R&D Tax Credit for Software Companies: 2026 Eligibility Guide
Published 2025-02-06
R&D Tax Credit for Software Companies: 2026 Eligibility Guide
Quick Answer
Software companies are among the largest beneficiaries of R&D tax credits. Qualifying activities include developing new features with technical uncertainty, optimizing performance where outcomes are unknown, building novel architectures, and solving scalability challenges. Typical credits range from $70,000-$150,000 per $1 million in developer wages. SaaS companies, product companies, and custom software developers all qualify—business model doesn’t matter, only the nature of development activities.
Key Takeaways
- Software development often qualifies - If technical uncertainty exists
- 70-90% of developer time may qualify - Depending on activity mix
- Cloud costs count as supplies - When allocated to R&D environments
- SaaS business model doesn’t affect eligibility - Only activities matter
- Document through existing tools - Git, JIRA, project management
- Typical credit: 7-14% of qualifying wages
Why Software Companies Qualify Strongly
Natural Alignment with 4-Part Test
| 4-Part Test | Software Development Alignment |
|---|
| Technological in nature | Computer science, engineering |
| Process of experimentation | Testing, iteration, debugging |
| Elimination of uncertainty | Unknown performance, scalability |
| Qualified purpose | New/improved software products |
Typical Credit Value
| Annual Developer Wages | Typical Credit Range |
|---|
| $500,000 | $35,000 - $70,000 |
| $1,000,000 | $70,000 - $140,000 |
| $2,000,000 | $140,000 - $280,000 |
| $5,000,000 | $350,000 - $700,000 |
Assumes ASC method, first-time or growing filer, 70-85% qualifying time
Qualifying Software Activities
What Typically Qualifies
| Activity Category | Examples | Why It Qualifies |
|---|
| Algorithm development | New search algorithms, optimization | Technical uncertainty |
| Architecture design | New system architectures | Unknown performance |
| Performance optimization | Speed, latency improvements | Experimental approach needed |
| Scalability work | Handling 10x users | Unknown if achievable |
| Security features | New authentication methods | Technical challenges |
| API development | Complex integrations | Uncertainty in approach |
| Database optimization | Query performance | Unknown outcomes |
| ML/AI development | Model training, optimization | Experimental by nature |
What Does NOT Qualify
| Activity Category | Examples | Why Not Qualified |
|---|
| Routine maintenance | Standard updates | No uncertainty |
| Bug fixes (simple) | Known solutions | No experimentation |
| Cosmetic changes | UI styling, colors | Aesthetic, not technical |
| Configuration | Setting up known systems | Standard process |
| Documentation | User guides, comments | Not technical research |
| Testing (QC) | Standard test execution | Not experimental |
| Training | Learning new tools | Not research |
| Project management | Planning, meetings | Not technical work |
The Gray Zone: Bug Fixes and Refactoring
| Scenario | Qualifies? | Analysis |
|---|
| ”Fix typo in error message” | No | Simple, known fix |
| ”Resolve race condition in distributed system” | Yes | Unknown cause, experimentation |
| ”Refactor for readability” | No | No technical uncertainty |
| ”Refactor to handle 100x load” | Yes | Uncertain if achievable |
| ”Standard library upgrade” | No | Known process |
| ”Migrate to new architecture” | Sometimes | Depends on uncertainty |
SaaS Company Specifics
Development vs. Production Allocation
For SaaS companies, properly separating R&D from production is critical:
| Environment | Typical Allocation |
|---|
| Development | 100% R&D |
| Testing/Staging (experimental) | 80-100% R&D |
| CI/CD pipeline | 70-90% R&D |
| Production | 0% R&D |
| Monitoring/Observability | 0% R&D |
Cloud Cost Allocation
Monthly Cloud Bill: $50,000
R&D Environments:
Development servers: $12,000 → 100% R&D = $12,000
Training/ML instances: $8,000 → 100% R&D = $8,000
Test environments: $6,000 → 85% R&D = $5,100
CI/CD: $4,000 → 80% R&D = $3,200
Non-R&D:
Production servers: $15,000 → 0% R&D
CDN: $3,000 → 0% R&D
Monitoring: $2,000 → 0% R&D
Monthly Supply QRE: $28,300
Annual Supply QRE: $339,600
Customer-Driven Development
| Feature Request Type | Qualifies? | Analysis |
|---|
| Simple feature with known approach | No | No uncertainty |
| Complex feature requiring innovation | Yes | Technical uncertainty |
| Integration with standard API | Usually no | Known process |
| Novel integration solution | Yes | Experimental approach |
Software Role Qualifying Percentages
| Role | Typical Qualifying % | Key Factors |
|---|
| Backend Developer | 75-95% | API, database, algorithms |
| Frontend Developer | 40-75% | Complex UI features may qualify |
| Full-Stack Developer | 65-90% | Depends on feature complexity |
| DevOps Engineer | 40-70% | Infrastructure for R&D |
| Platform Engineer | 60-85% | Development platform work |
| QA Engineer (test dev) | 50-80% | Test automation, performance |
| Data Scientist | 80-100% | Model development |
| ML Engineer | 85-100% | Model training, optimization |
| Tech Lead | 50-80% | Technical decision-making |
| Product Manager | 5-25% | Only technical requirements |
| UX Designer | 0-20% | Generally non-technical |
| Scrum Master | 0-10% | Not technical work |
Git as Documentation
Well-written commits support R&D claims:
✓ Good commit messages (showing experimentation):
- "Experiment with Redis clustering for session scale"
- "Test alternative approach to query optimization"
- "Prototype new caching strategy for 10x throughput"
- "Evaluate React vs. Vue for complex UI component"
✗ Poor commit messages (no evidence):
- "Fix bug"
- "Update code"
- "Improve performance"
- "Misc changes"
JIRA/Project Management as Evidence
| Artifact | R&D Documentation Value |
|---|
| User stories with technical detail | Shows uncertainty |
| Sprint planning | Shows project scope |
| Technical design docs | Demonstrates approach |
| Retrospectives | Documents lessons learned |
| Performance benchmarks | Proves experimentation |
Time Tracking Approaches
| Method | Pros | Cons |
|---|
| Weekly timesheets by project | Most defensible | Requires discipline |
| Project-level allocation | Simpler | Less precise |
| Role-based percentages | Easiest | Least defensible |
Recommendation: Track time at project level weekly
Software QRE Calculation Example
Company Profile
Business: B2B SaaS platform
Employees: 30 (22 technical)
Annual developer compensation: $2,800,000
Cloud costs: $400,000
Contractor payments: $120,000
QRE Calculation
Wage QRE:
Backend developers (8): $960,000 × 85% = $816,000
Frontend developers (4): $400,000 × 60% = $240,000
Full-stack developers (5): $600,000 × 80% = $480,000
DevOps engineers (2): $240,000 × 55% = $132,000
Data scientists (2): $320,000 × 90% = $288,000
QA engineers (1): $80,000 × 65% = $52,000
Total Wage QRE: $2,008,000
Supply QRE:
Cloud costs: $400,000
R&D allocation: 70%
Supply QRE: $280,000
Contract Research QRE:
Contractor payments: $120,000
Qualifying work: 100%
Contract Research QRE: $120,000 × 65% = $78,000
Total QRE:
$2,008,000 + $280,000 + $78,000 = $2,366,000
Credit Calculation
Prior 3-year average QRE: $1,200,000
Base amount (50%): $600,000
Current QRE: $2,366,000
Incremental QRE: $1,766,000
Federal credit (ASC): $1,766,000 × 14% = $247,240
Result: ~$247,000 annual federal credit
Machine Learning and AI Development
Strongly Qualifying Activities
| ML Activity | Why It Qualifies |
|---|
| Novel model architectures | Unknown performance |
| Feature engineering experimentation | Testing alternatives |
| Hyperparameter optimization | Systematic experimentation |
| Custom training pipelines | Technical challenges |
| Performance optimization | Uncertain outcomes |
| Transfer learning research | Unknown generalization |
ML-Specific Documentation
| Artifact | Value |
|---|
| Experiment tracking (MLflow, W&B) | Experimentation evidence |
| Model training logs | Process documentation |
| Hyperparameter records | Systematic testing |
| Performance benchmarks | Results evidence |
ML Role Qualifying Percentages
| Role | Typical % |
|---|
| Research Scientist | 90-100% |
| ML Engineer | 85-100% |
| Data Scientist | 80-95% |
| ML Ops Engineer | 50-75% |
| Data Engineer | 40-70% |
Internal-Use Software Considerations
The Internal-Use Rule
Software developed for internal use has additional requirements:
| Standard Internal-Use Test | Requirement |
|---|
| Innovative | Novel, not just new to you |
| Significant economic risk | Substantial resources committed |
| Substantial resources | High development cost |
Exception: Software for Sale/Lease/License
Most SaaS and product companies fall under this exception:
| Software Type | Internal-Use Rules Apply? |
|---|
| SaaS product | No - Software is licensed |
| Mobile app | No - Sold/distributed |
| Custom client software | No - Delivered to clients |
| Internal tools only | Yes - Must meet 3-part test |
Common Software Company Mistakes
Mistake 1: Claiming 100% Developer Time
| Reality | Fix |
|---|
| Not all development qualifies | Track actual time allocation |
| Some projects are routine | Identify qualifying vs. non-qualifying |
| Admin/meeting time is non-R&D | Separate technical from administrative |
Mistake 2: Missing Cloud Costs
| Reality | Fix |
|---|
| Cloud is often 20-30% of QRE | Track R&D vs. production environments |
| Allocation is required | Document allocation methodology |
Mistake 3: Not Documenting Uncertainty
| Reality | Fix |
|---|
| ”It worked” doesn’t show experimentation | Document what was unknown |
| Success hides the process | Record alternatives tested |
Mistake 4: Overlooking Contractors
| Reality | Fix |
|---|
| Contractors at 65% is still valuable | Include qualified contractor work |
| Documentation still needed | Track contractor R&D activities |
Mistake 5: Section 174 Confusion
| Reality | Fix |
|---|
| Section 174 ≠ Section 41 | Credits still available |
| Amortization affects deductions, not credits | Plan for cash flow impact |
State Credits for Software Companies
Top States for Software R&D
| State | Credit Rate | Notes |
|---|
| California | 15% | Transferable, major tech hub |
| New York | 9% | Growing tech scene |
| Massachusetts | 10% | Strong tech presence |
| Washington | B&O credit | No income tax |
| Texas | Franchise tax credit | No income tax |
| Colorado | 3-13% | Growing tech hub |
Multi-State Allocation
For companies with developers in multiple states:
Total QRE: $2,366,000
Allocation by Developer Location:
California: $1,200,000 (51%)
New York: $600,000 (25%)
Remote (various): $566,000 (24%)
State Credits:
California: $1,200,000 × 15% = $180,000
New York: $600,000 × 9% = $54,000
Total State Credits: $234,000
Combined with Federal: $247,000 + $234,000 = $481,000
Documentation Checklist
Frequently Asked Questions
Does open source development qualify?
Yes, if the development meets the 4-Part Test and serves your business purposes (not purely charitable). Contributing to frameworks while solving your own technical challenges can qualify.
What about DevOps work?
DevOps activities supporting R&D infrastructure can qualify. This includes building CI/CD pipelines for R&D, managing development environments, and optimizing R&D-focused infrastructure.
Can remote developers qualify?
Yes. Work location within the US doesn’t affect qualification. Track time the same as on-site developers. Foreign-based developers have different treatment under Section 174.
How do microservices architectures affect qualification?
Microservices development qualifies when it involves technical uncertainty (performance, scalability, reliability). Routine decomposition of known services may not qualify.
What about technical debt reduction?
Generally, technical debt work doesn’t qualify unless it involves solving technical uncertainty (e.g., “refactor to handle 10x scale” vs. “refactor for readability”).
Do QA activities qualify?
Test automation development and performance testing can qualify when they involve technical challenges. Routine test execution does not qualify.
Disclaimer: Software R&D credit claims involve complex technical and tax determinations. This guide provides general information. Consult a qualified tax professional for advice specific to your development activities.