Does quantum or advanced science work qualify for R&D tax relief?

This note covers quantum computing, sensing, photonics and communications, advanced materials and optics, and deep-tech research businesses generally.

Why these claims are harder than the science suggests

Three things make deep tech more difficult to claim correctly than sectors with far less impressive engineering in them.

The competent professional is genuinely hard to identify. The whole test turns on what a competent professional working in the field could readily deduce, and HMRC’s guidance says the phrase “has not been defined as the natural meaning is considered to be self-explanatory”. In a young field it is not self-explanatory at all. What is settled to a specialist in one sub-field may be an open question to an equally qualified specialist next door, and there may be only a few dozen people anywhere who could adjudicate.

Uncertainties are coupled across disciplines. A single project can carry live unknowns in control electronics, cryogenics or vacuum systems, materials, firmware and algorithms at the same time. Separating genuine uncertainty from difficult-but-known engineering has to be done discipline by discipline, and a claim that treats the whole apparatus as one uncertainty will not survive a competent reader.

Research and commercial work run in parallel. Grant-funded research, university collaboration and customer-funded development often proceed simultaneously, sometimes on the same hardware and the same people’s time. That makes the cost boundary the most error-prone part of the claim.

Framing the advance at the right level

This is the single most valuable thing to get right, and it is where most deep-tech narratives fail.

Take a project that uses a particular element in a novel device. Framed as “using that element”, there is no advance — the element and its properties have been understood for decades, and a reviewer can find that in an afternoon. Framed as the entire finished instrument, the claim is too broad to be evidenced. The advance is usually the narrow thing in between: not the material, but the material in that configuration, at that scale, under those constraints, where nothing published tells you whether it can be made to work.

The discipline is to state the advance as a capability the field lacked, and then check whether a competent professional could have said in advance how to achieve it. If the answer is obviously no because the statement is too grand, the frame is too wide. If the answer is obviously yes, it is too narrow. The level that survives is nearly always more specific than the company’s own description of what it is building — see what is an advance in science or technology.

The same discipline tells you what to leave out. Assembling a system from established, commercially available components is not R&D even where the application is novel, and applying a documented algorithm to a new dataset is not R&D, however striking the result. The advance has to be in the underlying science or technology, not in what it is pointed at.

Projects that run past the year end

Deep-tech projects routinely span several accounting periods, and the uncertainty is often still live at the year end. That is not a problem: “even if the advance in science or technology sought by a project is not achieved or not fully realised, R&D still takes place”.

So a period can properly contain design and development work on an advance that is resolved — or abandoned — a year later, and the claim for that period stands on the attempt. What the claim has to do is say honestly where the uncertainty had got to at the period end, rather than describing the project as though it were finished. An entry that reads as a completed success story, filed while the work is still running, is both less accurate and less persuasive than one that states the position as it stood.

Two consequences follow. The advance has to be re-tested against the published state of the art each period, because in a fast-moving field the baseline moves — something uncertain in one year may be published by the next. And the boundary of each period’s claim is a matter of record, not reconstruction.

Grants, collaborations and who claims

Deep-tech companies are the most heavily grant-funded claimants there are, and the rules changed.

For accounting periods beginning on or after 1 April 2024, a grant does not reduce an R&D claim in the way it once did. For earlier periods — still live on open years and amendments — it very much could, and the treatment depended on the terms of the award rather than on who funded it. Both positions are set out in how grant funding affects an R&D claim.

University collaborations and consortium projects need their own look, because the question of who is treated as carrying out the R&D is separate from who paid for it. Where a company directs the work and carries the risk, the claim is usually its own; where it has been engaged to perform someone else’s research programme, it usually is not.

One scheme point worth raising early rather than late: a loss-making company spending a high proportion of its total expenditure on R&D — which describes a great many deep-tech businesses before revenue — may fall within ERIS, which is more generous than the merged scheme. Whether the PAYE and NIC cap restricts the payable amount is worth computing at the same time rather than discovering afterwards.

The cost boundary, which is where the money is lost

In most sectors the qualifying-cost exercise is mechanical. In deep tech it is a judgement made repeatedly, because the same people, the same equipment and often the same written records cover qualifying and non-qualifying work in the same week.

The practical risk is documentary. Progress reports, grant returns and board packs in these businesses habitually describe research and commercial delivery together, in one narrative. Those are exactly the documents an HMRC officer will ask for, and an apportionment that cannot be reconciled with them will not hold, however reasonable it was when it was made. The apportionment therefore has to be defensible against the company’s own reporting, not merely arithmetically sound.

Consumables deserve a second look too. Cryogens, specialist gases, wafers, optics and targets consumed in experimental work are qualifying material; the same items consumed in producing something delivered to a customer are not.