A Qualification Is Not an Approval: What Actually Gates the Organ-on-Chip Market

The document that matters is a docket, not a datasheet

The organ-on-chip industry has spent a decade arguing that its models predict human biology better than animals. The argument is largely won on the science and largely irrelevant to the market, because the question a pharmaceutical company asks is not whether a model is more predictive. It is whether a regulator will accept the resulting data in a submission. Those are different questions, and the second one is answered by a qualification rather than by a publication.

That distinction – a qualification licenses a use, not a product – is the single most important structural fact about this market. It explains why a technology with a twenty-year research history has a commercial base year measured in hundreds of millions of dollars, and it explains why the companies most likely to win are not necessarily the ones with the best chips.

The organ-on-chips market report sizes the category at USD 265.0 million in 2025 rising to USD 1,178.0 million by 2033 at 20.50 percent. A rate that high on a base that small is not a statement about growth appetite. It is a statement about how much of the market is currently gated by evidence rather than by demand.

ISTAND, and what an accepted letter actually means

The FDA’s ISTAND programme – Innovative Science and Technology Approaches for New Drugs – is the mechanism through which a novel method can be qualified for a defined context of use. In September 2024 the agency accepted a letter of intent for a liver-chip model applied to drug-induced liver injury. That acceptance is a real milestone and it is routinely overstated.

A letter of intent is the start of a three-step process: the letter of intent itself, then a qualification plan, then a full qualification package, each reviewed by the agency. The output of the completed process is a statement that a specific model, used in a specific way, to answer a specific question, may be relied on in a regulatory submission. It is not an approval of the device. It is not a licence to sell. It does not travel: a qualification for liver injury in one class of candidate does not extend to kidney, to the blood-brain barrier, or to a different endpoint.

Which is why the correct way to read progress in this market is to count completed qualifications rather than to count accepted letters of intent. On that measure the position as of the end of 2024 was documented by the US Government Accountability Office in its May 2025 review: no organ-on-chip model had completed the qualification process. The same review recorded a finding that suppliers do not advertise: for some applications, work done with chip models cost more and took longer than the conventional approach it replaced.

Why “it takes longer and costs more today” is not a refutation

That GAO finding is quoted in the industry as a criticism and it should be read as a description of an early-stage technology doing qualification work. A new method must be run in parallel with the incumbent method before anyone will accept it, which doubles the cost and extends the timeline by definition. The parallel period ends only when the qualification is granted. So the economics of adoption are negative until the docket closes and positive afterwards, and no amount of instrument improvement changes the sequence.

That is the mechanism that makes this a qualification market rather than a device market. It also explains the shape of the capital: sustained private rounds in the tens of millions rather than a breakout, because the value is being accumulated in evidence that has not yet been recognised. The transient expression market shows what the same gating logic looks like in a different input category: a market that grows with the number of experiments started rather than with the capacity installed, as set out in why a non-clonal process has a structural ceiling.

The standards layer is behind the science

Qualification is harder than it needs to be because there is no common benchmark. The relevant work sits within ISO/TC 276, whose subcommittee SC 2 carries six standards for bioprocess measurement, and the standard most directly relevant to chip characterisation – ISO/CD 25693 – remains at committee draft stage. A committee draft is not a published standard. Until there is one, two laboratories validating the same model have no shared definition of what counts as equivalent performance.

That absence has a commercial consequence that is easy to miss. It means each supplier’s qualification package must build its own comparators, and it means a buyer cannot credibly evaluate two platforms against each other. The result is what the report calls fragmentation by scientific switching cost: a laboratory that adopts one platform does not casually adopt a second, because the two are not merely different products but different experimental systems.

The incumbent that framed the problem honestly

In July 2023 Charles River Laboratories joined the NXTGEN HIGHTECH consortium in the Netherlands, a programme of EUR 1 billion with a further EUR 450 million from the Dutch National Growth Fund spanning some 330 industry and academic partners, with organ-on-chip development and standardisation as one of its biomedical workstreams. Charles River’s own explanation of its interest is the most useful sentence in the category’s public record: the technology has existed for a while, and what has not been seen is the technology in everyday practice in drug discovery, because it needs to be better validated and made more accessible.

A contract research organisation saying that is a statement about where the bottleneck sits. It is not in the chip. It is in the validation work that turns a chip into something a sponsor can put in a submission – and that is a services business with different economics from a hardware business.

What a qualification is worth

Once granted, a qualified context of use is the most durable asset in this market. It cannot be copied, only repeated, and repeating it costs years. It attaches to a specific model and a specific question, and it can be used across programmes by any sponsor who chooses that model. That is a licence-like position without a licence, and it is why the report argues the likely acquirer of a leading pure-play is either a services company buying a validated platform to sell qualification work through, or a tool and reagent major buying a qualified assay to attach to an installed base.

Both would be paying for evidence rather than hardware. The signal to watch is therefore not a product launch. It is the first completed qualification attached to a named commercial platform, and the first transaction priced on the back of it – which is a far better indicator of this market’s trajectory than any instrument specification, including the ones that determine how much each experiment costs, the subject of the companion piece on what a chip costs per data point.