Hyperscale Data Shifts from Bitcoin Mining to AI Data Center in Michigan
Hyperscale Data, trading under the ticker GPUS on NYSE American, has officially shut down its bitcoin mining operations at its Michigan data center as of September 1. The company has opted to redirect its resources towards a more lucrative venture: transforming the Dowagiac facility into an AI data center.
Through its subsidiary, Alliance Cloud Services, Hyperscale signed a Master Services Agreement (MSA) that is initially valued at over $1.2 billion for a ten-year term, with options for two additional five-year extensions. The pivot begins with an initial allocation of 20 MW of AI compute capacity, with plans to deploy an additional 10 MW within 90 days of contract signing, and another 10 MW thereafter. The MSA also includes an option for an extra 32 MW, potentially raising the total contract value beyond $3 billion.
The Dowagiac campus spans 617,000 square feet across 34.5 acres and has a total potential power capacity of approximately 340 MW, meaning the current AI contract utilizes only about 20% of the site's capabilities. Hyperscale had been mining bitcoin at this location since at least March 2021.
CEO William Horne explained that the decision was driven by profitability, as AI colocation opportunities present higher margins compared to bitcoin mining. Discussions regarding this transition reportedly began as early as June 2026, indicating that the company had been planning its exit from bitcoin mining for several months.
However, Hyperscale is not completely abandoning bitcoin. The company has retained some bitcoin in its treasury and plans to selectively sell portions of its holdings to fund future developments at the Dowagiac site. Recent sales of approximately 65 BTC generated around $5.1 million, which will be reinvested into the AI infrastructure.
Updated 12:32 UTC
New Developments in AI Infrastructure
Equinix has announced the launch of the Equinix Inference Exchange, set to debut in Q1 2027. This initiative aims to create a global marketplace for running AI models at the edge of enterprise networks.
The Inference Exchange will utilize Nvidia’s enterprise reference architectures and Together AI’s inference platform, allowing enterprises to run inference workloads closer to their data and end users.
Equinix’s infrastructure will support over 200 open-source models, providing flexibility without vendor lock-in. This move is part of Equinix's strategy to enhance its AI capabilities, following previous initiatives like the AI Factory program.
Equinix CEO Adaire Fox-Martin emphasized the company's history of reliable infrastructure, while Nvidia’s VP Raj Mirpuri highlighted the potential of transforming Equinix’s platform into a global fabric for AI inference.
Equinix's data centers are designed for neutrality, enabling connections to multiple clouds and networks, which is crucial for enterprises employing hybrid or multi-cloud strategies.
Updated 12:33 UTC
Recent Developments in AI and Tech Earnings
- Dell Technologies reported fiscal Q2 2027 revenue of $46.97 billion, a 58% increase from the previous year.
- Dell's AI server revenue reached $16.4 billion in the quarter, prompting an increase in its full-year AI server revenue target to $74 billion.
- Dell's adjusted earnings per share (EPS) were $7.04, significantly above the expected $4.92.
- Palo Alto Networks posted fiscal Q4 2026 revenue of $3.41 billion, exceeding the consensus estimate of $3.35 billion.
- Palo Alto Networks' adjusted EPS was $1.02, slightly above the expected $0.98.
- For FY2027, Palo Alto Networks anticipates revenue between $14.1 billion and $14.2 billion, with adjusted EPS projected at $4.16 to $4.19.
- Dell's stock rose approximately 9% in after-hours trading, while Palo Alto's initial gains faded due to margin concerns.
Updated 12:33 UTC
New Insights on AI Data Centers
- Greg Abel, CEO of Berkshire Hathaway, highlighted the growing electricity demand from AI data centers as a "significant opportunity" for the company.
- In 2025, AI data centers accounted for approximately 8% of Berkshire Hathaway Energy's total load in Iowa, with expectations of a 50% growth in AI-related demand over the next five years.
- Berkshire Hathaway Energy is currently engaged in a $34 billion capital expenditure program to support infrastructure development for these demands.
- In June 2026, Berkshire Hathaway made a $10 billion investment in Alphabet, marking a strategic shift towards technology investments under Abel's leadership.
- Abel emphasized that while Berkshire will not become an AI company, it aims to benefit from the growth of tech giants like Google and Microsoft that are expanding their AI capabilities.
Updated 13:03 UTC
Recent Developments in Semiconductor Manufacturing
- TSMC's quarterly demand for semiconductor manufacturing tools has surged to approximately 1.9 times the projections made in December 2025.
- The company is currently constructing nearly 20 wafer fabrication plants simultaneously, a pace that is four to five times its normal rate.
- TSMC aims to produce 180,000 wafers per month on its advanced 3 nm process node by the end of 2026.
- The company's capital expenditure forecast for 2026 has been raised to between $60B and $64B, with a significant portion allocated for advanced production nodes.
- TSMC has committed an additional $100B towards its facilities in Arizona, bringing its total investment in the U.S. to $265B.
- Advanced-node supply is expected to remain insufficient to meet AI-related demand through at least 2028 to 2030.
- Global semiconductor equipment sales are projected to reach $133B in 2025, $145B in 2026, and $156B by 2027, driven by foundry spending and demand for high-bandwidth memory.
FAQ
Why did Hyperscale Data stop its bitcoin mining operations?
Hyperscale Data ceased its bitcoin mining operations to redirect resources towards a more profitable venture in AI data centers, as AI colocation opportunities present higher margins compared to bitcoin mining.
What is the value of the Master Services Agreement (MSA) signed by Hyperscale?
The Master Services Agreement (MSA) signed by Hyperscale is initially valued at over $1.2 billion for a ten-year term, with options for two additional five-year extensions.
How much AI compute capacity will Hyperscale initially allocate to its new data center?
Hyperscale will initially allocate 20 MW of AI compute capacity to its Dowagiac facility, with plans to deploy an additional 10 MW within 90 days and another 10 MW thereafter.
What is the total potential power capacity of the Dowagiac campus?
The Dowagiac campus has a total potential power capacity of approximately 340 MW, meaning the current AI contract utilizes only about 20% of the site's capabilities.
Will Hyperscale completely abandon bitcoin mining?
No, Hyperscale is not completely abandoning bitcoin mining. The company has retained some bitcoin in its treasury and plans to selectively sell portions of its holdings to fund future developments at the Dowagiac site.
Comments
Comments are moderated before publish.
No comments yet — be the first.