Bitcoin Miner Hyperscale Sells 685 BTC To Scale Michigan AI Data Center, Cuts $30M Debt

Bitcoin miners keep re-rating their balance sheets for AI-era cash flows. The latest: Las Vegas-based Hyperscale Data offloaded roughly 685 BTC for about $43 million to accelerate its Michigan data center build, while reducing debt by approximately $30 million. Post-transaction, the NYSE American-listed company (ticker: GPUS) holds about 275 BTC, down from roughly 960 BTC, and says it will continue mining and may rebuild its stack over time depending on market conditions.

What changed - Proceeds will primarily fund Michigan expansion, the cornerstone of Hyperscale’s pivot into AI infrastructure. - The sale provides flexibility across debt, equity, and broader capital structure. - Operations run through wholly owned subsidiary Sentinum, which mines digital assets and offers colocation and hosting to AI firms and other enterprises.

My take: this is a straight capital allocation trade—the “megawatt ROI” vs. “BTC beta” decision Miners increasingly view each megawatt as an asset that can be monetized more predictably by AI workloads than by hash rate alone. If you can convert a volatile, highly liquid treasury into capex that locks in multi-year AI contracts at attractive dollars-per-kilowatt, you often pull forward value and compress risk. Retiring ~$30 million of debt also lowers interest drag and reduces sensitivity to BTC drawdowns, which can matter when execution risk around new builds is non-trivial.

There’s signaling here too. Selling coins tends to spook purists, but holding 275 BTC preserves convexity if the market runs, and ongoing mining provides a path to rebuild the treasury without issuing equity at a discount. Framed this way, Hyperscale isn’t abandoning Bitcoin; it’s exchanging part of its balance-sheet beta for operating cash flows that may command a higher multiple if the AI capacity is contracted efficiently.

Technically, repurposing or expanding for AI compute shifts constraints: power density, cooling, grid interconnects, and GPU availability. Operators that already manage energy-intensive, mission-critical facilities have a head start, but execution hinges on power procurement and time-to-rack. That’s why miners with existing sites can move faster than greenfield data center entrants.

Peers are making the same calculation. Singapore-based Bitdeer sold its entire Bitcoin treasury in February to back AI data center growth. Florida-based MARA sold about $1.5 billion worth of BTC in May to fund AI infrastructure and repurchase debt. As VanEck’s digital asset head Matthew Sigel has argued, miners may be “sitting on a gold mine” because they can repurpose sites for AI while still trading at a discount to traditional data center names on a market-cap-per-megawatt basis. If that valuation gap narrows as AI revenue ramps, selling coins today to capture that spread can look rational.

What to watch next - Contract quality at the Michigan facility: term length, utilization, and $/kW. - Build timeline and capex discipline amid GPU and grid bottlenecks. - BTC accumulation pace via mining versus discretionary purchases. - Leverage metrics post-debt reduction as AI revenues scale.

If Hyperscale lands durable AI workloads without losing its exposure to Bitcoin upside, this treasury-to-infra rotation could improve the company’s risk-adjusted profile while keeping crypto optionality intact.