Skip to content
AI Usage3 min read

Oracle to Power Project Jupiter with 2.45 GW of Bloom Fuel Cells

Oracle and BorderPlex pivot to a massive water-efficient Bloom fuel cell microgrid to power their New Mexico AI campus, bypassing grid constraints.

AB

Author

AUG Bot

Published

Aerial view of a data center campus with fuel cell infrastructure

Oracle to Power Project Jupiter with 2.45 GW of Bloom Fuel Cells

New Mexico AI campus pivots to water-efficient microgrid to reduce local grid and environmental impact

Oracle and BorderPlex Digital Assets have announced a major shift in the power strategy for "Project Jupiter," a massive AI data center campus in Doña Ana County, New Mexico. The facility will now be powered by up to 2.45 GW of Bloom Energy fuel cell capacity, moving away from traditional grid-dependent and combustion-based power sources.

Key details

The updated design replaces previously planned gas turbines and diesel generators with a massive on-site microgrid powered by Bloom Energy’s solid oxide fuel cell technology. This shift represents one of the largest private power deployments of its kind, providing up to 2.45 GW of installed capacity.

The environmental and resource implications are significant:

  • Water Efficiency: The fuel cells utilize a non-combustion process that requires a negligible amount of water compared to traditional cooling and power generation methods.
  • Emission Reductions: The new microgrid is expected to reduce nitrogen oxide (NOx) emissions by approximately 92% compared to the originally planned gas turbines.
  • Grid Independence: By operating as a single microgrid campus, the project aims to preserve local electricity rates for residents by avoiding direct competition for existing utility capacity.

Why this matters

The pivot to fuel cells addresses two of the most critical bottlenecks in AI infrastructure: energy availability and water scarcity. In the arid environment of southern New Mexico, the ability to scale compute capacity while minimizing water consumption is a vital prerequisite for long-term operational viability and community acceptance.

Context

As AI workloads demand unprecedented levels of power, data center operators are increasingly forced to become their own power utilities. Project Jupiter follows a broader trend where hyperscalers are exploring alternative energy sources—ranging from small modular reactors (SMRs) to advanced fuel cells—to bypass grid constraints and meet sustainability targets.

What happens next

BorderPlex will continue the build-out of the Project Jupiter infrastructure in the New Mexico desert. The deployment of the Bloom fuel cells will likely serve as a high-profile test case for whether massive AI campuses can successfully decouple from local utility grids while maintaining the high reliability required for next-generation model training.


Source: Oracle Published on AI Usage Global, author: AUG Bot

Related

Read more

More posts that expand on the topics, companies, and AI trends covered in this story.

US EPA headquarters and digital representation of AI data center infrastructure
AI Usage

US EPA Declines National Standards for AI Data Centers

The US EPA announces it will not set national standards for data center resource consumption, leaving regulation of AI's water and energy footprint to the states.

Digital representation of high-density memory chips and AI infrastructure capital investment
AI Usage

Micron Shares Surge 11.7% as AI Storage Demand Hits Critical Bottleneck

Micron stock jumps as AI infrastructure demand drives a memory supply shortage, with prices projected to rise through 2027 as hyperscalers lock in long-term contracts.

Digital representation of a hyperscale data center and water efficiency metrics for Amazon
AI Usage

Amazon Reports 2.5 Billion Gallons of Data Center Water Use

Amazon's first water disclosure reveals 2.5 billion gallons used in 2025 and a 52% efficiency gain since 2021 through an air-cooling first strategy.