R&D Tax Credits for UK Startups

Does software development qualify for R&D tax credits?

Yes — software development is one of the most commonly qualifying categories of R&D in the UK, and accounts for a significant proportion of total SME claims. HMRC's information and communication sector (which covers software, SaaS, and digital product companies) is consistently the second-largest sector by number of claims, with over 13,000 claims per year.

Qualifying software R&D typically includes building novel algorithms that solve computational problems without a known prior solution, developing new approaches to data processing, storage, or retrieval that are not available from existing tools, creating AI or machine learning systems whose architecture and performance characteristics were uncertain at the outset, and integrating disparate systems in ways that require solving genuine interoperability challenges.

Routine software development — building a standard e-commerce site using established frameworks, configuring existing software, or styling front-end components — does not qualify. The dividing line is whether your engineers were solving known problems with known tools, or genuinely uncertain technical problems requiring investigation.

Which startup sectors generate the most qualifying R&D?

While any sector can generate qualifying R&D, certain industries produce consistently high volumes of eligible expenditure per company. Life sciences, pharmaceutical, and medical device startups routinely qualify for large claims due to the scientific nature of their development work. Advanced manufacturing, materials science, and engineering companies qualify for process and materials R&D. Fintech companies qualify for algorithm development, risk modelling systems, and security infrastructure work.

Gaming studios qualify for engine development, procedural generation systems, and graphics rendering work that goes beyond what existing engines provide. Cleantech and climate technology companies qualify for energy system modelling, sensor development, and novel material applications. Agricultural technology companies qualify for precision farming systems, crop monitoring algorithms, and biotechnology applications.

What does not qualify for R&D tax credits?

Certain activities are explicitly excluded from HMRC's definition of qualifying R&D, regardless of how technically complex they appear. Arts, humanities, social sciences, and economics are excluded — qualifying R&D must be in natural science, engineering, or technology. Market research, consumer testing, and commercial feasibility studies do not qualify, even if they involve data analysis.

Capital expenditure on land or buildings does not qualify, though capital expenditure on software and equipment used in R&D does. The cost of obtaining patents and licences is excluded. Work that applies existing techniques to solve known problems — even if the application is sophisticated — does not qualify if a competent professional in the field could have reached the same solution without genuine uncertainty.

Can a loss-making startup claim R&D tax credits as cash?

Yes — and for most early-stage startups, the loss-making position makes R&D credits more valuable, not less. Under the merged scheme, a company in a loss position receives the credit as a direct cash payment from HMRC at the applicable rate (20% or 27%), rather than as a reduction in a Corporation Tax bill that does not yet exist.

This means a startup spending £300,000 on qualifying R&D in a year where it is loss-making receives either £60,000 (at 20%) or £81,000 (at 27% if R&D-intensive) as a cash payment from HMRC, typically within four to six weeks of a well-prepared claim submission. For pre-revenue startups, this is often the largest single cash inflow in the financial year.

The only restriction on payable credits under the merged scheme is the PAYE/NIC cap: the payable credit cannot exceed £20,000 plus 300% of the company's total PAYE and NIC liability for the period. For startups with a payroll, this cap rarely binds, but very early pre-employee startups should be aware of it.

How do you calculate the size of an R&D tax credit claim?

Calculating your R&D credit starts with identifying qualifying expenditure across four main cost categories: employee costs (salaries, employer NIC, and pension contributions for employees directly engaged in R&D), externally provided workers (agency workers and contractors at 65% of the payment for merged scheme claims), consumable materials used up or transformed in the R&D process, and software licences and cloud computing costs directly and necessarily used in qualifying R&D.

Not all employee time qualifies — only time spent directly on qualifying R&D projects. If an engineer spends 60% of their time on qualifying R&D and 40% on routine product maintenance, only 60% of their employment costs are includable. Documenting this allocation is one of the most important parts of claim preparation, and one of the areas where specialist accountants add the most value compared to a general practitioner preparing the claim without a technical review.

Subcontractor costs are included at 65% of the qualifying payment under the merged scheme, reflecting HMRC's assumption that the subcontractor receives some benefit from performing the R&D work. Payments to connected parties (related companies, directors) are capped at the lower of the actual payment and the connected party's own cost.

Staff costs (the largest qualifying category for most startups)

Qualifying staff costs include gross salaries, employer National Insurance contributions, and employer pension contributions for employees who directly engage in qualifying R&D activities. Agency workers and contractors are included at 65% of the qualifying payment.

The allocation of staff costs between qualifying and non-qualifying activity is done on a time-apportionment basis. If a developer spends 70% of their time on qualifying R&D projects and 30% on bug fixes and routine maintenance, 70% of their employment cost is includable. Larger companies typically use timesheets; smaller companies can use a reasonable estimate supported by documentation of the projects involved.

Directors who work on R&D qualify in the same way as employees. Founder-engineers who spend significant time on technical development are often the largest single qualifying cost in an early-stage startup's claim, and are frequently excluded from self-prepared claims because founders do not realise their own time qualifies.

Software and cloud computing costs

Software licences used directly in qualifying R&D are eligible expenditure under the merged scheme, as are cloud computing costs — including AWS, Google Cloud, and Azure — when those costs are directly and necessarily incurred in performing qualifying R&D activities.

The key qualifier is 'directly and necessarily.' Cloud compute costs for running production infrastructure do not qualify. Cloud compute costs for running model training jobs, running test environments for qualifying R&D, or performing qualifying data processing operations do qualify. Splitting cloud costs between qualifying and non-qualifying use is typically done on a usage basis (CPU hours, storage consumed) rather than cost allocation.

Licence costs for tools like Matlab, specialist simulation software, or scientific computing libraries qualify when the licence is used primarily for qualifying R&D. General-purpose tools like office software do not qualify, even if they are used incidentally in R&D projects.

Consumable materials

Consumable materials that are used up or transformed in the R&D process qualify at 100% of cost. This category is most significant for hardware, manufacturing, life sciences, and materials science startups where physical prototyping is part of the development process.

A biotech startup consuming reagents in laboratory experiments qualifies for the reagent costs. A hardware startup destroying prototype circuit boards in testing qualifies for the component costs. A food technology company consuming ingredients in recipe development qualifies for the ingredient costs. The material must be consumed in the R&D process — materials that are retained, sold, or incorporated into a finished product do not qualify.

How does HMRC review R&D claims — and what triggers an enquiry?

HMRC processes most well-prepared R&D claims without opening a formal enquiry. Claims that are submitted with a full technical narrative, properly documented cost analysis, and accurate CT600L data are typically processed within four to six weeks of submission.

Claims that trigger HMRC enquiries tend to share common characteristics: high claimed costs relative to the company's apparent technical complexity, vague or generic technical narratives that do not specifically describe the uncertainties addressed, inclusion of clearly non-qualifying costs, or material inconsistencies between the technical narrative and the financial records. HMRC's compliance unit has become significantly more active in R&D enquiries since 2022, and poorly prepared claims now face substantially higher scrutiny than they did in previous years.

The best defence against an enquiry is a claim that would satisfy a competent professional reading it cold. The technical narrative should describe specific projects, specific uncertainties, and specific approaches taken to resolve them — not generic descriptions of the company's technology. The cost analysis should be supported by payroll records, timesheets or time estimates, and software contracts that a reviewer can check against the narrative.

How long does an R&D tax credit claim take from preparation to payment?

A well-prepared claim submitted as part of the Corporation Tax return typically receives payment or tax credit from HMRC within four to six weeks of submission. HMRC's processing target for R&D repayment claims is 28 days for straightforward submissions.

Claims that require additional information from HMRC, or that are selected for compliance checks, take considerably longer — typically three to six months for a standard compliance check, and potentially 12 to 18 months for a formal enquiry. The risk of compliance checks has increased significantly since 2022, making claim quality the most important determinant of timeline.

SaaS and software companies

SaaS and software startups typically have the highest proportion of qualifying R&D to total expenditure of any sector, because their primary cost is engineering labour and their primary activity is technical development. The qualifying activity threshold is crossed whenever engineers are solving problems that required genuine investigation rather than applying known patterns.

Commonly qualifying activities include building novel database architectures that solve performance or scalability problems not addressed by existing solutions, developing distributed systems with novel consistency or fault-tolerance properties, creating machine learning pipelines where the model architecture and training approach were uncertain, and building real-time processing systems with novel latency or throughput characteristics.

Frequently missed costs include founder-engineer time (directors who are also the technical leads qualify in full), cloud computing costs for model training and test environments, open-source library development where the uncertainty relates to the library's technical approach, and integration work that involved resolving genuine API incompatibilities rather than applying documented integration patterns.

Life sciences, medtech, and healthtech

Life sciences and medtech companies typically have large, well-documented qualifying R&D activities due to the scientific nature of their work, but sometimes underestimate their qualifying costs by excluding work that does not result in a regulatory filing. Pre-clinical research, failed experiments, and discontinued development programmes all qualify if they involved genuine scientific uncertainty.

Qualifying activities include compound or device development, clinical data analysis involving novel statistical methodologies, software development for diagnostic or monitoring devices, and manufacturing process development for novel biologics or devices. Where a company uses Contract Research Organisations (CROs) or Contract Development and Manufacturing Organisations (CDMOs), the subcontractor costs qualify at 65% of the qualifying payment under the merged scheme.

Gaming studios

UK gaming studios qualify for both R&D tax credits and Video Games Tax Relief (VGTR) — two separate schemes that can be claimed simultaneously for qualifying projects. R&D credits apply to technical development work (engine development, rendering systems, physics simulation, AI/NPC behaviour systems) that involves genuine technical uncertainty. VGTR applies to the overall production costs of qualifying British video games.

The distinction between qualifying R&D and routine game development is important. Building a new rendering technique that solves a performance problem not addressed by existing engines qualifies. Configuring Unreal Engine to render a specific art style does not. Developing a novel procedural generation system whose statistical properties were uncertain qualifies. Using established procedural algorithms to generate levels does not.

Fintech and financial technology

Fintech startups frequently underestimate their R&D credit eligibility because they assume financial modelling and algorithm development are not sufficiently 'scientific' to qualify. HMRC's definition explicitly includes technology — and financial technology involves genuine scientific uncertainty when developing novel risk models, fraud detection systems, trading algorithms, or regulatory compliance automation.

Qualifying fintech R&D typically includes developing novel credit scoring algorithms whose predictive accuracy was uncertain during development, building real-time fraud detection systems using machine learning approaches that required experimental validation, creating regulatory reporting systems that involved solving novel data reconciliation challenges, and developing financial simulation models with novel statistical methodologies.