On January 21, 2025, the Stargate Project, a “groundbreaking” artificial intelligence initiative, was announced at the White House by U.S. President Donald Trump, OpenAI CEO Sam Altman, Oracle co-founder Larry Ellison, and SoftBank CEO Masayoshi Son. With an estimated budget of $500 billion, the initiative aims to achieve artificial general intelligence (AGI), with transformative applications like personalised cancer treatments at its forefront.
Construction for the project is already underway, with Texas hosting the initial buildout. While SoftBank will lead financing, OpenAI will head software development, and key partners include Arm, Microsoft, NVIDIA, Oracle, and OpenAI. This ambitious venture is expected to generate 100,000 jobs and marks a significant evolution from the earlier Stargate collaboration between OpenAI and Microsoft, announced in March 2024.
The previous Stargate effort focused on diversifying AI accelerator usage to reduce reliance on NVIDIA while leveraging innovative energy solutions, such as nuclear power and renewable energy from companies like Lancium and Crusoe Energy Systems. Those partnerships have since expanded, with Oracle and Blue Owl Capital supporting advanced data centre infrastructure in Texas.
A high-stakes gamble on AI dominance
The updated Stargate Project shifts Microsoft’s role to one of several technology partners while maintaining their strategic relationship with OpenAI through a modified exclusivity agreement until 2030. Initial funding of $100 billion is being pooled from stakeholders, including OpenAI, SoftBank, Oracle, and MGX, with additional financing expected from debt and other investors.
Some IDC analysts caution that this initiative may be counter-intuitive, given the AI industry’s anticipated shift away from massive Large Language Models (LLMs) toward smaller, more specialised models which require far less infrastructure, as research is focusing on sparse and low-precision models that can maintain accuracy while significantly reducing infrastructure demands. Recent developments from Chinese firm DeepSeek also suggest that substantial reductions in GPU capacity may lead to comparable AI model performance, further questioning the direction of the Stargate Project.