Project Stargate. OpenAI, Oracle, and Softbank Team Up for $500B AI Infrastructure Initiative.

OpenAI, SoftBank, and Oracle have formed a joint venture called the Stargate Project, initially committing $100 billion—and potentially up to $500 billion over four years—to build AI data centers across the U.S. This massive effort, which will start in Abilene, Texas, aims to create “hundreds of thousands” of jobs and bolster “American leadership in AI.” According to the companies, SoftBank will take on financial responsibility while OpenAI leads operationally; Masayoshi Son, the head of SoftBank, is set to chair the venture. Oracle, Nvidia, Microsoft, Arm, and Middle East AI fund MGX are also involved, underscoring the scale and significance of the initiative.

These new facilities will cater to OpenAI’s escalating demands for AI compute, with rumors suggesting that OpenAI may eventually deploy its own custom chips built in partnership with Broadcom and TSMC. Microsoft is a key partner as well, having collaborated closely with OpenAI to develop large-scale infrastructure and supercomputers for AI. The Abilene site, expected to reach nearly a gigawatt of power by mid-2026, will be the first in a series of planned data centers around the country—a buildout that could cost several billions of dollars per site.

Although some critics argue that data centers can strain environmental resources and often employ fewer workers than promised, the drive to expand AI infrastructure shows no sign of slowing. With estimates suggesting that AI could account for 19% of data center power demand by 2028, investors and tech giants alike see an urgent need for more compute. OpenAI’s CEO Sam Altman has advocated for reducing regulatory and bureaucratic hurdles to accelerate construction, aligning with President Donald Trump’s push for large-scale investments in U.S. infrastructure, including advanced facilities like those under the Stargate Project.

🚀 Supercharge Your Workday with ChatGPT: Free Ultimate Guide

Transform your productivity overnight with our comprehensive guide to mastering ChatGPT in the workplace. Join thousands of professionals who have already revolutionized their workflow with this game-changing resource.

Inside this free guide, you'll discover:

• A crystal-clear roadmap to understanding and implementing ChatGPT

• Real-world applications that instantly boost your efficiency

• 100 ready-to-use prompts for tasks like email writing and data analysis

• Expert strategies for overcoming common AI challenges

Whether you're new to AI or looking to enhance your existing skills, this guide delivers practical solutions for modern workplace demands.

OpenAI’s agent tool may be nearing release

OpenAI might soon release "Operator," an AI tool capable of autonomously performing tasks on a user’s computer. References to Operator were recently uncovered in OpenAI’s macOS ChatGPT client and website. While hidden for now, these findings align with earlier reports from Bloomberg about Operator being an “agentic” system designed to handle tasks like coding and booking travel. Some internal benchmarks comparing the Operator's performance to other AI systems were also found on the OpenAI website, suggesting the tool is promising but not yet fully reliable across all tasks.

Benchmarks reportedly show the Operator performing well on some tests, such as website navigation, but struggling in areas like signing up for cloud services or creating Bitcoin wallets, where it succeeds only 60% and 10% of the time, respectively. On OSWorld, a benchmark mimicking real computer environments, the "OpenAI Computer Use Agent" (likely powering Operator) scored 38.1%, surpassing rivals but still trailing human performance at 72.4%. While Operator appears competitive in certain web tasks, its performance varies, highlighting the challenges in creating a robust, general-use AI agent.

The imminent release of Operator comes amid increasing competition from companies like Anthropic and Google, all aiming to dominate the nascent AI agent market projected to reach $47.1 billion by 2030. However, concerns remain over the safety of such tools. Leaked charts indicate Operator has undergone extensive safety evaluations, but critics argue OpenAI has prioritized commercialization over caution. OpenAI co-founder Wojciech Zaremba recently criticized competitors for releasing inadequately tested agents, emphasizing the risks of rapid technological advancements without sufficient safeguards.

Trump revokes Biden executive order on addressing AI risks

U.S. President Donald Trump on Monday repealed a 2023 executive order issued by Joe Biden that required developers of AI systems, especially those posing national security or economic risks, to report the results of safety tests before releasing those systems to the public. Biden’s directive had aimed to address concerns over AI’s potential impact on consumers, workers, and infrastructure, particularly in fields like cybersecurity and public health. The 2024 Republican Party platform had vowed to remove that requirement, arguing it restricts innovation and free speech.

Biden’s now-revoked order also instructed federal agencies to set standards for testing AI systems and to tackle threats involving chemical, biological, radiological, nuclear, and cybersecurity issues. Meanwhile, the U.S. Commerce Department recently imposed new limits on exporting AI chips and related technology, a move that drew criticism from industry leaders including Nvidia. Generative AI, with its ability to create text, photos, and videos from basic prompts, has generated both enthusiasm over its capabilities and anxiety about job displacement and other unintended consequences.

Although Trump has rescinded Biden’s AI oversight order, he left intact another Biden directive from last week that addresses the high energy demands of advanced AI data centers. That second order calls for federal support and for granting access to select sites owned by the Departments of Defense and Energy. As U.S. lawmakers continue to wrestle with passing comprehensive AI legislation, the federal government’s role in regulating and fostering AI development remains at the forefront of public debate.

Other stuff

• The 2025 state of AI Development

• The second wave of AI coding is here

• Moore’s law may be coming to an end. What happens to AI progress if it does?

• Perplexity launches Sonar, an API for AI search

• NVIDIA Project DIGITS Explained: AI Power in a Compact Package

• Friend delays shipments of its AI companion pendant

• Paul Schrader Thinks AI Can Mimic Great Storytellers: ‘Every Idea ChatGPT Came Up with Was Good’

• Google invests further $1bn in OpenAI rival Anthropic

• Microsoft is no longer OpenAI’s exclusive cloud provider

Hope you enjoyed today's newsletter

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.

OR