Bitfarms, one of the historic giants in Bitcoin mining that is publicly traded, has decided to change its game. The company announced it will gradually shut down its cryptocurrency farms between 2026 and 2027 to transform into a provider of infrastructure for high-performance computing (HPC) data centers and artificial intelligence. This move comes after a series of multimillion-dollar losses and amid the global boom in demand for GPUs for AI.
From mining blocks to serving GPUs
The announcement was made alongside the third quarter 2025 results, in which Bitfarms reported net losses of $46 million—almost double the $24 million from the same period in 2024. The company currently operates 12 data centers across North America, with an installed energy capacity of 341 megawatts (MW), originally designed to power Bitcoin mining ASICs.
Now, it aims to repurpose this electrical infrastructure—which is becoming increasingly valuable in a computing-hungry world—to host GPU servers focused on generative AI, 3D modeling, scientific simulation, and other high-performance use cases.
In a statement to specialized media, Bitfarms CEO Ben Gagnon explained that the plan involves converting one of its key sites, the Washington data center (United States), to support future NVIDIA GB300 GPUs cooled by state-of-the-art liquid cooling solutions. According to the executive, just that site, operating under a “GPU-as-a-service” model, could generate more operating profit than the company’s entire historical Bitcoin mining activity.
The roadmap communicated to investors is clear: Bitcoin mining will be “phased out” gradually over 2026 and 2027 as new AI computing contracts come online.
An energy portfolio turning into a competitive advantage
Bitfarms’ biggest asset is no longer ASICs or mining farms but its access to relatively cheap, large-scale energy. Besides the Washington facility, the company has recently restructured a $300 million credit line with Macquarie to fund the transformation of its Panther Creek data center in Pennsylvania, with a minimum potential capacity of 350 MW.
When adding all projects in its pipeline, the company speaks of an “energy pipeline” of around 1.3 gigawatts (GW). In a context where tech giants, AI startups, and investment funds are competing for every available megawatt to deploy GPUs, this figure positions Bitfarms favorably to become a significant player in the new AI data center market.
The business logic is obvious:
- Bitcoin mining margins have compressed following the 2024 halving, increased mining difficulty, and BTC prices that don’t always move upward.
- AI, on the other hand, pays very well per megawatt: large language models, training recommendation systems, or cloud gaming platforms require vast amounts of accelerated computing and are willing to sign long-term contracts.
For Bitfarms, this is about shifting from an extremely volatile activity, dependent on Bitcoin’s price, to a more predictable one based on capacity contracts and large-scale GPU leasing.
A transformation amid sector reorganization
Bitfarms is not alone in this move. Over recent months, several publicly traded miners have begun diversifying into AI and HPC. Companies like Core Scientific, Hut 8, and IREN have announced agreements or projects to allocate part of their electricity capacity to AI-focused data centers, leveraging existing energy contracts and already approved sites.
Additionally, Marathon Digital (MARA), a major sector player, has announced it will actively explore the AI compute business as an additional growth avenue.
However, Bitfarms stands out as one of the first significant actors to set a clear date for the end of mining as its primary activity and to explicitly plan a near-total abandonment of Bitcoin mining.
In practice, what’s happening is a conversion from cryptocurrency “farms” to artificial intelligence “farms”. The foundational infrastructure—land, substations, high-voltage lines, cooling systems, physical security—is very similar. What changes is the “hardware”: from specialized SHA-256 ASICs to general-purpose GPU racks.
Washington and Panther Creek: two key pieces of the puzzle
Though the company hasn’t laid out a detailed technical roadmap, it has provided some clues:
- Washington data center
It will be the first major testing ground. The goal is to redesign parts of the technical rooms to host racks of Nvidia GB300 with high-density liquid cooling, capable of dissipating several tens of kilowatts per rack. This type of solution is already becoming standard in large AI hyperscale data centers and specialized startups. - Panther Creek project (Pennsylvania)
With at least 350 MW potential and specific funding through the Macquarie loan conversion, this site could, in the medium term, host one of the largest private GPU facilities in North America if Bitfarms manages to close capacity sale agreements with major AI clients.
The company emphasizes to investors that there’s already a “constant incoming demand” for its sites and that interest in its energy portfolio is high—precisely because few locations have permits, power, and connectivity ready to convert square meters into useful megawatts for AI in a relatively short time.
The stock market, between caution and anticipation
Despite the narrative of opportunity, the market’s initial reaction was lukewarm. After the announcement, Bitfarms’ shares (ticker BITF) closed about 18% lower at $2.60, with over a 50% decline in the past month.
Part of this correction relates to:
- The direct impact of quarterly losses.
- The uncertainty around executing the transformation plan—necessary capex, timelines, potential regulatory delays.
- The risk that the AI cycle won’t maintain the current growth rate over the next decade.
Nevertheless, the movement is happening during a time when markets are starting to penalize “pure” Bitcoin miners and reward those capable of telling a diversification story into AI, energy storage, or digital infrastructure services.
What does this mean for the data center ecosystem?
Bitfarms’ case illustrates a broader phenomenon: the emergence of a new player in the AI data center market. Until now, the key players were:
- Hyperscalers (large public clouds)
- Traditional colocation providers
- Infrastructure funds that buy and develop energy campuses
Now, a fourth profile is emerging: former cryptocurrency miners recycling their energy footprint to sell high-value computing capacity.
If these types of operations succeed, they could:
- Partially alleviate the energy scarcity for new AI projects by reusing already connected sites.
- Increase competition in the “GPU-as-a-service” segment, pushing rental prices downward for startups and mid-sized companies.
- Accelerate the transition of part of the crypto sector towards less speculative business models more tied to the real economy of data.
It remains to be seen whether Bitfarms can execute its plan swiftly enough and secure long-term compute contracts to stabilize its revenue. For now, the clear message to the market is: the golden era of Bitcoin mining as the company’s sole business seems to be over, with the future relying on a strong push into AI data centers.
Source: decrypt