top of page

Bitcoin miner MARA to acquire majority stake in Exaion in AI, HPC play

 Bitcoin miner MARA to acquire majority stake in Exaion in AI, HPC play
Published date:
Source:
BB Finews
8/15/25, 4:32 PM

Bitcoin miner MARA Holdings is making its biggest AI play to date, signing a $168 million deal to buy a 64% stake in Exaion, a subsidiary of French state-owned Électricité de France, one of the world’s largest low-carbon energy producers.

The agreement, announced on Tuesday, includes an option for MARA to raise its stake to 75% by 2027 with an additional $127 million investment, contingent on performance milestones.

Exaion works with Nvidia

Exaion develops high-performance computing (HPC) data centers and provides AI and cloud infrastructure in partnership with the likes of AI heavyweight Nvidia and Big Four accounting firm Deloitte.

The agreement now positions Exaion to transition to larger-scale international deployment to serve enterprise and public-sector clients, MARA said.

The transaction is expected to close in the fourth quarter, subject to necessary approvals.

MARA’s CEO and chairman Fred Thiel added: “As data protection and energy efficiency become top priorities for both governments and enterprises, MARA and Exaion’s combined expertise would enable us to deliver secure and scalable cloud solutions built for the future of AI.”

AI expansion comes as BTC mining difficulty surges

MARA’s expansion in AI comes amid a steep rise in Bitcoin (BTC) mining difficulty, driving up energy consumption and putting pressure on miner profitability margins unless they adapt with more efficient equipment or lower energy costs.

MARA is the largest Bitcoin miner by Bitcoin production, network hash rate, and market cap, but has been one of the last players in the industry to make a significant push into AI.

Related: Michael Saylor is not sweating the rise of Ethereum treasury companies

In additional comments on Monday via X Spaces, Thiel noted that MARA “deliberately chose not to be in the first wave” of expanding into AI and HPC data center businesses like other Bitcoin miners.

“Rather than trying to retrofit mining facilities, we’re investing in a partner who already has the expertise, has the customer base and has the track record in the space,” the MARA boss said, while adding that “it allows us to move quickly, intelligently and credibly.”

MARA’s new BTC production competitor

In July, MARA mined 703 Bitcoin — a rare loss to competitor firm IREN, which posted a company record 728 Bitcoin in the same month. MARA’s drop in production was likely driven by fewer machines being active for Bitcoin mining.

However, MARA saw its revenue rise 64% year-on-year to $238 million in the second quarter, while it still holds 50,000 Bitcoin worth nearly $6 billion, making it the second-largest Bitcoin treasury after Michael Saylor’s Strategy.

Magazine: US risks being ‘front run’ on Bitcoin reserve by other nations: Samson Mow

Financial News

Cardano risks a 6.5% drop if $0.90 support breaks – Can it happen?

BB Finews

XRP dips after 470M tokens sold – $2.60 in sight IF…

BB Finews

2.9B flows into Ethereum ETFs – Is ETH’s dip a buying signal?

BB Finews

Bitcoin has ‘greater than 50% chance’ of $150K before bear hits: Exec

BB Finews

China is quietly upstaging America with its open models

BB Finews

Galaxy secures $1.4B loan to fast-track Texas Helios AI datacenter

BB Finews

DeFi revival fuels AAVE’s rally: Is $400 on the cards?

BB Finews

BtcTurk halts withdrawals after $48M crypto hack sparks panic!

BB Finews

BONK falls, Pump.fun holds fort – Can Bonk_fun claw back control?

BB Finews
  • Page 3

Disclaimer:

This article is an original work by BBFinews, with copyright owned by Jinse Finance. Unauthorized reproduction is prohibited. Authorized media must indicate: “Source: BBFinews” when using this content. Violators will be held legally accountable.

 

Risk Warning:

Investment involves risks. Please exercise caution when entering the market. This content does not constitute investment or financial advice.

bottom of page