HIVE Digital Technologies is betting big on Canada’s AI future. Its subsidiary, BUZZ High Performance Computing, has announced plans to build a 320 MW AI “gigafactory” in the Greater Toronto Area, a facility that would rank among North America’s largest domestically controlled AI clusters.
The project carries a price tag of roughly CAD $3.5 billion and is targeting the second half of 2027 to bring its first operations online.
What the gigafactory actually looks like
The facility is designed to host more than 100,000 GPUs across a vertically integrated supercomputing platform. BUZZ HPC has reportedly acquired a 25-acre land parcel for approximately CAD $46 million to house the campus. The site sits within the Toronto-Waterloo innovation corridor.
The facility will draw roughly 320 MW of utility power. Ontario’s grid, which runs heavily on nuclear and hydroelectric generation, gives the project a low-carbon energy story. The campus will also feature a sustainable, low-water-use cooling system.
On the employment side, the project is expected to create over 800 construction jobs during the build phase and hundreds of permanent high-skill positions once operational.
From Bitcoin mining to AI compute
HIVE Digital Technologies built its reputation in Bitcoin mining, operating GPU and ASIC rigs across facilities in Canada, Sweden, and Iceland. The pivot toward high-performance computing and AI infrastructure represents a strategic evolution. A CAD $3.5 billion capital commitment reflects the scale of that shift.
The “sovereign AI” branding addresses Canada’s reliance on US data centers controlled by American hyperscalers for AI training and deployment workloads. BUZZ HPC is positioning itself as domestically controlled infrastructure for government agencies, defense contractors, and regulated industries that need data to stay on Canadian soil.
What this means for investors
The financing structure for this project will be a critical detail to watch. Whether HIVE funds this through equity dilution, debt, government subsidies, strategic partnerships, or some combination will determine how much value actually flows to existing shareholders versus new capital providers.
The 2027 timeline introduces execution risk. GPU architectures evolve rapidly, and a commitment to 100,000 units locks in certain hardware assumptions that may need revisiting before the first rack powers on.