Microsoft claims Crusoe’s new 900 MW DC campus, boosting its data‑center portfolio.

Crusoe’s latest expansion adds two data halls, each rated for 336 MW of critical IT load, and a dedicated on‑site power plant capable of delivering a full 900 MW of behind‑the‑meter electricity to the facilities. According to The Register, the power plant will be the primary source of energy for the new campus, although the company has not disclosed the generation technology; given the aggressive timeline—operational by mid‑2025—natural‑gas turbines or fuel‑cell generators are the most plausible options (The Register). The on‑site generation strategy is intended to sidestep the chronic grid‑capacity constraints that have plagued large‑scale AI deployments in the United States, allowing Microsoft to provision a stable, low‑latency power supply for its AI workloads.

The new campus will sit adjacent to Crusoe’s 1.2 GW “Stargate” complex, a facility originally earmarked for Oracle and OpenAI under a $500 billion initiative announced early last year (The Register). While Oracle and OpenAI had initially planned to lease the additional capacity, negotiations stalled, opening a slot that Microsoft has now filled. The Register notes that the combined capacity of the two campuses will reach roughly 2.1 GW, though only about 200 MW of the original 1.2 GW is currently online; the remainder is slated to come up throughout 2026. This phased rollout means that Microsoft’s AI models will initially rely on the newly built 900 MW segment while the broader infrastructure ramps up.

From a networking perspective, the co‑location of Microsoft, Oracle, and OpenAI creates a high‑density AI hub that can exploit low‑latency interconnects and shared fiber pathways. The Register reports that work on the second campus—designated for Microsoft’s AI services—is still in the “land‑clearing and site preparation” phase, suggesting that the physical layout will incorporate modern data‑center standards such as 48‑VDC power distribution, liquid‑cooling loops, and high‑density rack designs. By integrating these technologies with on‑site power generation, Microsoft can achieve power usage effectiveness (PUE) figures that rival the most efficient hyperscale facilities, a critical metric for AI training clusters where energy consumption can exceed 10 MW per rack.

The strategic implications extend beyond raw capacity. Microsoft’s partnership with Crusoe positions the tech giant to sidestep the regulatory and permitting delays that have hampered new AI‑focused data centers in other regions. The Register highlights that senators are pressing datacenter operators to be transparent about power consumption, and Microsoft’s on‑site generation could serve as a model for compliance, providing verifiable, real‑time energy data directly from the plant. Moreover, the proximity to OpenAI’s existing lease may enable tighter integration of Azure’s AI services with OpenAI’s models, reducing data‑transfer overhead and potentially lowering inference latency for enterprise customers.

Finally, the expansion underscores the broader industry shift toward “industrial AI”—large‑scale, purpose‑built facilities that combine compute, storage, and power under a single roof. Crusoe’s “bit barns” terminology reflects this trend, emphasizing the convergence of traditional datacenter design with the high‑density, power‑intensive requirements of modern large language models. As The Register points out, the 900 MW addition will bring the site’s total capacity to 2.1 GW, but only a fraction of that capacity is currently active. The staged commissioning schedule, coupled with Microsoft’s commitment to occupy the new campus, suggests that the industry’s demand curve for AI‑grade compute will continue to outpace the rollout of grid‑scale power infrastructure, making on‑site generation an increasingly attractive solution for the next generation of AI workloads.