initial public offering, analysts suggest that OpenAI may be attempting to demonstrate financial discipline to potential investors. Building data centers requires billions of dollars in upfront spending, with returns spread over many years. By renting capacity instead, OpenAI can show more immediate profitability.

Alvin Nguyen, a senior analyst at Forrester, a global research firm, noted that the optionality helps OpenAI's financial positioning. The company can scale its computing needs up or down based on demand without being locked into massive fixed costs associated with owning and operating data center infrastructure.

How Are Other Companies Filling the Data Center Gap?

OpenAI's departure from data center ownership has created opportunities for emerging companies. Nscale, a British startup founded just two years ago, has become one of Europe's hottest technology companies, valued at $14.6 billion. The company specializes in building "neoclouds," data centers designed specifically for artificial intelligence workloads.

Nscale's rapid growth demonstrates the enormous demand for AI computing infrastructure. The company raised $2 billion in March 2026, the largest funding round of its kind in European history. It now has five data centers in various stages of construction across the United States, United Kingdom, and Norway. Investors are lining up in anticipation of a possible initial public offering later in 2026.

"I've never seen a startup take off like that before," said Nvidia CEO Jensen Huang, who invested $683 million in Nscale.

Jensen Huang, CEO at Nvidia

What Makes Data Centers So Critical to AI Development?

Data centers are the physical foundation of the artificial intelligence revolution. These massive facilities house thousands of high-performance computer chips that train and run large language models like GPT-4 and GPT-5. The computational demands are staggering; more than 800 data centers are currently under construction worldwide, on every continent except Antarctica.

The global data center build-out represents one of the largest infrastructure investments in history. The world's biggest technology companies are set to spend approximately $7 trillion on data centers by 2030, according to consulting firm McKinsey. Together, these facilities will consume roughly the same amount of electricity annually as the nation of Malaysia.

How to Understand the Data Center Economics

• Location Advantages: Northern Norway offers surplus hydroelectric power at 3 to 4 cents per unit, far below the European average of 10 cents, plus natural cooling from the Arctic climate that reduces energy costs for cooling computer chips.
• Scale Requirements: OpenAI hopes to build data centers totaling roughly six times the annual energy consumption of New York City by 2030, demonstrating the massive infrastructure needed to support advanced AI models.
• Financial Models: Companies like Nscale rely on long-term contracts averaging five years with major technology companies to cover upfront construction costs, then profit by renting excess capacity on the open market after contracts expire.

The shift in how OpenAI accesses computing power reflects broader changes in the artificial intelligence industry. Rather than vertically integrating every aspect of AI development, companies are increasingly relying on specialized infrastructure providers. This approach allows OpenAI to focus on developing and improving its language models while outsourcing the complex business of building and operating data centers.

For investors watching OpenAI's path to an initial public offering, the company's decision to rent rather than build signals a focus on profitability and operational efficiency. Meanwhile, companies like Nscale are positioned to become the critical infrastructure backbone of the AI era, much like Amazon Web Services became essential to cloud computing.