U.S. AI Data Center Project 'Stargate' Still Not Up and Running

According to The Financial News, a claim has emerged that the artificial intelligence (AI) data center project proposed by U.S. President Donald Trump shortly after the launch of his second administration in January last year has been stalled for more than a year. The project has gone off course because of disagreements among the private companies in charge of carrying it out.
On the 23rd (local time), U.S. tech outlet The Information reported this, citing multiple sources. The outlet said the Stargate Project, which is being pursued by AI company OpenAI, Japanese tech and investment firm SoftBank, and U.S. tech company Oracle, is not functioning properly due to clashes among the three firms over how to divide roles and structure their partnership.
The three companies’ leaders had gathered at the White House on January 21 last year and announced that they would jointly invest to establish a new AI joint venture called the Stargate Project. The project’s goal was to build data centers and other infrastructure in the United States for use by AI companies. Under the initial plan, once the data centers were completed, OpenAI would use the facilities while Oracle would operate them. The Stargate Project was to set up a separate board of directors and hire a new CEO, but day-to-day management would effectively be handled by OpenAI. SoftBank was to be responsible for raising capital. SoftBank Chairman Masayoshi Son said the companies would first invest a total of 100 billion dollars in the Stargate Project immediately, and then inject an additional 400 billion dollars over the following four years. Trump pledged full support, saying he wanted to "keep AI infrastructure in the United States."
A source who spoke with The Information said the Stargate Project initially planned to invest 100 billion dollars to build computing facilities consuming 10 gigawatts (GW) of power. However, the project has reportedly not even managed to hire staff, nor has it begun developing the data centers that OpenAI was supposed to use.
Because securing computing resources had become urgent, OpenAI tried to build its own data centers. That effort was put on hold, however, as lenders questioned OpenAI’s ability to repay its debts. OpenAI ultimately changed course, abandoning the three-way joint structure and instead signing separate bilateral deals with SoftBank and Oracle, under which those companies would own the data centers while OpenAI would retain control over facility design.
As a result, OpenAI agreed with Oracle to develop data centers across the United States with a combined power capacity of 4.5 GW, and decided to work with SoftBank to build a 1 GW data center in Milam County, Texas.
In the process, OpenAI also recruited Sachin Katti, formerly chief technology officer (CTO) at Intel, to secure control over intellectual property (IP) related to data center design.
OpenAI came under significant financial strain last year because of delays in the Stargate Project. Its gross profit margin fell below expectations as it was forced to procure expensive computing resources on short notice, and its projected computing costs through 2030 were revised upward from 450 billion dollars to 665 billion dollars.


pjw@fnnews.com Park Jong-won Reporter