HIVE Digital Technologies just announced one of the largest AI infrastructure projects in Canadian history. Through its subsidiary BUZZ High Performance Computing, the company plans to build a 320 MW AI campus in the Greater Toronto Area, a facility designed to house more than 100,000 GPUs when fully operational.

The estimated price tag: roughly CAD 3.5 billion. The targeted completion window: the second half of 2027. For a company that built its reputation mining Bitcoin and Ethereum, this is about as dramatic a pivot as you’ll find in the digital infrastructure space.

What HIVE is actually building

BUZZ HPC, the subsidiary leading the project, has already acquired approximately 25 acres of land in the Greater Toronto Area for CAD 58 million. That’s the foundation for what HIVE is calling an AI gigafactory, a term borrowed from Tesla’s playbook that signals ambitions well beyond a standard data center.

At 320 megawatts, the campus would rank among the largest purpose-built AI compute facilities in North America. To put that power draw in perspective, 320 MW is enough electricity to power roughly 250,000 homes. Instead, it will power GPUs, the specialized chips that train and run AI models.

The 100,000-plus GPU target is significant. For context, training a frontier AI model like GPT-4 required an estimated 10,000 to 25,000 GPUs. A facility with six figures’ worth of chips could theoretically support multiple large-scale training runs simultaneously, or serve as a massive inference engine for deploying AI applications at scale.

HIVE hasn’t disclosed which GPU vendor will supply the hardware, though the market for AI-grade chips remains dominated by a small handful of manufacturers. The choice of supplier will matter enormously for the facility’s capabilities and its attractiveness to potential tenants.

From crypto mining to AI compute

Here’s the thing about HIVE’s transformation. It’s not happening in a vacuum. The company is part of a broader migration among former crypto miners who are discovering that their core competency, managing massive amounts of power-hungry compute hardware, translates remarkably well to AI workloads.

The logic is straightforward. Crypto mining operations already know how to negotiate cheap power contracts, cool dense racks of processors, and manage facilities at scale. AI data centers need all of those same skills. The difference is that AI tenants, cloud providers, and enterprise customers tend to sign longer-term, higher-margin contracts than the volatile returns of proof-of-work mining.

HIVE has been signaling this strategic shift for some time. The creation of BUZZ HPC as a dedicated subsidiary suggests the company wants a clean organizational separation between its legacy mining operations and its AI ambitions. It also gives BUZZ the flexibility to pursue partnerships, financing, and customer relationships without the baggage of being perceived as purely a crypto company.

The Toronto location is a deliberate choice. Canada’s largest metro area offers proximity to a deep talent pool in AI research, anchored by institutions like the University of Toronto, where much of the foundational work in deep learning originated. The region also provides relatively stable power infrastructure and a regulatory environment that has generally been welcoming to data center development.

The CAD 3.5 billion question

Look, CAD 3.5 billion is a staggering sum for a company of HIVE’s size. The capital expenditure required to build 320 MW of AI-ready infrastructure, fill it with over 100,000 GPUs, and connect it to sufficient power and cooling systems is the kind of investment that typically requires either deep pockets or creative financing arrangements.

HIVE hasn’t detailed how it plans to fund the full build-out. The land acquisition alone consumed CAD 58 million, which is essentially a down payment on a much larger commitment. Investors should expect some combination of equity raises, debt financing, and potentially pre-lease agreements with anchor tenants to make the numbers work.

The competitive landscape is also worth watching closely. HIVE isn’t the only former miner eyeing AI infrastructure. Several publicly traded mining companies have announced similar pivots over the past two years, and they’re all competing for the same pool of power capacity, GPU allocations, and enterprise customers.

The timeline adds another layer of risk. Targeting the second half of 2027 for completion means HIVE is working against a clock in a market where demand for AI compute is intense right now. Every month of delay is a month of lost revenue and a month where competitors can establish themselves. Construction timelines for facilities of this scale have a habit of slipping, particularly when supply chains for specialized electrical equipment and cooling systems are already stretched thin.

For investors evaluating HIVE’s prospects, the key metrics to track will be financing milestones, pre-lease announcements with named tenants, and construction progress updates. A CAD 3.5 billion project that secures committed customers before completion looks very different from one that’s building speculatively. The land is purchased. The vision is articulated. Now comes the hard part: actually building it.