Why SR&ED Matters Even More in 2026
SR&ED has quietly become one of the most powerful non‑dilutive funding tools for Canadian SaaS and AI startups. In 2026, new rules around capital expenditures and higher CCPC limits mean you can recover more of your dev and infrastructure spend than ever before.
Why SR&ED Matters Even More in 2026
If you are building SaaS, AI, or deep‑tech in Canada, SR&ED is now one of your most important funding levers. Recent federal changes have expanded both what you can claim and how much you can recover each year.
SR&ED was already the backbone of Canadian R&D incentives, but the 2024–2025 policy updates—now live for 2026 tax years—add hundreds of millions in new support and reverse a decade‑old cut on capital expenditures. For a technical founder, that means more runway, less dilution, and the ability to invest in serious infrastructure instead of cutting corners.
The Big Shift: Capital Expenditures Are Back
For years, SR&ED basically ignored your hardware. From 2014 onward, most capital expenditures were excluded, so you could claim dev salaries but not the high‑end servers, GPUs, or lab gear that made the R&D possible.
Late‑2024 and 2025 announcements change that. The federal government is restoring capital expenditure eligibility for SR&ED, largely aligning with the pre‑2014 rules. This applies to qualifying depreciable property and certain leases acquired or entered into after the specified effective date (in practice, property acquired from late 2024 onward and used in 2025–2026 SR&ED projects).
What You Can Now Claim
You still need the asset to be used primarily (more than 50%) for SR&ED work, but the scope is much broader than the last decade. Examples that can now fall inside the SR&ED envelope include:
- High‑performance GPU servers and storage dedicated to training models, running large‑scale simulations, or stress‑testing your platform.
- Networking and data‑centre equipment that forms the backbone of your experimental environment (for example, high‑throughput switches and specialized routers in CCA Class 46).
- Specialized hardware rigs or lab setups used for experimental work—robotics test beds, sensor arrays, custom benches, or industrial prototypes.
- Leased equipment or compute where the lease is mainly for SR&ED activity (for instance, a one‑year lease on a GPU cluster or dedicated test hardware).
Instead of treating this as “just capex,” you can allocate the SR&ED‑eligible portion of those assets and effectively get a refund or tax credit on part of the hardware bill.
Did you know you can now claim key hardware again—like on‑prem GPU servers, high‑end dev machines, and specialized test rigs—when they are primarily used for R&D?For infra‑heavy SaaS, especially AI workloads that need serious compute, this change alone can translate into hundreds of thousands back per year.
Enhanced Expenditure Limits for CCPCs
The second major upgrade is how much qualifying spend a Canadian‑Controlled Private Corporation (CCPC) can put through the enhanced federal rate. Historically, CCPCs received a 35% refundable Investment Tax Credit (ITC) on up to $3M of SR&ED expenditures per year, with the rest at 15% and less refundability.
New measures, tied to the 2025 budget and now applying to 2026 tax years, increase that enhanced expenditure limit significantly. Proposals moved the ceiling first to $4.5M and then to $6M of annual qualifying spend at the 35% rate, with a higher phase‑out range for larger companies.
What This Means in PracticeFor a qualifying CCPC:
- Under the old $3M limit, the maximum annual federal refundable credit was $1.05M (35% of $3M).
That is an additional $1.05M per year if your R&D budget can support it. And this is before adding provincial SR&ED or related innovation credits that can push your combined effective rate above 50–60%.
The enhanced limit still phases out as your prior‑year taxable capital and income rise, but the thresholds are more forgiving (for example, phase‑out beginning around $15M in taxable capital instead of $10M, and fully ending at a higher upper bound such as $75M). This keeps the richest support focused on small and mid‑market innovators while giving scale‑ups more headroom.
What Actually Qualifies as SR&ED for SaaS and AI
Even with better limits and capex rules, SR&ED is not a blank cheque. Your work still has to meet the fundamental tests of technological uncertainty, systematic investigation, and advancement.
For a SaaS or AI founder, typical qualifying work might include:
- Designing a new distributed architecture to achieve performance, reliability, or scalability levels that can’t be met with known techniques.
- Developing novel algorithms or ML models where you are experimenting with data structures, training strategies, or optimization methods beyond standard practice.
- Building low‑latency, high‑throughput pipelines (for telemetry, event streams, or edge/IoT data) where the interplay of components is technically uncertain.
- Implementing privacy‑preserving analytics or security controls where you are pushing past documented, off‑the‑shelf solutions.
On the other hand, routine software development—CRUD screens, straightforward integrations, standard cloud migrations, or cosmetic UI changes—does not qualify on its own. The art is in clearly separating experimental work (SR&ED) from productization or maintenance (non‑SR&ED) in your documentation and time tracking.
Example: A Hardware‑Heavy AI SaaS in 2026Imagine your CCPC is building an AI observability platform that must run on‑prem in customer environments, with custom agents, tight latency requirements, and privacy constraints.
To push the boundaries, you:
- Purchase a cluster of GPU servers and specialized network gear to test new training and inference strategies.
- Lease additional off‑site compute for 12 months to experiment with distributed scheduling and fault‑tolerant deployment patterns.
- Have a team of ML engineers and backend devs running structured experiments to see whether you can hit ambitious latency and accuracy targets in hostile conditions.
Under the 2026 rules:
- A significant portion of the hardware and lease costs can now be treated as SR&ED capital or lease expenditures, as long as they are primarily used for those experiments.
- The engineers’ time, qualifying contractors, and a slice of overhead are all SR&ED‑eligible current expenditures.
- Up to $6M of this combined spend can earn 35% federal refundable ITCs, yielding as much as $2.1M back into your bank account, with provincial credits on top.
This is the kind of leverage that can fund an entire extra product team, another year of runway, or the GPU capacity you thought you couldn’t afford.
How to Operationalize SR&ED as a Technical FounderTo really capture this value, you need processes—not just a year‑end scramble with your accountant.
- Classify hardware and leases up front. Tag GPU purchases, servers, specialized rigs, and experimental cloud resources as potential SR&ED assets as soon as you budget for them.
- Map work to uncertainties. For each major initiative, write down the technological uncertainties, hypotheses, and experiments. Tie JIRA/Linear tickets and Git branches to specific SR&ED projects.
- Track time consistently. Use lightweight time codes or sprint‑based allocations so you can attribute engineer time to SR&ED vs non‑SR&ED in a defensible way.
- Layer in provincial programs. Provinces like Quebec, BC, and Ontario add their own refundable or non‑refundable credits; your effective subsidy can be dramatically higher than the federal number alone.
Don’t Leave 35% of Your Dev Costs on the Table
The combination of restored capital expenditure eligibility and a higher enhanced limit for CCPCs means SR&ED in 2026 is structurally better for technical founders than it has been in years.
If you are paying Canadian engineers, experimenting with new architectures, training models, or investing in GPUs and servers to push technical boundaries, there is a good chance a large chunk of that work qualifies.
Don’t leave 35% of your dev costs on the table. Check your SR&ED eligibility for 2026—especially for any hardware, leases, or deep‑R&D projects you’ve budgeted this year.