Frank Holmes, Executive Chairman of HIVE Digital Technologies, addressed an audience of operators who have navigated multiple halving cycles and regulatory shifts. The central argument was not that AI is a better business than mining, but that the specific operational discipline required to manage Bitcoin mining at scale is the necessary precursor to succeeding in artificial intelligence infrastructure. It is a narrative of adaptation, where a company born in the crypto era leverages its existing power contracts and facility footprints to enter a far more capital-intensive market. Joining Holmes was Luke Rossy, who provided context on the technical realities of this expansion.
This discussion, recorded at Mining Disrupt Dallas 2025 in Irving, Texas, on November 12–14, 2025, captures Hive mid-transition. Bitcoin was trading at $104,994 that week. As of October 10, 2026, it stands at $82,795, well below that level, yet the strategic logic Holmes outlined remains intact. The argument has aged well because the fundamental challenge of AI infrastructure has not changed: it is expensive, complex, and requires a level of operational maturity that most new entrants lack. Hive’s approach suggests that the companies that succeed will not be those with the most capital, but those with the most refined operational muscle.
The 33x Reality: Why AI Infrastructure Is Not Just Bigger Mining
Holmes opened by acknowledging the popularity of the AI narrative. He noted that Hive was the first crypto mining company to go public in Canada before Riot, and the first to balance the grid in Sweden. However, he warned against the common misconception that moving into high-performance computing is simply a matter of buying larger silicon. The operational reality, he argued, is exponentially more complex. In Bitcoin mining, the feedback loop is immediate and tangible. In AI, the delay between capital expenditure and revenue generation is significant, requiring a different kind of financial patience and engineering precision.
when we get an [Antminer], a brand new one, it takes us 6 hours to unpack it and plug it in and get cash flow
Frank Holmes · 0:39
This immediate liquidity is the bedrock of the mining business model. It allows for rapid scaling and quick response to market changes. The AI business, by contrast, involves a much longer lead time. Holmes highlighted the specific delay associated with Nvidia hardware, which is central to the current AI boom. The transition from a simple plug-and-play model to a complex deployment scenario is not just a logistical hurdle; it is a cultural shift for the engineering teams.
When we get an Nvidia H100, it is 6 weeks before you can even think of you're going to get cash flow from it
Frank Holmes · 1:24
Holmes described the process as being 33 times greater in complexity than Bitcoin mining. He used a vivid comparison to explain the mindset shift required for his team. The difference is not merely in the cost of the hardware, but in the delicacy of the engineering. The companies making big splashes in the AI space often underestimate this gap. Hive, with its three years of experience in the sector, has learned that the transition is far more challenging than the initial announcements suggest. Despite the hurdles, the company has managed to scale its AI division to $20 million in revenue with 80% margins, a clear indicator of the operational efficiency they brought from their mining roots. Rossy noted that this efficiency is driven by a rigorous internal audit process that treats every megawatt as a critical asset.
The Paraguay Gamble: Turning Hydro Surplus into Fiscal Stability
A significant part of Hive’s strategy relies on securing cheap, abundant power. Holmes detailed the company’s move into Paraguay, a decision driven by global shifts in energy policy following the invasion of Ukraine. The war caused hydroelectricity prices to rise globally, making it difficult for green miners to maintain growth momentum in traditional locations. Paraguay, however, presented a unique opportunity. The country is endowed with the largest dam in the Western Hemisphere, a massive infrastructure project that produces over 10 gigawatts of electricity.
It is 7 kilometers or 5 miles long and it produces over 10 [gigawatts] of electricity
Frank Holmes · 3:32
Paraguay's politics and economy are complex. Holmes noted that the country’s existence is tied to a historical conflict in 1870, and its current stability is partly due to the influence of the United States. The president, a former rugby player who studied in America and worked at the IMF, has focused on scaling the country’s economic potential. For Hive, the appeal is not just the power, but the fiscal stability. Paraguay has a low tax rate, with income taxes at only 10%, which is significantly lower than neighboring countries like Colombia. This fiscal environment allows Hive to hire local engineers and integrate university programs, creating a sustainable ecosystem for their data centers.
From Jeep to Ferrari: Engineering the Mindset Shift
Holmes’s analogy of the Jeep versus the Ferrari captures the essence of the engineering challenge. Bitcoin mining hardware is rugged, durable, and designed to run continuously in harsh conditions. It is a workhorse. AI hardware, particularly high-end Nvidia chips, is delicate, expensive, and requires precise environmental controls. The engineering teams must shift from a mindset of brute force to one of precision and care. This shift is not just about technical skills but about corporate culture.
The financial implications of this shift are significant. Holmes noted that while ASIC chips for Bitcoin mining generate 12 to 14 cents per hour, Nvidia chips can generate more than a dollar per hour. Newer models like the H200 are expected to reach $2 per hour, and Blackwell chips could push up to $5 per hour. This revenue differential justifies the higher capital expenditure and the operational complexity. For Hive, the AI business is not a side project but a core growth driver, with plans to scale revenue to $100 million in the next 12 months. The company has already secured 10,000 customers worldwide who rent their GPU chips, demonstrating a strong market demand for their services.
The Canadian Pivot: Repurposing Bitcoin Facilities for HPC
In Canada, Hive is leveraging its existing infrastructure to enter the HPC market. The company has purchased a Bitcoin mining facility in Toronto with less than 10 megawatts of capacity. Instead of starting from scratch, Hive is repurposing this facility for HPC, a strategy that significantly reduces costs and accelerates deployment. The permitting process for converting an existing facility is much faster than building a new one. Holmes emphasized that while new construction can take three years, repurposing a Bitcoin facility takes only nine months. That speed to market matters in AI.
the permitting is not 3 years for something that's brand new
Frank Holmes · 7:46
Hive is also partnering with Bell Canada to expand its HPC capabilities in Canada. Bell Canada, which previously viewed data centers as a cost center, is now recognizing them as a profit center. This partnership allows Hive to tap into Bell’s distribution network and customer base. The company is planning to turn on machines in Manitoba, followed by Toronto and the east coast. This phased approach allows Hive to manage its capital expenditure while building a strong presence in the Canadian market. Bell Canada itself works with Cohere, the Canadian large language model developer, and brings a 600-person sales force.
San Antonio and the New Regulatory Center of Gravity
Holmes concluded by discussing Hive’s strategic move to San Antonio, Texas. The company has relocated its head office to the city, citing the favorable regulatory environment and the deep talent pool. San Antonio is a major hub for hyperscaling, largely due to the University of Texas at San Antonio, which has 10,000 cybersecurity engineers. This talent pipeline is crucial for the AI industry, which requires a skilled workforce to manage and secure its infrastructure. The city is also attracting significant investment from other tech giants, including Microsoft, which is building a $3 billion data center. This concentration of resources makes San Antonio an ideal location for Hive to expand its operations.
The move to San Antonio reflects a broader trend in the industry, where companies are seeking jurisdictions that offer both energy stability and regulatory clarity. For Hive, the relocation is a strategic decision that aligns with its long-term goals. By positioning itself in a key AI hub, Hive can better access the talent, capital, and partnerships needed to drive its growth. The company’s journey from Bitcoin mining to AI infrastructure shows how a miner can reuse its power, sites and operating habits for AI. Other miners weighing the same move can look at how Hive sequenced the transition.
This conversation is part of a broader dialogue on the future of AI infrastructure, which continues at Intelligence Disrupt 2027, taking place March 22 to 24 in Irving, Texas, where operators and builders meet again.