Bitfarms, one of the largest publicly listed Bitcoin mining companies in the United States, has announced a radical transformation: it will abandon its original business of Bitcoin mining and pivot entirely to providing High Performance Computing (HPC) services for artificial intelligence companies. The decision comes after the company reported a staggering loss of $46 million in Q3 2025, highlighting the deteriorating profitability of the Bitcoin mining sector following the latest halving.
From Bitcoin Mines to AI Data Centers
CEO Ben Gagnon revealed during the earnings call that the pivot is expected to be completed by Q4 2026, aligning with the planned shipment of NVIDIA's next-generation AI GPU, codenamed Vera Rubin. To fund this transition, Bitfarms has raised over $588 million through a convertible note offering and secured a $300 million debt facility. These funds will be used to retrofit existing mining facilities with high-density power and cooling systems capable of supporting the enormous power demands of Vera Rubin GPUs.
As part of the restructuring, Bitfarms recently sold its last Latin American site in Paraguay for $30 million, effectively transforming itself into a company with a fully North American energy footprint. The company now focuses exclusively on markets with high demand for AI workloads, primarily in Canada and the United States. According to Gagnon, the Washington state facility will serve as the primary deployment site for Vera Rubin GPUs and could "potentially produce more net operating income than we have ever generated with Bitcoin mining, providing the company with a strong cash-flow foundation."
The Vera Rubin Gambit: Betting on Next-Generation Hardware
Unlike many competitors that are adapting their facilities for NVIDIA's current Blackwell generation GPUs—which are already powering most AI workloads including those at OpenAI—Bitfarms has decided to skip this cycle entirely. Instead, it is placing an all-or-nothing bet on Vera Rubin, NVIDIA's next architecture that CEO Jensen Huang confirmed is already in production at CES 2025.
Unconfirmed industry reports suggest that each Vera Rubin GPU will have a Thermal Design Power (TDP) of 2.3 kW, compared to 1.4 kW for second-generation Blackwell chips. This dramatic increase in power consumption poses significant challenges for existing data centers, as current power and cooling infrastructure may not be sufficient. Bitfarms aims to be the first provider ready for these chips, believing that early movers will capture premium pricing.
"A lot of the infrastructure that’s being built right now is not going to be compatible with the next generation," Gagnon explained during the Q3 earnings call. "And as companies allocate all this money into those Vera Rubin GPUs, they’re going to be very economically incentivized to deploy them." He questioned whether AI players would prefer to sit on idle GPUs or pay higher rates to monetize assets immediately, suggesting that demand for ready infrastructure will outweigh cost concerns.
Financial Resources and Timeline
Bitfarms' war chest of over $588 million from convertible notes plus a $300 million debt facility gives it substantial capital to upgrade its facilities. The company plans to use these funds to install advanced cooling systems such as direct liquid cooling, upgrade electrical transformers, and purchase networking equipment capable of handling the high-bandwidth demands of GPU clusters.
The sale of Latin American assets—including sites in Argentina and Paraguay—is part of a broader strategy to streamline operations. By exiting these high-risk jurisdictions, Bitfarms reduces regulatory and political uncertainty while concentrating resources on stable North American markets with abundant hydroelectric and wind power. The company now operates facilities in Quebec, Ontario, and Washington state, regions known for cheap and renewable energy that can support the enormous energy appetite of HPC data centers.
Gagnon stated that the transition timeline targets Q4 2026, the same quarter when Vera Rubin GPUs are expected to begin shipping. This synchronized schedule is intended to minimize idle time: once the infrastructure is ready, Bitfarms can immediately install the latest GPUs and begin offering HPC services to customers who are eager to deploy their pre-ordered chips.
Market Reactions and Industry Perspectives
Following the announcement, Bitfarms' stock experienced increased volatility. Some analysts view the pivot as a necessary move given the structural challenges facing Bitcoin mining—declining block rewards, rising energy costs, and intense competition from large-scale miners. Others caution that betting entirely on a single GPU generation, especially one that hasn't yet shipped, carries enormous execution risk. If Vera Rubin faces delays or if demand fails to materialize as expected, Bitfarms could be left with expensive, underutilized facilities.
Industry peers such as CoreWeave and Hut 8 have also diversified into AI, but most have adopted a gradual approach, retaining some Bitcoin mining capacity while building out HPC services. Bitfarms' decision to completely abandon mining is seen as the most aggressive among major players. Whether this gambit pays off depends on NVIDIA's ability to deliver Vera Rubin on time and market appetite for next-generation AI compute.
Nevertheless, Gagnon remains confident. "Would you rather sit on your GPUs and not deploy them? Or would you rather pay a higher infrastructure expense in order to deploy them and start monetizing the asset?" he asked rhetorically. The coming months will determine whether Bitfarms becomes a pioneer in the next wave of AI infrastructure or a cautionary tale of overambition.