North Dakota-based Bakken Energy plans to deploy a fleet of portable AI data centers—micro data centers—across North Dakota’s oil fields in a partnership with Armada, a startup that builds small, stackable computing modules designed for remote industrial sites. The units, prefabricated containers 20 to 40 feet long, can be trailered in, spun up in roughly half a day, and packed out in about two hours, according to Armada. Multiple modules can be stacked into a larger cluster, with combined capacity that the company says can top 100 megawatts of compute.
North Dakota’s approach reflects a broader market shift. Micro data centers, which are factory-built compute blocks designed for near-source processing, are growing quickly as enterprises push AI inference and data analytics to the edge. Market estimates forecast compound growth rate for micro data centers at approximately 29% through 2032. In parallel, the broader edge computing market saw revenues between $16.8 and $33.9 billion in 2023, and is expected to expand at roughly 28% CAGR through 2032.
The revenue driver is compelling: billions of IoT sensors, cameras, and industrial controllers now generate data that is too latency-sensitive or bandwidth-heavy to ship first to centralized regions.
The Advantage of Proximity
For Bakken Energy, the draw is bringing data analytics to the oil wellhead. Oil field operations are highly data intensive, from monitoring methane leakages to tracking safety protocol. But spinning up a full-fledged data center isn’t practical in this use case. Also, the micro data center are largely off-grid. Instead of using fiber networks, the sites will rely on Starlink for connectivity.
Cooling will rely on air and a water-based coolant, though Armada declined to specify volumes or sourcing. On power, Bakken says that site data supports production while locally generated electricity (potentially from gas that would otherwise be flared) feeds the computer racks.
The initiative arrives as North Dakota leaders court AI infrastructure. At last week’s Petroleum Council conference, Gov. Kelly Armstrong touted potential new synergies between Silicon Valley and the state’s operators around using natural gas to power AI. Armada, which in July disclosed $131 million in funding from sources that include Microsoft, is pitching these field systems as a faster path to AI adoption in heavy industry.
Still, the deployment is in early days: There are currently no micro data centers operating in the state. Questions remain about siting, volume, regulatory treatment, and customers. Bakken says it will supply power, handle permitting, and line up buyers, while Armada will build and maintain the facilities. Pricing is unspecified and may vary by configuration.
Small is Beautiful
Local reception will likely play a major role in whether the units succeed. North Dakota’s recent experience with large data campuses has been mixed. A site in rural Ellendale generally was seen positively for revitalizing a small community, though residents voiced concerns. In contrast, a project near Harwood drew heavy scrutiny over planning transparency and environmental impact. Bakken and Armada argue their smaller footprint, which uses existing industrial land and is sized far below hyperscale, will avoid those challenges.
In sum, these micro data centers fly under the radar and likely won’t incur the resident pushback of full-scale facilities.
A near-term milestone will be the first pilot: exactly where the micro facilities land, how their power sources perform, and how these tiny data centers handle the winter freeze. If the units deliver measurable efficiency gains without the friction of a full-scale campus, North Dakota’s patch could become a test bed for containerized AI at the energy edge.
