Hyperscale Data is pushing forward with plans to lease AI compute capacity from its Michigan Data Center Campus, marking the latest step in a broader transformation from crypto miner to infrastructure-as-a-service provider. The company is evaluating long-term expansion options for a facility that could eventually consume 340 MW of power.
From mining rigs to AI racks
Hyperscale Data’s Michigan campus spans 600,000 square feet across 34.5 acres. It currently operates at roughly 30 MW of capacity, generating revenue through digital asset mining and colocation services.
Expansion plans call for scaling to 70 MW by the second quarter of 2027, with a full 340 MW buildout targeted for the third quarter of 2029. To fuel that growth, Hyperscale Data reached agreements in February 2025 for 300 MW from an electric utility and an additional 40 MW from a natural gas utility.
The company is also building out its GPU cloud offering. An on-demand NVIDIA GPU cloud platform is set to launch in the first half of 2026, providing access to NVIDIA H100, B200, and B300 models for AI training, inference, and high-performance computing workloads.
Hyperscale Data already operates NVIDIA GPU clusters for an existing Silicon Valley cloud provider.
Capital and partnerships
Hyperscale Data has filed a $125 million shelf registration to facilitate the capital needed for development.
On the partnership front, the company has teamed up with AGIBOT PTE. LTD. for AI and robotics operations.
Hyperscale Data’s expansion plans include hiring more than 500 employees over the next three years.
What this means for investors
The risks are equally real. A $125 million shelf registration means dilution is coming. The timeline from 30 MW today to 340 MW by Q3 2029 involves substantial execution risk, both in construction and in actually filling that capacity with paying customers. Investors should watch for announcements about anchor tenants, contract terms, and actual revenue from GPU cloud services once the platform goes live in 2026.