Canadian AI: Cohere eyes a German merger, and a Toronto 911 AI startup goes stealth-to-exit in under a year.
And more on NVIDIA's first Canadian OEM partner, Shopify's AI Toolkit, four venture rounds and a bit on "AI psychosis."
Good morning! Welcome to the Canadian AI Newsletter, a weekly rundown for founders, operators and investors.
Was this forwarded to you? Subscribe below.
I am Raif Barbaros, Partner at Mistral Venture Partners. Views are my own.
This was the week Canadian AI went international. Cohere surfaced in advanced merger talks with Germany’s Aleph Alpha. Motorola acquired Toronto’s Hyper, a 911 AI startup that went from stealth to exit in under a year. Montréal’s Hypertec became NVIDIA’s first Canadian OEM partner. The Bank of Canada convened the Big Six to discuss whether Anthropic’s latest model is a cybersecurity threat (I have thoughts). Scotiabank became the fourth Big Five bank to launch an enterprise AI platform. And Shopify shipped an open-source AI Toolkit that hands full store control to AI agents. Let’s get into it.
💰 Deals & Milestones
Cohere (Toronto) is in advanced merger talks with Germany’s Aleph Alpha, per Germany’s Handelsblatt, confirmed by BNN Bloomberg.
Combined entity headquartered in both countries. The German government supports the deal and would become an anchor customer. Neither company has officially confirmed. Ties directly into the Canada-Germany Sovereign Technology Alliance signed in February.
Cohere is roughly 3x the size (~1,000 employees vs ~350 per LinkedIn). And while Aleph Alpha was once hailed as Europe’s answer to OpenAI, it abandoned its own model development in late 2024, pivoting to an “AI operating system” after CEO Jonas Andrulis acknowledged that just having a European LLM wasn’t a viable business model. I can see how their pivot can complement Cohere.
(Disclosure: Cohere is a portfolio company.)
Hyper (Toronto) acquired by Motorola Solutions. Stealth to exit in under a year.
Agentic AI voice technology to screen non-emergency 911 calls. Emerged from stealth less than a year ago with a US$6.3M seed led by Toronto’s Ripple Ventures. Handles 100+ non-emergency scenarios in 30+ languages.
Clients included Toronto Police Service, Manitoba police, and San Diego County Sheriff’s Office. All 18 employees join Motorola. Co-founders Ben Sanders (ex-Clearco) and Damian McCabe (ex-Connected/ThoughtWorks) are Canadian repeat entrepreneurs. Company originated in the Yukon.
Stealth in 2025, seed round, paying clients across two countries, acquisition by a $70B public company. In under a year. That’s a tempo worth paying attention to.
Hypertec (Montréal) became NVIDIA’s first Canadian OEM partner for AI servers, through its Ciara division.
Pre-release access to NVIDIA products, direct engineering collaboration, and marketing/sales support. Hypertec will manufacture NVIDIA-Certified AI systems domestically. Bell Canada has already selected Hypertec for five new BC data centres.
CEO Simon Ahdoot: “Canada can take a bigger slice of sovereign revenue.” Canadian-made NVIDIA AI servers, manufactured in Canada, for Canadian data centres. The physical layer just got more Canadian.
Soma Energy (Vancouver) emerged from stealth with $7M USD for AI-powered data centre energy optimization.
Seed led by Category Ventures with Haystack, RRE Ventures, TO VC, Uncork Capital; Panache Ventures at pre-seed. Founded by ex-AWS energy leaders. Already optimizing 2 GW of electricity across 5 data centre customers. 18 employees.
As data centres keep getting bigger, someone has to optimize the power. Soma is betting it can be the intelligence layer for all of it.
Sonibel (Vancouver) raised $1.6M USD pre-seed for AI welding defect detection.
UBC graduates built a machine learning model that identifies weld defects in real-time via acoustic sensing. A proprietary sensor mounts on the welding torch and listens for defect signatures.
Led by Maple VC with Champion Hill Ventures and Dorm Room Fund. First paid pilot is a Fortune 100 company. AI that listens to a weld and tells you if it’s good. Canadian deep-tech at its most practical.
Harmix (Toronto) raised $1M USD, pivoting from music AI to AI agents for SMBs.
Originally a multimodal AI company with patented music/image/video search technology (clients: Red Bull, Disney, Warner Brothers). Now building “PAM,” a proactive AI manager for SMBs using agents to solve software fragmentation across Google Drive, Slack, and other tools.
CEO Nazar Ponochevnyi is a Vector Institute graduate researcher. ~20 employees. The original music search business remains profitable.
🏢 Large Company / Big Tech
Shopify (Ottawa/Toronto) launched an open-source AI Toolkit that hands full store control to external AI agents. Also on GitHub.
Free, MIT-licensed. Connects Claude Code, OpenAI Codex, Cursor, Gemini CLI, and VS Code directly to the Shopify platform. First time a major e-commerce platform has handed full operational control of a live store to external AI coding agents.
Ships with 16 skill files covering admin, products, themes, Hydrogen storefronts, Functions, and partner tools. Live documentation access, real-time code validation, and store operation execution through CLI.
Scotiabank (Toronto) launched Scotia Intelligence, an enterprise AI platform integrating capabilities, governance, and infrastructure.
Key component: Scotia Navigator for assistive AI. AI already handles 40%+ of client queries in contact centres. In commercial banking, AI processes ~90% of commercial emails for routing. First Canadian bank with a dedicated Data Ethics team.
Fourth Big Five bank to make a major enterprise AI announcement: RBC (Issue 2), TD (Issue 3), BMO (Issue 7), now Scotiabank. CIBC, you are now officially the last one standing.
Microsoft Canada detailed the implementation phase of its $19B CAD cloud and AI infrastructure commitment with a “Community-First approach” for Ontario data centre expansion.
Five principles: no electricity price increases for Canadians, full cost coverage for grid upgrades, water stewardship, energy-efficient design, and workforce development. Invest Ontario welcomed the investment, supporting 1,250 jobs.
🔬 Research
Mila and Sweden’s RISE Research Institutes signed a landmark MOU for AI collaboration in forestry, mining, and energy.
Joint Steering Committee established. Mila CEO Valérie Pisano: “We are focusing AI development on critically strategic domains like our forests, mines, and energy systems.” Swedish royalty visited Mila for the signing.
Between the Mozilla partnership (Issue 7) and now RISE, Mila is quietly building a web of international AI research alliances anchored in open-source and sovereignty. Worth watching.
UWaterloo received $250K from Princess Margaret Cancer Centre for “MedDataOS,” a multi-agent AI framework for biomedical data analysis on head and neck cancers. Professor Ana Crisan (Cheriton School of Computer Science) is leading.
McGill Office for Science and Society published a detailed investigation into “AI psychosis,” the phenomenon where sycophantic AI chatbots reinforce delusional thinking. Tests conducted April 7-9 on Gemini and Claude.
🏛 Policy
Bank of Canada convened the Big Six banks and regulators to discuss cybersecurity risks from Anthropic’s Claude Mythos Preview model, which can identify and exploit software vulnerabilities.
The CFRG brought together Canada’s Big 6, Desjardins, OSFI, Department of Finance, and TMX Group. Meeting followed a similar US session called by Treasury Secretary Bessent. OSFI said it won’t issue short-term guideline changes but is in “active conversations.”
My take: I’m skeptical about the severity narrative. Anthropic has a pattern of being dramatic about its own models’ capabilities to drive attention and demand. If Mythos is truly as dangerous as claimed, how does limiting access to 12 companies meaningfully contain the risk? And is anyone suggesting no other lab has similar capabilities? Mythos sounds like another impressive step toward AGI-level reasoning. The cyber risk is real. The apocalyptic tone probably isn’t.
Liberal Convention voted to restrict AI chatbots for minors under 16 at the party’s national convention in Montreal (4,500 delegates).
Two non-binding resolutions: (1) age of 16 for social media; (2) a Quebec-originated resolution to limit “all AI chatbots and other potentially harmful forms of AI interaction” to Canadians over 16. McGill’s Taylor Owen warned the upcoming Online Harms Act must include AI chatbots, citing the Tumbler Ridge case. Not binding on cabinet, but signals where the party base is heading.
AI in Canadian law enforcement made headlines twice this week.
RCMP AI-drafted police reports: The RCMP is piloting Axon’s Draft One AI software across 10 detachments (8 in BC, 2 in Alberta) to auto-generate draft incident reports from body camera audio. $200,000 budget, ~800 reports generated, ~380 officers participated. Officers must change a minimum 10% of any AI draft. Professor Christopher Schneider (Brandon University) raised concerns about AI “hallucinations” in court evidence.
Edmonton Police AI facial recognition: CBC obtained documents via ATIP on Edmonton Police Service’s AI facial recognition bodycam pilot (~50 officers, December 2025). Facial recognition model supplied by Corsight AI. A “critical fault” system outage during the pilot. Privacy assessment submitted only the day before cameras went live. Watchlist of ~7,000 people. Kate Robertson (U of T Citizen Lab) called it “likely the most high-risk algorithmic surveillance program I have observed to date in Canada.”
In brief
Google launched AI Mode in Canada, one of the first countries outside the US to receive it. Uses a custom Gemini 2.5 model. Agentic restaurant booking also expanded to Canada.
Environment Canada announced a hybrid AI weather model. Six-day forecasts as accurate as current five-day. 8 to 24+ hours earlier detection of major weather systems. Practical, tangible, affects every Canadian. More of this, please.
Alberta’s AI data centre wars continue. Synapse reapplied to the AUC for a 1.4-gigawatt facility in Olds designed for AI hyperscalers. Meanwhile, Wonder Valley, the $70B data centre park in Grande Prairie backed by Kevin O’Leary Ventures, was deemed exempt from environmental impact assessment.
Shared Services Canada updated its federal AI strategy, detailing CANChat, a GenAI chatbot for public servants powered by GC LLM, a Canadian-trained large language model.
Budget 2025 SR&ED changes could supercharge Canadian hard tech and AI hardware startups, allowing companies with physical footprints to reclaim R&D investments more effectively.
Sorintellis, UdeM, Université Laval, and IVADO announced a collaboration for AI-powered clinical trial intelligence. Mila researchers building recommendation engines spanning reinforcement learning, causal inference, and medical ML. Montréal’s pharma-AI pipeline keeps growing.
Edmonton’s Sarcomere Dynamics and NTWIST are among the companies tapped for NGen’s $79.5M AI-manufacturing investment (Issue 8). Sarcomere is deepening its work with Magna on AI robotics for automotive production. Good to see Edmonton getting its share of the NGen pie.
📅 Upcoming Events
ALL IN Vancouver — April 15, 2026
SASK AI EXPO (Saskatoon) — April 27, 2026
The AIA (Montréal) — May 5, 2026
Upper Bound (Edmonton) — May 19–22, 2026 (Sold out. 8,000+ attendees.)
Canadian AI Conference 2026 (Vancouver) — May 25–29, 2026
ALL IN Toronto — May 28, 2026
ALL IN Main Event (Montréal) — September 16–17, 2026
Thank you for reading! Subscribe, and please share any feedback 🙏🏼
— Raif Barbaros