The Canadian AI Funding Guide
SR&ED, IRAP, ESSOR, Scale AI, Mitacs — there's more money available for Canadian AI projects than most teams realize. Here's how to find and access what you qualify for.
Most Canadian businesses are sitting on money they don't know exists.
Between federal tax credits, Quebec grants, and BDC financing, there are real dollars available for teams adopting AI — whether you're building something new or implementing tools that already exist. The problem isn't access. It's that nobody has explained it clearly.
I put this together because I keep having the same conversation with marketing and business leaders in Montreal: they assume funding is only for tech startups doing deep R&D. It's not. If you're using AI to automate workflows, improve reporting, or train your team — there's likely a program for you.
Use the eligibility quiz below to get a personalized match — takes under 2 minutes.
Federal Programs
Available to registered Canadian businesses across all provinces.
SR&ED — Scientific Research & Experimental Development
The granddaddy of Canadian innovation funding, and it got significantly better in 2026. The expenditure limit doubled to $6M, and capital costs are now eligible again.
The key question with SR&ED is whether your work creates new technical knowledge — not just uses existing tools. If you're implementing ChatGPT with a custom system prompt, that's probably not SR&ED. If you're building a proprietary model, fine-tuning on your data, or developing a new AI-powered process where the outcome isn't guaranteed, you likely qualify.
- Document your technical uncertainties throughout the project — meeting notes, code logs, design decisions, failed experiments
- Calculate eligible expenses: salaries, 55% overhead on qualifying salaries, materials, subcontractor fees (80%)
- File Form T661 with your corporate tax return within 18 months of your fiscal year-end
- New for 2026: pre-claim approval lets you confirm eligibility before filing, cutting review time from 180 to 90 days
A SaaS company spent 6 months building a recommendation engine, testing three different collaborative filtering approaches before landing on a hybrid deep learning model. No guaranteed outcome, documented experiments, novel technical challenge. Claim: $180K in salaries + $25K cloud compute. Refund at 35%: ~$68K.
IRAP — Industrial Research Assistance Program
NRC IRAP is one of the most underused programs in Canada. Non-repayable funding, covers up to 80% of salary costs and 50% of subcontractors. In 2026 they launched a dedicated AI Assist stream with $100M over five years.
- Find your regional NRC-IRAP advisor (free, no commitment)
- Prepare a project proposal: business problem, AI approach, team, timeline, budget
- Submit through NRC portal — decision in 4–6 weeks
- Get reimbursed monthly or quarterly as you spend
A 50-person manufacturer proposed a computer vision system for predictive maintenance. Team: 1 ML engineer ($90K) + 1 manufacturing engineer ($70K) + $40K subcontractor. IRAP covered 80% of salaries = $128K non-repayable. Their co-investment: $72K.
BDC — Data to AI Program
Not a grant — a loan. But worth knowing because it covers things grants don't: consulting fees, software acquisition, implementation costs. Up to $5M, up to 8 years to repay. If you're doing a serious AI implementation and need capital to fund it, cleaner than a traditional business loan.
BDC Data to AI ↗Scale AI Supercluster
Montreal-based innovation supercluster. Multiple funding streams including an AI Training Program (up to $1M per company) and industry-led project funding covering up to 40% of eligible expenses. Projects need to be collaborative — multiple stakeholders involved.
Check current intake status before applying.
Scale AI ↗Mitacs Accelerate
Connects your business with university AI researchers. $15,000 per 4–6 month internship, with enhanced funding for AI projects. Lower dollar amounts but a good way to get research capacity without hiring full-time.
Mitacs ↗CDAP — Canada Digital Adoption Program
Quebec Programs
Quebec has built one of the richest AI funding ecosystems in North America — a deliberate strategy to attract and retain talent and businesses in the province.
ESSOR — Investissement Québec
Quebec's most accessible grant for AI adoption. Three components depending on where you are: feasibility studies ($20K–$50K), digital diagnostics ($10K–$20K), and digital implementation ($20K–$50K). Stream 1 is open through March 2027. Explicitly covers recommendation engines, predictive models, computer vision, AI-powered SaaS.
- Start with Component 1B (digital diagnostic): hire a qualified consultant, claim 50% back
- Use diagnostic output to build an implementation proposal
- Submit to Investissement Québec — 4–8 week decision
- Execute and submit invoices for reimbursement at 50%
A Quebec retail chain did a $18K diagnostic (received $9K grant). Diagnostic recommended AI inventory forecasting. Implementation budget: $45K. ESSOR covered $22.5K. Result: 18% inventory reduction, $400K annual savings.
CDAEIA Tax Credit
Quebec's AI-focused evolution of the former e-business tax credit. 30% of eligible salaries — 22% refundable, 8% non-refundable. Applies to fiscal years starting after December 31, 2025. If you have Quebec employees working on AI projects, this is money sitting in your tax return.
Track employee time on eligible AI work throughout the year. Claim on your Quebec corporate tax return. Works cleanest with a tax specialist who knows the program.
A Quebec fintech (15 employees) built an AI loan assessment system. 8 employees contributed, $1.2M eligible salary pool. CDAEIA at 30% = $360K. Refundable portion: $264K back in cash.
PROMPT — Collaborative Innovation
Quebec's research partnership fund. Up to $500K/year for collaborative AI projects. Dedicated PARTENAR-IA stream for AI development and commercialization. Requires partnering with a research center. If you're doing anything at the R&D end of AI, worth the process.
PROMPT Innov ↗Can you stack these programs?
Yes — and you should. The savviest companies combine multiple programs to cover different expense types. Here's how it works.
The core rule:The same dollar cannot be claimed in two programs. But different programs cover different things — salaries, overhead, implementation costs, training — so they naturally complement each other. Watch two caps: federal programs combined cannot exceed 75% of your total eligible costs; Quebec's ESSOR caps cumulative government aid at 80%.
IRAP covers 80% of salaries as you go. SR&ED covers overhead and materials retrospectively. Example: $400K AI R&D project. IRAP covers $280K (salaries). SR&ED claims $120K (overhead + materials) and returns $42K at 35%. Total government support: $322K. Your co-investment: $78K. Note: IRAP funding is "government assistance" and reduces your SR&ED eligible base dollar-for-dollar — track carefully with an accountant.
ESSOR covers 50% of implementation costs. CDAEIA covers 30% of salaries on the same project. Different expense categories — they layer cleanly. Example: $100K implementation project + $160K in employee salaries. ESSOR: $50K. CDAEIA: $48K. Total government support: $98K on a $260K project.
Use IRAP for R&D phase (build it). Use ESSOR for the Quebec deployment phase (roll it out). Clean sequential logic, different expense types, no overlap.
CDAEIA claims 30% of qualifying AI staff wages. SR&ED claims the R&D overhead and remaining eligible costs on the same project. Different portions of the same work — both apply.
Frequently asked questions
If you're using it as-is, no. If you're building custom AI on top of it — fine-tuning, developing novel processes, solving a technical problem that isn't guaranteed to work — yes. The key is engineering effort and documented uncertainty, not the software itself.
Only if you're building proprietary AI tools — a custom attribution model, a demand forecasting engine, something novel. Using existing tools to deliver client work doesn't qualify. The distinction: Are you doing R&D to build something, or using tools to deliver a service?
Almost everything: software licences, cloud costs, consultant fees, training. The distinction is between deductible (reduces taxable income) and SR&ED-eligible (generates a cash refund on top of deduction). Maximize SR&ED first — it's more valuable. Deduct the rest normally.
60 days if accepted, up to 180 days if audited (90 days with new pre-claim approval). In practice: expect 4–6 months. If you need money faster, IRAP or Quebec grants pay during the project.
Plugging in a ChatGPT API? No. Building a custom solution where GPT is one component of novel work with documented uncertainty? Possibly SR&ED or IRAP eligible. The bar is effort and uncertainty, not the technology stack.
Yes. Scale AI (when open) funds custom training programs. ESSOR covers training as part of implementation. IRAP can include training if it's project-specific. Standard training costs are deductible regardless.
IRAP is a grant — upfront, non-repayable, paid as you go, requires an application and NRC advisor. SR&ED is a tax credit — filed with your return, covers broader expenses, paid months later. Use both: IRAP funds the people doing the work, SR&ED refunds you for what you spent.
No, but first-timers usually benefit from one. A $2K–$5K advisor fee often uncovers $20K–$100K+ in eligible work you'd otherwise miss. Worth it once.
Yes — that's the whole stacking strategy. Watch cumulative caps (75% federal, 80% Quebec), don't double-claim the same expense, and make sure IRAP funding is disclosed on your SR&ED form.
CRA asks for documentation: project timelines, expense records, meeting notes, code history. Most audits end in a negotiated settlement — you keep 70–90% of what you claimed. Poor documentation is the main reason claims get reduced. Don't panic, do get professional help if it happens.
ESSOR covers 50% of consulting as part of implementation. SR&ED and IRAP cover contractor fees if they're doing the technical R&D work. Pure advisory consulting (telling you what to do) is deductible but rarely grant-eligible on its own.
Contact your regional NRC-IRAP advisor for a free preliminary assessment — no fee, no obligation, answer within a week. Ask yourself first: Is there genuine technical uncertainty? Commercial value? Can it be done in 12–24 months with a real budget? If yes to all three, it's worth the call.
Where to start
Five steps. No fluff.
Rough spreadsheet: who's working on AI, at what salary, for how long? What software, cloud costs, consultants? This determines which programs are worth pursuing.
Building new AI internally → SR&ED + IRAP. Implementing existing solutions in Quebec → ESSOR + CDAEIA. Need capital for a large rollout → BDC Data to AI. Want to train employees → Scale AI or Mitacs.
NRC-IRAP advisor: free, no commitment. Investissement Québec consultant: same. SR&ED specialist: most offer a free 30-min chat. Don't guess — these conversations cost nothing and save months.
SR&ED has no application process — just claim it on your next tax return. CDAEIA same. File both while you're pursuing grant applications.
IRAP: 1–2 page project summary to your NRC advisor. ESSOR: portal submission. BDC: online application. All have 4–6 week decision timelines. These programs are meant to be accessible.
Last updated: April 2026. Funding programs change — always verify directly with the program before applying. I update this page quarterly.