Two publicly listed companies are liquidating crypto treasury holdings to bankroll expansion into AI infrastructure, in a sign of how corporate balance sheets built on digital assets are increasingly being redeployed toward the AI buildout. Tokyo-listed Quantum Solutions (2338) and NYSE American-listed Hyperscale Data (GPUS) both disclosed treasury moves this week tied to data center projects.
Quantum Solutions sold 1,000 ETH on July 30, 2026, at $1,903 per token, generating $1.903 million after fees. That price marks a steep 47% discount to the $3,595.02 average cost basis the company had reported for its ethereum holdings back in June.
Nearly a Third of June's Holdings Now Gone
The July 30 sale followed an earlier disposal of 904 ETH on June 16 for $1.61 million, bringing Quantum's total ethereum sold to 1,904 ETH — about 29% of the 6,668.8 ETH it held as of June. Current holdings now stand at 4,764.8 ETH.
The company's board has since raised its cumulative sale authorization to 4,375 ETH through October 30, equal to nearly 66% of the June balance, leaving room for an additional 2,471 ETH in disposals if the board proceeds. Of the ethereum still on the books, 3,050 tokens are pledged as collateral to an unnamed Singapore financial-services firm, while 1,714.8 ETH sits in a trading account.
Where the Money Is Going
Quantum says proceeds are earmarked for data center deposits, GPU servers, networking equipment, and general working capital. The company signed a nonbinding agreement in June with Hong Kong-based Integrated Capital to explore building a data center in Japan, though no investment figures or construction timeline have been disclosed yet.
Hyperscale Data Takes a Different Route
Hyperscale Data is pursuing a similar goal through a different mechanism. Rather than selling outright, the company monetized approximately 100 bitcoin and used it to establish a bitcoin-backed credit facility carrying a variable interest rate of 4.5% to 5%. The financing is intended to accelerate an AI data center project already under development in Michigan.
The two approaches illustrate a broader tension facing crypto-treasury companies as AI infrastructure demand accelerates: sell holdings outright and crystallize losses against high cost bases, as Quantum has done, or borrow against them and keep exposure intact, as Hyperscale has chosen. Both routes convert digital-asset treasuries into the capital-intensive hardware and power contracts that AI data centers require.
