Quick answer: Data and cloud costs are a qualifying R&D expenditure category since 1 April 2023, covering licensed datasets and cloud computing services such as storage, processing and platform access used directly in qualifying R&D. Where the same resource also serves non-R&D purposes, the cost must be apportioned to the R&D share.
What are data and cloud costs for R&D purposes?
Data and cloud costs are a qualifying R&D expenditure category introduced by Finance Act 2022 for accounting periods beginning on or after 1 April 2023. The category covers the cost of licensing datasets used directly in qualifying R&D and the cost of cloud computing services, including data storage, data processing, hardware-as-a-service, and software-as-a-service platform access, where used for qualifying activity. Costs must be apportioned to the R&D use where the same data or cloud resource is also used for non-R&D purposes.
How does HMRC define data and cloud costs?
HMRC guidance is at CIRD82500 of the CIRD Manual, which details the scope and apportionment requirements. The legislation is at section 1125A of the Corporation Tax Act 2009. Note that where data is licensed and then sublicensed or resold, or used for staff training outside the qualifying project, that portion is excluded.
What do data and cloud costs look like in practice?
An AI company spends £80,000 on AWS compute and £20,000 on a licensed training dataset directly used to train a novel machine-learning model that forms the core of a qualifying R&D project. With appropriate apportionment evidence showing the AWS usage was 85% R&D, the claim includes £68,000 of cloud costs and the full £20,000 of data costs.
Worked example: apportioning shared cloud infrastructure
Where the same AWS or Azure account serves both the R&D project and production traffic, HMRC expects a defensible split - usage logs, compute-hours, or headcount allocation are all accepted evidence. Extending the AI company example above: if the same £80,000 AWS bill also covered £15,000 of unrelated production hosting, the apportionment falls to £65,000 (81%) rather than £68,000, changing the qualifying cloud total from £88,000 to £85,000 - close to the 2023-24 average SME-scheme claim of roughly £85,400.
Data and cloud costs alongside other categories
Data and cloud costs rarely stand alone in a claim; they usually sit next to staff costs and consumables for the same project. Companies unsure whether their AWS, Azure or GCP spend is apportioned correctly should check the wider eligible expenditure categories before finalising a claim. Because cloud spend is usually itemised monthly by the provider, the strongest apportionment evidence ties each invoice line to a specific project code or tagging convention set up at the start of the accounting period, rather than an estimate produced retrospectively when the claim is being prepared. Providers such as AWS, Azure and Google Cloud all offer cost-allocation tagging at the project or resource-group level, and setting this up before the accounting period begins is far less work than reconstructing an apportionment percentage from raw billing data months later when the claim is being prepared.
Related terms
Frequently asked questions
Only where the platform access is used directly for the qualifying R&D activity itself, not for general business administration. A data-science team's model-training platform can qualify; the company's accounting software does not. Use the free eligibility calculator for an indicative view of how cloud and data costs contribute to a claim.
Usage logs, compute-hours, or headcount allocation are all accepted evidence for a defensible split between R&D and non-R&D use of the same cloud account.
For accounting periods beginning on or after 1 April 2023, following the change introduced by Finance Act 2022.
Data that is sublicensed or resold, or cloud or data use for staff training outside the qualifying project, is excluded from the category.