Cloud Costs and R&D Credits: A Growing Opportunity and a Big Risk
Cloud computing now sits at the core of how many companies conduct R&D. Simulation, modeling, AI and ML training, and software development all run on cloud infrastructure. The question tax teams increasingly face is whether those costs qualify as qualified research expenses under IRC Section 41. They can, but only with careful structuring and documentation. This article covers the legal framework, how different cloud arrangements are treated, and what tax teams should do to position cloud costs for defensibility.
The Legal Landscape: Outdated Rules in a Modern World
The core statute is Section 41(b)(2)(A)(iii) of the Internal Revenue Code. It allows QRE treatment for:
“Any amount paid or incurred to another person for the right to use computers in the conduct of qualified research.”
This provision was originally designed in the pre-cloud era to help smaller companies that couldn’t afford to buy mainframes or servers outright.
The corresponding Treasury Regulation (1.41-2(b)(4)) is also narrow, allowing QRE treatment only for:
- Computers owned and operated by someone other than the taxpayer
- Located off the taxpayer’s premises
- Where the taxpayer is not the primary user of the computer
This regulatory language has not kept pace with modern cloud business models, creating uncertainty and audit risk for taxpayers.
Standard Cloud Computing: A Strong Case
In typical public cloud computing, for example using AWS, Microsoft Azure, or Google Cloud:
- Servers are owned and operated by the cloud provider.
- They are located off the taxpayer’s premises.
- The taxpayer’s workload is routed dynamically across shared infrastructure, making it hard to argue the taxpayer is the “primary user” of any specific server.
This fact pattern aligns well with the requirements of Section 41(b)(2)(A)(iii) and the Treasury Regulation.
Many companies are already successfully claiming these cloud costs as QREs.
Best Practice
- Carefully document how your cloud provider allocates resources.
- Maintain purchase records and technical descriptions showing that the services used meet the statutory and regulatory tests.
- Link cloud costs clearly to qualified research projects and business components.
Private Cloud Computing: More Complex, Still Possible
Things get trickier with private cloud arrangements, where a company pays for dedicated servers, often for security or regulatory reasons.
In these cases:
- The taxpayer may inspect or help maintain the servers.
- The servers may be dedicated exclusively to the taxpayer’s use.
This raises questions about whether the taxpayer is now the primary user or even the effective operator of the hardware.
However, there are many fact patterns where QRE treatment is still viable:
- If the taxpayer is essentially renting capacity in a large data center, with flexible use of multiple servers (rather than one dedicated box used 24/7), the analogy to classic time-sharing still holds.
- If the taxpayer’s use of any specific server does not exceed 50% of its useful life, you have a stronger case.
Best Practice
- Document the terms of the cloud contract carefully.
- Show that the taxpayer does not operate or primarily control the physical servers.
- Demonstrate that server allocation is dynamic and governed by the cloud provider—not by the taxpayer.
Pilot Models and Cloud: A Powerful Alternative Path
There’s another path that many companies overlook:
Section 41(b)(2)(A)(ii) allows QRE treatment for:
“Any amount paid or incurred for supplies used in the conduct of qualified research.”
If your cloud spend is used to run a pilot model. For example testing a prototype software platform or running simulations as part of product development, you may be able to claim those costs as supply costs instead of under the cloud computing provision.
This can be particularly attractive for industries like aerospace, automotive, semiconductors, and advanced manufacturing where simulation-driven R&D is common.
Best Practice
- When using cloud to run a pilot model, classify costs carefully in your study.
- Maintain clear technical documentation showing how the cloud usage supports prototype or pilot model development.
- Consult with your tax advisor to structure this properly.
What Tax Teams Should Do Now
Cloud computing is an increasingly large part of modern R&D. The good news is that you can claim these costs, but you need to do it thoughtfully.
Here’s where to start:
1. Map Cloud Usage to Business Components
Work with engineering and IT teams to connect cloud spend to specific qualified business components. That connection is what establishes the nexus between cost and credit. Without it, cloud costs are difficult to defend under examination.
2. Align Documentation to Current Expectations
Treasury Regulations in this area have not kept pace with modern cloud models, but IRS exam teams still expect granular, activity-level documentation. Tax teams should not assume that outdated regulatory language reduces their documentation burden. It does not.
3. Train SMEs on What Qualifies
SMEs need to explain how specific cloud usage supports qualified research, not just confirm that cloud tools were used. That distinction matters under IRS review. SMEs who understand what the IRS expects produce documentation that holds up. SMEs who do not create gaps that are difficult to close after the fact.
4. Monitor Policy Developments
This area continues to evolve. The IRS has not updated its regulations to address modern cloud arrangements, and future guidance could clarify or restrict what qualifies. Tax teams should track developments and be prepared to adjust their approach as the regulatory landscape shifts.
Bottom Line
Cloud computing is no longer a niche expense; it’s at the heart of how many companies conduct R&D. With the right structuring and documentation, these costs can contribute meaningfully to your R&D tax credit.
If your team would like to review how your cloud spending is currently treated or explore strategies to strengthen your credit claims, we’re here to help. Contact us for a consultation.