Several important Canadian companies made big announcements this week that highlighted their partnerships with chip giant Nvidia.
The announcements, which ranged from Cohere’s custom models for Nvidia’s architecture to Kepler Communications’ Nvidia-powered satellites, came as part of Nvidia’s GTC conference, which wrapped up in San Jose on Thursday.
Nvidia’s dominance makes its GTC conference a central pillar of the AI industry.
Nvidia sells the shovels for the AI gold rush. It handily holds the majority of market share for graphics processing units (GPUs), the chips essential for AI computing and development, contributing to more than $4-trillion USD in market capitalization. This dominance makes its GTC conference, where the company shows off its upcoming products, a central pillar of the AI industry.
At this year’s conference, Toronto-based large language model (LLM) developer Cohere announced it will build custom models optimized for Nvidia’s latest architecture, Toronto-based Kepler Communications revealed how Nvidia powers its satellite data centres, and other Canadian companies like Telus and Vention spotlighted their Nvidia-powered AI developments.
Below, we’ve highlighted the notable Canadian announcements from the conference and why they matter.
Cohere
Cohere’s custom models for Nvidia’s architecture and software ecosystem will target areas where frontier models (large-scale models like ChatGPT, Claude, or Gemini) “still face critical limitations,” like specialized enterprise workloads or specific hardware, the company said in a statement.
Cohere said the new models meet the “growing demand for AI systems that operate securely, on‑premise, and within national borders for tightly regulated industries.”
The optimized models and Cohere’s flagship North platform will be available on Nvidia’s DGX Spark, the company’s pint-sized “personal AI supercomputer.”
Kepler Communications
In January, Kepler deployed its first wave of orbital data centre satellites, becoming the first company in the world to launch a low-earth orbit satellite system based on an optical relay network.
This week at GTC, Kepler revealed its on-orbit compute capacity is powered by 40 Nvidia Jetson Orin modules, deployed across 10 satellites and interconnected through Kepler’s network. Kepler said each satellite is capable of supporting AI and other compute workloads.
“By leveraging NVIDIA AI infrastructure in our optical network, data can be processed, routed, and acted on in orbit rather than waiting to return to Earth,” Kepler CEO Mina Mitry said in a statement.
Telus
The Canadian telecommunications giant used last year’s GTC to announce that Nvidia would upgrade the AI compute power of its Rimouski, Que. data centre and turn it into a “Sovereign AI Factory.”
This year brought an iterative update: Telus has partnered with Silicon Valley-based cybersecurity company Fortanix on a new “Confidential AI solution” built on Nvidia hardware. The result aims to ensure proprietary AI models, sensitive data, and agent credentials remain encrypted throughout training, fine-tuning, and inference.
“This latest innovation with Fortanix and NVIDIA adds a critical, final security layer to protect data while it is actively being processed,” Telus chief information officer Hesham Fahmy said in a statement.
Vention
Montréal-based industrial tech scaleup Vention announced the commercial launch of Rapid Operator AI, a robotic arm that automates picking up items out of bins for manufacturing settings.
The robot is powered by Vention’s Generalized Robotic Industrial Intelligence Pipeline (GRIIP), which mixes Vention’s proprietary AI models with Nvidia’s Isaac robotic models (specifically FoundationStereo and FoundationPose) to match robotic perception to the associated motion.
Vention said its robotic arm achieves up to 99 percent first-pick success rates, depending on part geometry and configuration of the bin it’s picking from, and automatically retries when it fails.
Vention secured $110-million USD ($150-million CAD) in Series D funding this past January to scale its physical AI platform into manufacturing plants worldwide.
Other announcements
Ottawa-based supply chain management firm Kinaxis and Toronto-based advanced materials startup OTI Lumionics both said they saw large speed gains in their products using Nvidia hardware during GTC.
Kinaxis said that a test of its decision-making Maestro platform on Nvidia’s AI architecture reduced total supply chain planning calculation time by a factor of 12. The test model covered more than 40,000 SKUs across a six-quarter daily planning horizon, and shortened planning from more than three hours to approximately 17 minutes, according to Kinaxis.
OTI, which develops materials for displays in consumer electronics using quantum simulations, said it successfully ran its material design simulation algorithm on a single Nvidia Blackwell GPU.
The company said high-precision simulations of complex molecular systems often require extensive use of supercomputing clusters, but that it increased performance by a factor of 90 by migrating its workloads to this Nvidia GPU. OTI claimed that this will enable more efficiency, speed, and accuracy in its simulations than ever before.
Feature image courtesy Nvidia via LinkedIn.