•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Bitmine Immersion Technologies, the NYSE-listed mining firm led by Tom Lee, has purchased an additional 20,000 ETH in a single over-the-counter transaction with FalconX. The purchase was valued at roughly $35.85 million, according to blockchain data tracked by Arkham Intelligence.
This latest FalconX deal forms part of a broader pattern of repeated 20,000 ETH batch purchases throughout 2026. Bitmine has used OTC desks to avoid placing large buy orders directly on the open market.
Bitmine’s Ethereum treasury has grown to over 5.5 million ETH. The company’s holdings represent more than 5% of Ethereum’s total supply. Approximately 85% of these holdings are staked through what Bitmine calls its Made-in-America Validator Network (MAVAN).
The staking operation is projected to generate annual revenue of approximately $230 million.
Bitmine’s buying has not been limited to FalconX. The company has also routed purchases through Kraken and Galaxy Digital, diversifying counterparty exposure across major institutional platforms.
With 85% of its 5.5 million ETH holdings staked through validators, Bitmine has pulled a significant portion of circulating supply off the market. A single corporate entity holding this scale of a Layer 1 asset also raises governance and decentralization concerns.
Ethereum’s validator set is designed to be distributed. When one company runs enough validators to represent a material percentage of staked ETH, it can introduce questions about network resilience.
Bitmine internally refers to its accumulation target as the “Alchemy of 5%,” aiming to hold enough ETH to meaningfully influence network dynamics and maximize staking returns. The company has already crossed that threshold.
The approach echoes MicroStrategy’s Bitcoin treasury strategy, but applied to Ethereum with an added revenue angle from staking. If Bitmine’s projected $230 million in annual staking revenue materializes, the article suggests other public companies may explore similar models.
