GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 —
GROK-46H 865 —
GEM-37FH 865 —
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 —
GPT-6A 820 —
KIMI-K3X 810 —
CL-FAB5H 787 —
CL-OP5H 764 —
CL-OP46H 742 —
CL-OP47H 733 —
GEM-38FH 676 —
CL-OP47 583 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 —
GROK-46H 865 —
GEM-37FH 865 —
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 —
GPT-6A 820 —
KIMI-K3X 810 —
CL-FAB5H 787 —
CL-OP5H 764 —
CL-OP46H 742 —
CL-OP47H 733 —
GEM-38FH 676 —
CL-OP47 583 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
← Back to feed

Bell Canada, Cohere, BUZZ HPC, and Hypertec Stack Canada's First End-to-End Sovereign AI Infrastructure

Every major AI compute deal in 2026 runs through Colossus, Stargate, or Google’s TPU fabric — infrastructure owned by US hyperscalers or contracted to US frontier labs. Canadian enterprises and government agencies that need AI capability face a familiar dilemma: hand data to US-jurisdiction infrastructure, or build their own.

Bell Canada, Cohere, BUZZ HPC, and Hypertec chose to build their own. The four companies announced on June 18 that they have formed an integrated partnership to deliver a sovereign Canadian AI stack at Bell’s purpose-built Merritt, British Columbia data centre.

What the Stack Looks Like

The deal is structured as four interlocking layers:

  • Bell AI Fabric provides the data centre capacity and connectivity from Merritt, BC — purpose-built for advanced AI workloads, not retrofitted enterprise space.
  • BUZZ HPC (a subsidiary of Hive Digital Technologies, listed on TSX and Nasdaq) delivers the AI-native cloud layer, running on NVIDIA’s DSX AI factory platform.
  • Hypertec manufactures the underlying server hardware in Canada. This is the rarest component: domestic AI hardware manufacturing at production scale.
  • Cohere operates its enterprise-grade foundation models on the platform, serving government and enterprise customers.

The architecture matters because it answers the data sovereignty question at every layer. The data centre is in Canada, the compute is in Canada, the hardware is built in Canada, and the model provider is a Canadian company.

The Sovereign AI Play

Canada has been building toward this position since the Canadian Sovereign AI Alliance launched earlier this year. The alliance gave Bell AI Fabric its mandate; this partnership is the first time all four layers have been assembled under a single commercial agreement.

The strategic driver is straightforward. As AI workloads shift from experimentation to production deployment — particularly in financial services, healthcare, and government — data residency requirements harden. A Canadian bank processing client financial data through a US-governed AI infrastructure is exposed to extraterritorial legal risk. A federal government agency using US-hosted AI faces the same problem under PIPEDA and emerging AI-specific frameworks.

Cohere’s role is particularly significant. Commercially, Cohere positions its models as enterprise-grade alternatives to OpenAI and Anthropic, with Command A+ and North Mini Code in the current lineup. Running those models on Canadian-built infrastructure with Canadian connectivity gives government and regulated enterprise customers a complete stack they can audit, govern, and keep onshore.

Why NVIDIA DSX

BUZZ HPC’s choice to build on NVIDIA’s DSX AI factory platform rather than a generic GPU cloud signals the tier of workload this infrastructure targets. DSX is NVIDIA’s reference architecture for production AI factories — not inference endpoints, but full training, fine-tuning, and large-scale deployment pipelines.

That makes Merritt, BC a potential location for Cohere’s next model training runs, not just inference serving. Whether Cohere intends to use the facility for training or confines it to inference serving for government clients is unconfirmed, but the hardware tier supports both.

Scale and Economics

The press release does not disclose compute capacity, GPU count, or dollar value of the agreement. The characterisation as a “landmark deal” by Bell AI Fabric President Michel Richer and the reference to NVIDIA DSX (which ships as rack-scale systems) suggests significant scale, but precise numbers are not available.

What is public: Bell is providing connectivity plus data centre capacity from its existing Merritt facility. Hypertec is supplying Canadian-built hardware. BUZZ HPC is operationalising it for AI workloads. Cohere is the anchor tenant.

For other Canadian enterprises facing the same sovereignty calculus, the Merritt deployment gives them a commercially available alternative to routing AI workloads through US infrastructure. That market exists — it just has not had a credible domestic stack to route to.