Canada has increasingly recognized the need for domestic AI compute capabilities, shifting away from reliance on foreign facilities. BUZZ HPC, a subsidiary of HIVE Digital Technologies, is set to spearhead this transformation with plans for a massive AI infrastructure project in the Greater Toronto Area. With an investment of approximately CAD $3.5 billion, this initiative aims to establish a facility that can accommodate over 100,000 GPUs at capacity, marking a significant step toward developing a homegrown AI ecosystem.
What does the construction entail?
The new facility is being referred to as an AI gigafactory, boasting an impressive utility capacity of 320 MW. This capacity dwarfs that of typical large data centers, which generally operate in the range of 30 to 50 MW, making BUZZ HPC’s project six to ten times larger.
A core feature of the facility will be its liquid-cooled GPU infrastructure. Leveraging NVIDIA GPUs, the facility's software management will utilize Kubernetes and Slurm, two prominent platforms equipped to oversee large-scale compute clusters effectively.
Collaboration is key for this project, with partnerships formed with major industry players. Bell Canada will provide telecommunications and fiber expertise, while Dell Technologies will contribute its hardware capabilities. VAST Data, specialized in high-performance storage solutions, will also play a vital role.
How does this fit into HIVE’s larger strategy?
HIVE Digital Technologies has a strong presence in the high-performance computing sector since 2017 and continues to grow its influence as an NVIDIA Cloud Partner. This new GTA gigafactory represents an expansion of existing operations that have previously included infrastructure across various provinces, including New Brunswick, Manitoba, and British Columbia. The overall strategy aims to scale liquid-cooled GPU capabilities nationwide, with the GTA location offering favorable fiber connectivity, cost-effective power, and proximity to leading AI research centers such as the Vector Institute and the University of Toronto.
In an environment where Canadian AI infrastructure has frequently been hosted abroad, the new facility presents a promising domestic alternative for government, enterprise, and research clients. With this initiative, Canadian operators can avoid the constraints of foreign regulations and pricing, enhancing their operational efficiencies.
What implications does this have for investors?
With a capital investment of CAD $3.5 billion, HIVE is making a significant commitment to establishing itself as a vital domestic infrastructure provider rather than just a competitor among hyperscalers. The existing infrastructure already supports about 130,000 GPUs across the country, laying a baseline for revenue. However, the scale of this project requires a scrutiny of how the company manages its financial resources moving forward. Observers should monitor how HIVE addresses the financial demands of this major undertaking, including funding strategies such as debt, equity, or partnerships.
Successfully constructing and filling a facility of this magnitude requires meticulous coordination in several areas: securing power procurement, adhering to construction timelines, managing hardware supply chains, and attracting clientele. While the GPU supply has improved since the shortages of 2023, the competitive nature of the market remains a challenge. Delays in data center construction projects are also common, adding another layer of complexity to this ambitious endeavor.