Six months after its White House announcement, the massive “Stargate” project aimed at boosting America’s AI ambitions is facing major difficulties and disputes between key partners SoftBank and OpenAI, leading to a sharp scaling back of its plans.
Article contents:
- Introduction: From a flashy announcement to a harsh reality
- What is the “Stargate” project?
- What went wrong? Disputes and scaled-back plans
- Altman moves forward without SoftBank
- The public facade: A “fantastic partnership”
- Masayoshi Son’s faltering ambition
- Conclusion: Massive ambitions face a complex reality
Introduction: From a flashy announcement to a harsh reality
In January, Japanese billionaire Masayoshi Son and Sam Altman stood side by side with President Trump at the White House to announce “Stargate,” a $500 billion effort aimed at accelerating United States AI ambitions. But six months later, reports reveal that this ambitious project has struggled to get off the ground and has significantly scaled back its near-term plans.
What is the “Stargate” project?
“Stargate” was designed as a joint venture between SoftBank (led by Son) and OpenAI (led by Altman) to build the physical infrastructure needed for the next generation of artificial intelligence: massive data centers. The two companies pledged to invest $100 billion “immediately” out of a total $500 billion by 2029, to build the equivalent of 5 gigawatts of data center capacity this year alone.
What went wrong? Disputes and scaled-back plans
According to people familiar with the matter, the newly formed company created to manage the project has not been able to finalize a single data center deal yet. The primary reason is disagreements between SoftBank and OpenAI over crucial partnership terms, including site locations. As a result, this year’s ambitious target has been scaled back. Instead of investing $100 billion, the goal is now much more modest: building a single small data center by the end of the year, likely in Ohio.
Altman moves forward without SoftBank
Amidst these struggles, Sam Altman—who is desperate to secure the computing power needed to support future generations of ChatGPT—has not waited. He has moved forward and signed data center deals with other companies. The most prominent of these was with Oracle, where OpenAI will pay the software and cloud computing company more than $30 billion annually starting in three years. This deal alone accounts for 4.5 gigawatts of capacity, coming close to the target Stargate was aiming for this year.
Surprisingly, Altman still uses the “Stargate” name on projects not funded through the SoftBank partnership, even though the trademark is owned by SoftBank.
The public facade: A “fantastic partnership”
Despite these internal difficulties, the leaders of both companies are trying to project that everything is fine. Last week, they appeared in a video during a SoftBank event, where Altman stated they have an initial goal to build 10 gigawatts of data centers together, describing it as a “fantastic partnership.” In a joint statement, the companies said they are “moving at lightning speed to deliver the AI infrastructure that will power the future and serve humanity.”
Masayoshi Son’s faltering ambition
The slow rollout of Stargate represents a setback for Masayoshi Son’s sweeping ambitions. Despite spending billions over the past few years, he has largely played catch-up in the fast-evolving AI sector. His commitment to invest $30 billion in OpenAI earlier this year was the largest startup investment ever, a massive bet that has driven SoftBank to take on new debt and sell assets. The Stargate project was supposed to grant him a major role in physical AI infrastructure, which has yet to materialize.
Conclusion: Massive ambitions face a complex reality
The story of Stargate is a reminder that massive announcements and ambitious promises in the tech world often face a much more complex reality. Building data centers is no easy task; it requires finding locations, developing structures, purchasing expensive chips, and securing massive amounts of electricity. It appears that disputes between the two main partners have stalled one of the most ambitious projects in AI history, at least for now.
Frequently asked questions
Does this mean the project has completely failed?
Not necessarily. It means it is facing significant difficulties in getting off the ground and short-term goals have been severely scaled back. It could still be revived in the future, but currently, it is not delivering on the promises initially announced.
Why does OpenAI need all this computing power?
Training better AI models requires heavier data processing. Every new generation of ChatGPT is larger and more complex, thus requiring immense computing power to train and operate.
Who are the other partners mentioned in the initial announcement?
The UAE’s MGX was mentioned as a partner, but its role remained unclear. The CEO of Oracle recently stated, “Stargate has not been formed yet,” underscoring the uncertainty surrounding the project.
