OpenAI Says Stargate Has Already Surpassed Its 10-Gigawatt US Compute Goal

OpenAI is framing compute as the central infrastructure problem of the AI era

OpenAI says it has already surpassed the 10-gigawatt U.S. AI infrastructure milestone that Stargate was originally meant to secure by 2029. In a new policy-oriented update, the company said that more than 3 gigawatts of capacity have been added in the last 90 days alone, a pace it presents as evidence of just how quickly demand for AI compute is rising.

The announcement matters because it reframes Stargate from a long-range aspiration into an active buildout campaign. When OpenAI introduced the initiative in January 2025, the commitment was to secure 10 gigawatts of AI infrastructure in the United States by the end of the decade. Just over a year later, the company says that threshold has already been crossed, and that it is now evaluating additional sites across the country as it plans beyond the original target.

Compute as the bottleneck

OpenAI's argument is direct: more people using AI means more compute is required, and the only responsible response is to bring more capacity online faster. The company describes compute as the critical input behind training stronger models, serving them reliably, improving performance, lowering costs over time, and widening access. It also presents compute as the center of an economic flywheel in which more infrastructure enables better models, which drive more usage, which in turn support more reinvestment.

That framing is important because it places physical infrastructure, not only software progress, at the center of the AI story. In other words, model capability is now being tied explicitly to power availability, data-center construction, supply chains, financing structures, and local permitting. OpenAI is not claiming that infrastructure is merely supportive. It is arguing that infrastructure is the constraint.

A partner-heavy buildout

The company also emphasizes that the effort is intentionally partner-centric. It says no single company can build the infrastructure for what it calls the Intelligence Age alone, and that success will require coordination across utilities, energy providers, chipmakers, cloud providers, neoclouds, construction firms, investors, skilled trades, public-sector actors, and local communities.

That ecosystem language serves two purposes. Operationally, it reflects the reality that multi-gigawatt AI campuses cannot be built by model developers acting in isolation. Politically, it signals that OpenAI wants Stargate to be seen as national-scale infrastructure development rather than a narrow corporate expansion plan. The company says partnership structures and financing models may evolve, but the core goal remains the same: capacity that comes online at scale, on time, and with flexibility as technology changes.

Why the timeline matters

Surpassing a 2029 target in 2026 is not just a symbolic milestone. It suggests the underlying demand curve has steepened enough that previously ambitious infrastructure plans are now treated as baseline requirements. OpenAI says demand is accelerating across consumers, businesses, developers, and governments. If that assessment is accurate, the next several years of AI competition may hinge less on who announces the boldest model roadmap and more on who can secure power, land, equipment, and build speed.

The statement also hints at continued geographic expansion. OpenAI says it and its partners are evaluating potential data-center locations across the country as planning moves beyond the initial 10-gigawatt goal. That means the current milestone is being used less as a finish line than as a platform for another round of siting and capacity growth.

The broader industry message

OpenAI's update lands in a year when AI infrastructure has become one of the most contested parts of the technology stack. Demand for chips, power, cooling, and data-center space is rising while governments and utilities face growing pressure to balance industrial development, grid reliability, and community concerns. Against that backdrop, a claim of already surpassing 10 gigawatts in the United States is meant to convey momentum and seriousness at a scale few AI companies can match.

Whether that pace can be sustained is a separate question. But the message is unambiguous. OpenAI is betting that the future of advanced AI will be determined as much by successful infrastructure execution as by breakthroughs in model design. In that view, compute is not a backend detail. It is the foundation the rest of the industry will stand on.

This article is based on reporting by OpenAI. Read the original article.