The next great technological revolution is unfolding, and OpenAI is at its helm- leading an ambitious new initiative called Stargate. Announced as a groundbreaking collaboration between OpenAI, Oracle, and SoftBank, the project aims to build a network of AI supercomputing data centers. With an initial commitment of $100 billion, potentially increasing to $500 billion, the partnership seeks to create the infrastructure necessary to support large-scale AI innovation.
While Sam Altman, CEO of OpenAI hailed the initiative as “the most important project of this era”, the project’s feasibility remains uncertain. Construction on the first data center in Abilene, Texas, is already underway, but significant financial and logistical challenges loom over its success.
Funding Uncertainty and Investor Skepticism
In her article ‘Sam Altman’s Stargate is science fiction’, Kylie Robison questions the feasibility of Stargate, highlighting funding concerns and logistical challenges. While OpenAI and its partners have committed $100 billion, only $45 billion has been secured so far, leaving a significant gap and raising concerns among critics, while investors remain cautious. According to a report by the Financial Times, the proposed infrastructure project to develop data centers in the United States lacks a fully formulated plan and has yet to secure the necessary funding.
Compounding these concerns, OpenAI continues to burn through billions without a stable business model. OpenAI is projected to incur a loss of approximately $5 billion in 2024, with revenues estimated at $3.7 billion. The company faces intense competition from other AI research firms and tech giants striving for market dominance. The emergence of competitors like China’s DeepSeek, which has developed an OpenAI-level model with fewer resources, poses a significant challenge to OpenAI’s leadership in the industry.
The Power Crisis of AI Data Centers
Beyond funding, one of the biggest obstacles to Stargate’s success is energy consumption. AI operates through massive data centers, not the “cloud,” and the power demands of modern AI models are skyrocketing. Kylie Robison also emphasizes the immense energy requirements of AI infrastructure, noting that current data centers may not be equipped to handle the power demands of AI GPUs. Experts warn that the transition to these facilities will not be easy, fast, or cheap.
Further expanding on these concerns, Lynette Bye, in her article “Can We Build a Five Gigawatt Data Center?“, discusses the industry-wide challenge of AI’s growing energy demands. By 2030, AI data centers could require the power equivalent of an entire city. As AI labs outgrow existing infrastructure, companies like Microsoft are planning facilities ranging from 1 to 5 gigawatts, a scale previously unheard of. However, these projects face major hurdles in terms of space, cost, chip production, and energy supply.
Constructing a 5-GW data center could cost over $100 billion, and chip shortages remain a critical bottleneck. But the biggest issue is power availability. At present, Goldman Sachs Research estimates the power usage by the global data center market to be around 55 gigawatts (GW). This is comprised of cloud computing workloads (54%), traditional workloads for typical business functions such as email or storage (32%), and AI (14%)
A High-Stakes Gamble
OpenAI is already scouting locations across the U.S. to expand its Stargate project, considering sites in 16 states to accelerate development. While CEO Sam Altman is securing political and corporate alliances, the challenges ahead are monumental.
Despite promises of medical breakthroughs and advancements in national security, the success of Stargate hinges on raising an unprecedented amount of capital and solving the power crisis, placing additional strain on an already burdened power grid. If OpenAI and its partners fail to overcome these barriers, Stargate may not be the future of AI—but rather its most expensive miscalculation.
