Mila and Inovia Launch $100 Million Fund to Turn Canadian AI Research Into Companies

Mila, the Montreal-based artificial intelligence research institute founded by Yoshua Bengio, has partnered with venture capital firm Inovia Capital to launch a $100 million USD early-stage fund aimed squarely at the gap between Canada's world-class AI research output and the country's much smaller share of AI venture capital investment. The Venture Scientist Fund, announced this month, is designed to invest in roughly 55 AI-native companies over the next several years, all of them emerging from the research pipelines at Canada's federally funded AI institutes. The announcement is the clearest policy signal yet that the federal government and the Quebec AI ecosystem have accepted a blunt diagnosis of Canada's AI economy: strong researchers, weak commercialisation.

The gap the fund is trying to close

Canada represents roughly 10 per cent of the world's top AI research talent, a share that has held steady through the last decade of global AI investment boom. Canada captures less than 2 per cent of global AI venture capital. The five-to-one gap between research excellence and capital formation is the reason a generation of AI breakthroughs pioneered at the University of Toronto, the University of Montreal, and the University of Alberta have translated into American companies rather than Canadian ones. Google's deep learning division in Toronto and New York both trace their lineage to Geoffrey Hinton's lab. OpenAI, Anthropic, and DeepMind have all hired heavily from Canadian PhD programmes.

The Venture Scientist Fund is, in the plainest terms, an attempt to keep the next cohort of that research onshore and commercialised in Canada. The fund will invest at the earliest stages of company formation, often before there is a company at all, engaging with researchers while their work is still at the PhD or postdoctoral stage. The pitch to researchers is that Mila and Inovia will provide capital, operational support, recruiting, and a bridge to later-stage funding rounds without requiring them to choose between academic careers and entrepreneurship.

Why this partnership, now

Inovia Capital is one of the larger Canadian venture firms with a track record of software and AI investments, including stakes in companies like Lightspeed and AlayaCare. The firm's decision to anchor the Venture Scientist Fund reflects a strategic bet that the pipeline from Mila and sister institutes is deep enough to support systematic early-stage investment rather than opportunistic deals. Inovia partners have been unusually frank in public remarks that Canada's AI sector has been under-invested in by domestic VCs and that the country's capital markets were not equipped to absorb the dealflow being generated by the research institutes.

Mila, for its part, has been exploring the commercialisation question for years. The institute has spun out individual companies before, most notably in drug discovery, climate modelling and language technology, but without a dedicated fund of its own the spinouts have generally had to look outside the institute for seed capital. The Venture Scientist Fund formalises an in-house path from research to product and is expected to dramatically accelerate the pace of spinout activity.

The national AI institutes network

The fund is positioned as a national resource rather than a Montreal-only vehicle. Mila is one of three federally funded national AI institutes created under the Pan-Canadian AI Strategy launched in 2017. The other two are Toronto's Vector Institute, whose affiliated faculty includes Hinton and most of the senior deep learning researchers at the University of Toronto, and Edmonton's Alberta Machine Intelligence Institute, or Amii, which grew out of Richard Sutton's reinforcement learning group at the University of Alberta.

All three institutes have shared a similar commercialisation problem: generating research at the frontier of AI while watching most of the resulting companies incorporate in the United States or get acquired early by American firms. The Venture Scientist Fund's announcement includes explicit language about investing across all three institutes, a design choice that will be tested in practice by where the dealflow actually concentrates. Montreal has a head start because Mila anchors the fund. Toronto and Edmonton will need to organise their pipelines to match.

Quebec's parallel investment

The provincial context matters. The Quebec government committed an additional $36 million CAD in core operating funding to Mila in March, a renewal and expansion of the province's long-standing support for the institute. Mila, Hypertec, and infrastructure developer 5C separately announced a $250 million LaSalle campus and sovereign AI research hub, a facility that will co-locate the institute with compute infrastructure and commercial AI tenants. The combination of public operating support, private infrastructure investment and a new commercialisation fund represents the most concentrated AI policy push any Canadian province has made.

The federal position is more diffuse. Ottawa continues to fund the national AI institutes through the Pan-Canadian AI Strategy, and Innovation, Science and Economic Development Canada has been consulted throughout the Venture Scientist Fund planning. The federal government has not announced matching capital in the fund structure, a choice that reflects both the Liberal government's broader capital programme constraints and the principle that commercialisation investment should be led by the private sector.

What sectors the fund is targeting

The public announcement identifies four priority areas for investment: AI infrastructure and foundational models, AI applications in life sciences and drug discovery, AI in climate and energy, and AI in enterprise automation. Each reflects a cluster of existing research strength at one or more of the national institutes. Drug discovery ties into Mila's existing partnerships with Valence Labs, the Montreal-based computational drug discovery company. Climate work connects to the federal Net Zero Accelerator and provincial decarbonisation priorities. Enterprise automation is the most crowded global market but also the one where Canadian enterprise software buyers have shown early interest in domestically developed tools.

The fund has signalled that it will not try to compete on agentic AI or generalist consumer applications, two areas where the capital scale required is beyond what a $100 million Canadian fund can support. The implicit bet is that the next generation of commercially viable AI companies will be domain-specific, grounded in scientific research, and built around a defensible technical advantage rather than a general-purpose model race.

The talent retention question

The most-watched question among Canadian AI policy analysts is whether the fund will slow the talent drain to American companies. Salaries offered by frontier US labs to top AI researchers routinely exceed $500,000 USD and can reach into the millions for senior hires. No Canadian startup fund can match those salaries at the early stage, and the Venture Scientist Fund has not pretended otherwise. What it offers instead is the chance for researchers to commercialise their own work with meaningful equity stakes, operational support and the ability to stay based in Canada.

Whether that proposition wins against American cash offers will be determined case by case. For researchers who have built their careers around specific Canadian research groups, have immigration and family reasons to stay, or want to build companies rather than work inside the labs of others, the Venture Scientist Fund is a significantly better offer than anything that existed in the Canadian market a year ago. For researchers whose primary orientation is frontier research at maximum scale, the California offer will continue to win.

The Canadian investor ecosystem

The announcement is also a signal to other Canadian capital pools that AI is a sector worth taking seriously. Pension funds, which collectively manage over $3 trillion in Canadian capital, have been cautious on AI venture investing, in part because the returns have historically accrued to American companies they could not access at early stages. If the Venture Scientist Fund produces a meaningful cohort of Canadian AI companies at scale, follow-on rounds will require capital that is currently not being deployed by Canadian institutional investors.

CDPQ, Quebec's large pension fund, has been more active than most Canadian pensions in Quebec-based technology. OMERS, PSP, and CPP Investments have dabbled in direct AI venture but at much smaller scale than their American counterparts. The Venture Scientist Fund's success will partly be measured by whether it creates a compelling enough deal pipeline to change that calculus.

Sovereign AI and the compute question

Running alongside the commercialisation push is a separate debate about whether Canada should invest in sovereign AI compute infrastructure, meaning domestically owned, domestically hosted training and inference capacity for large models. The Mila-Hypertec-5C LaSalle campus announcement includes sovereign AI research hub language specifically because the Quebec partners have concluded that Canadian AI companies and public-sector AI users should not be entirely dependent on American cloud providers for the compute that powers their systems.

The Venture Scientist Fund does not itself build compute infrastructure, but it invests in companies that need it. How the fund's portfolio companies decide where to run their training workloads, and what the cost and availability tradeoff looks like between Canadian sovereign capacity and American hyperscaler capacity, will shape the economics of Canadian AI commercialisation in ways that are not always visible from the investment announcements. Federal and provincial co-investment in compute capacity, currently under active discussion in Ottawa and Quebec City, is the less-visible partner to the venture fund and may matter more over the next five years than the direct investment dollars do.

What's next

The fund will begin deploying capital immediately, with the partners indicating that first investments will be announced through the spring and summer. A dedicated programme for researchers to engage with the fund before formal incorporation is being built inside Mila, and similar engagement programmes are expected to launch at Vector and Amii in the coming months. The number of investments publicly disclosed in the first year will be the earliest signal of whether the fund is finding the deal volume its thesis requires.

For Canada's AI sector, the Venture Scientist Fund is the most tangible attempt yet to solve a problem the country has been diagnosing for a decade. Its results will matter beyond the AI industry: if Mila and Inovia can successfully convert a Canadian research advantage into Canadian companies at scale, the template they build will be studied by every other Canadian research commercialisation effort that has watched American firms monetise Canadian science.