Futures
Access hundreds of perpetual contracts
TradFi
Gold
One platform for global traditional assets
Options
Hot
Trade European-style vanilla options
Unified Account
Maximize your capital efficiency
Demo Trading
Introduction to Futures Trading
Learn the basics of futures trading
Futures Events
Join events to earn rewards
Demo Trading
Use virtual funds to practice risk-free trading
Launch
CandyDrop
Collect candies to earn airdrops
Launchpool
Quick staking, earn potential new tokens
HODLer Airdrop
Hold GT and get massive airdrops for free
Launchpad
Be early to the next big token project
Alpha Points
Trade on-chain assets and earn airdrops
Futures Points
Earn futures points and claim airdrop rewards
Layoffs, selling coins, developing AI: MARA's transformation is just a typical example of what mining companies are doing.
Zhu Qian, Golden Finance
Abstract
On April 3, 2026, the Bitcoin miner MARA laid off 15% of its workforce to help drive the company’s strategic shift from a pure-play Bitcoin mining firm to an energy and digital infrastructure company, intensifying its push into AI infrastructure. The company had previously entered the AI compute market by acquiring a 64% equity stake in Exaion; now, with its Bitcoin mining business continuing to post massive losses and AI compute demand growing explosively, both have become the dual core forces propelling its transformation. It’s not just MARA—global miners’ AI transformation journey has long been underway……
MARA, one of the world’s largest Bitcoin mining firms (NASDAQ: MARA), cut about 15% of employees, affecting full-time workers across multiple departments and some contract workers. In an internal memo, MARA CEO Fred Thiel said the layoffs were not purely a financial decision, but part of the company’s strategic transition from a pure-play Bitcoin miner to an energy and digital infrastructure company.
This move reflects that MARA is proactively “right-sizing,” shifting resources away from traditional mining operations toward the AI space, which offers greater growth potential.
I. From a miner to a digital infrastructure provider: MARA’s transformation path
On February 26 this year, MARA Holdings, Inc. announced it had reached a strategic agreement with Starwood Capital Group (“Starwood”) and its dedicated data center development platform, Starwood Digital Ventures (“SDV”). The partnership will help MARA upgrade some of its data centers, building next-generation digital infrastructure to meet the growing needs of enterprise, hyperscale, and artificial intelligence customers.
SDV leads with design, development, tenant recruitment, construction, and facilities operations, while Starwood provides investment expertise to improve project economics. MARA contributes dedicated, energy-efficient data centers. The two sides plan to deliver about 1 gigawatt of IT capacity, and are expected to ultimately reach more than 2.5 gigawatts.
MARA sits at the intersection of energy and computing, and SDV’s development engine provides strong execution and operational capabilities. These capabilities are crucial to MARA’s ability to convert and expand at that intersection into scalable and sustainable digital infrastructure.
The dual-purpose design of these data centers allows them to run artificial intelligence/enterprise/high-performance computing workloads and Bitcoin mining at the same time, enabling operational flexibility amid a constantly changing market environment. This modular approach enables Marathon to continue mining operations while also securing “highly attractive economic terms” from data center customers with higher profit potential.
MARA’s AI buildout can be traced back to 2025.
In August 2025, broker HC Wainwright noted that Bitcoin miner MARA would acquire a 64% stake in Exaion, a high-performance computing (HPC) company owned by French energy giant EDF, and could increase its stake to 75% before 2027. In February this year, an announcement on MARA’s website showed that the transaction for MARA France to acquire the 64% stake in Exaion had been completed; EDF remains a minority shareholder and customer. NJJ invested in MARA France with a 10% stake. Exaion provides HPC data centers and secure cloud/AI, with a board that includes Xavier Niel and MARA CEO Fred Thiel, and it aims to accelerate expansion in Europe.
This marks MARA’s first substantive entry into the AI/HPC arena—shifting from a miner into a participant in compute-as-a-service.
II. Why the transformation?
1. Mining business losses
At the same time it announced the transformation in February, MARA also released its 2025 fourth-quarter results: despite operational improvements, it still posted a huge loss.
In the 2025 fourth quarter, MARA reported a net loss of $1.7 billion (or $4.52 loss per share). This sharply contrasts with the $528 million net profit in the same period last year. Revenue declined 6% year over year to $202 million, below analysts’ expectation of $253.65 million.
MARA’s fourth-quarter performance reflects the tough challenges faced by Bitcoin miners, with multiple unfavorable factors weighing on its profitability. The company’s financial and operational overview shows that its core operating metrics were under overall pressure.
Despite year-over-year compute growth of 25% to 66.4 EH/s and a 20% increase in Bitcoin supply to 53,822 BTC, production fell 19% to 2,011 BTC due to rising network difficulty. MARA succeeded in improving cost efficiency, reducing the cost per daily PET of compute power by 4% to $30.50. However, that was not enough to offset the impact from Bitcoin price volatility and intensified network competition.
Due to major impairment charges and operational pressure, adjusted EBITDA plunged from $796 million in the 2024 fourth quarter to negative $1.5 billion. The company holds about $5.3 billion in cash and Bitcoin, but faces substantial total debt of up to $3.64 billion, and in the past 12 months its leveraged free cash flow consumed $1.77 billion.
2. The rise of AI
MARA’s adjustments also aim to align with the current broader trend of AI’s rise.
Power demand for AI data centers is expected to grow from about 50 gigawatts in 2025 to 200 gigawatts by 2030—an increase of up to 255%—requiring investments of tens of trillions of dollars in capital.
According to a Goldman Sachs research report: by 2030, global data center power demand will be about 165%–200% higher than today, with the share of AI-related loads continuing to rise. McKinsey & Company stated that the cumulative investment demand for AI infrastructure (compute + data centers + power) could reach the scale of tens of trillions of dollars over the next several years.
In the AI wave, MARA will either continue to absorb losses stemming from BTC uncertainty, or pivot toward a compute demand market driven by more essential, durable needs. And since a BTC mining site is effectively a natural AI compute infrastructure, MARA’s transformation looks more like an industrial upgrade by riding the trend.
III. Miners are collectively taking the road to transformation
MARA’s transformation is not isolated; it is a typical snapshot of the broader miner community.
Over the past year, as profitability in BTC mining has gotten smaller and compute demand has exploded due to the rise of AI, miners worldwide are going through a new round of transformation.
According to data released by S&P in February: although revenue from high-performance computing (HPC) and artificial intelligence (AI) to date has been relatively limited, infrastructure investment is accelerating. Analysts predict that from 2026 onward, HPC will bring significant revenue contribution. HPC is no longer a side business: for multiple mining companies, it is expected to become a main pillar of growth in the coming years. In particular, IREN, Terawulf, and Core Scientific, which now are almost entirely focused on HPC development; analysts predict these businesses will drive most of these companies’ revenue growth in 2026.
By 2026, high-performance computing (HPC) revenue will account for 13% of Riot’s total revenue. Other companies’ shift is even more pronounced: IREN’s HPC revenue is expected to jump from 3% in 2024 to 71% of total revenue; Core Scientific is expected to rise from 5% to 71%; HIVE from 7% to 15%; Cipher Mining and Terawulf are expected to reach 34% and 70%, respectively, while their contributions in 2024 were almost negligible.
This shift highlights the industry’s strategic transformation from relying on cryptocurrencies to being driven by artificial intelligence and high-performance computing. Miners are positioning themselves as providers of high-performance computing infrastructure, offering hosting services such as power, cooling, and physical infrastructure.
The following introduces transformation case studies of crypto miners.
1.Core Scientific, Inc.
Core Scientific was founded in 2017 in Seattle, but later moved its operating headquarters to Austin. Its founders include former Microsoft COO B. Kevin Turner. The company initially focused on Bitcoin mining using renewable energy and digital asset infrastructure.
However, due to a sharp fall in the Bitcoin price and high debt levels, Core Scientific applied for bankruptcy protection under Chapter 11 of the U.S. Bankruptcy Code at the end of 2022. During the bankruptcy proceedings, the company’s operations continued. In January 2024, after large-scale restructuring and reorganization, the company exited its bankruptcy predicament.
Since 2024, the company has placed increasing emphasis on AI high-performance computing (HPC). In 2025, it signed a $10 billion data center operations contract. In July 2025, CoreWeave announced plans to acquire Core Scientific for $9 billion.
In March 2026, Core Scientific announced it would sell about $175 million worth of Bitcoin to accelerate the expansion of its AI infrastructure. On the other hand, its Bitcoin mining business will be stopped. In addition to selling Bitcoin, the company also secured a $1 billion loan to build new data centers across multiple states in the U.S.
As of March 2026, Core Scientific currently operates 10 data centers distributed across seven U.S. states.
2.CoreWeave, Inc.
CoreWeave was founded in 2017 in New Jersey by three commodity traders—Michael Intrator, Brian Venturo, and Brannin McBee—and Peter Salanki. The company was originally named Atlantic Crypto, a cryptocurrency mining company that used graphics processing units (GPUs) to mine Ethereum. After the cryptocurrency crash in 2018, the company rebranded in 2019 as CoreWeave, leveraging its massive GPU inventory to begin providing cloud computing infrastructure for enterprises.
As market demand for AI processing kept growing in 2022 and 2023, CoreWeave’s business—backed by exclusive rights to use Nvidia GPUs—saw significant growth. In February 2025, CoreWeave became the first cloud service provider to offer cloud services providing Nvidia GB200 NVL72 chips via the cloud. IBM announced it would use the GB200 cluster to train its Granite AI.
In January 2026, CoreWeave received a $2B investment from NVIDIA, with an acquisition price of $87.20 per share; the two sides expanded their collaboration to support CoreWeave’s data center buildout. In February 2026, CoreWeave sought new financing of $8.5 billion, using a large AI infrastructure contract signed with Meta Platforms as collateral.
3.IREN
IREN’s predecessor was Iris Energy, founded in 2018 by brothers Daniel & Will Roberts. In its early years, the company focused on 100% hydropower/wind power-driven Bitcoin mining businesses, promoting “green mining.” When it went public in 2021, it expanded its compute capacity to 50 EH/s (top 5 globally).
During the crypto winter in 2023, it paused mining expansion and reserved Texas power interconnection rights. It also renamed itself to IREN, downplaying its crypto label.
In October 2025, IREN signed a 5-year, $9.7 billion AI cloud services contract with Microsoft. In March 2026, it signed a $3.5 billion contract with Dell and increased orders by 50k NVIDIA Blackwell B300 units.
4.Terawulf
Terawulf was established in 2019 and focuses on Bitcoin mining and clean energy.
In 2024, it established the WULF Compute subsidiary dedicated exclusively to AI/HPC hosting, fully converting mining sites into liquid-cooled AI data centers.
2025 was its milestone year for a surge in orders: in August it signed a 450MW, 10-year, $6.7 billion contract with Fluidstack, backed by Google. In December, it also reached a collaboration of 72.5MW, 10 years, $1.1 billion with the UAE’s G42/Core42. Total signed business for the year reached 522MW of HPC, total contract value was $12.8 billion, and AI/HPC revenue was $16.9 million, accounting for 10% of total revenue for the year. It also received a $3.2 billion equity and debt investment package from Google and a total funding package of $6.5 billion.
5.HIVE
HIVE’s full name is HIVE Digital Technologies Ltd. It was founded in 2017 by Frank Holmes, Aydin Kilic, and others. Its core team combines experience in cryptocurrencies, energy, and technology. From the very beginning, it set a development direction of “clean energy + crypto mining.”
In 2024, HIVE officially launched an AI compute transformation strategy. HIVE has become North America’s third-largest AI transformation miner (only behind IREN and Terawulf). In Canada’s sovereign AI cloud market, it has a first-mover advantage. The results of the transformation are gradually becoming visible, forming a stable development pattern driven by “mining + AI” on two wheels.
Summary
From the cases above, it’s clear that the crypto mining firms’ transformation wave has already started. Miners are turning themselves into AI training centers, GPU cloud service platforms, and HPC hosting facilities. There are also miners switching from previously holding cryptocurrencies to selling coins to invest in AI. This can also be seen as a re-pricing of compute assets: in the past, compute was consumed by mining, so compute depended on coin prices; now, compute is starting to serve real industrial needs such as AI model training and inference. Such a shift is not only the most direct market reflection of the broader environment of a downturn in the crypto industry, but also a structural optimization that the market receives as the AI era arrives.