Buy or differentiate? A cost model for iPSC neurons

The short answer: differentiating your own iPSC neurons costs roughly $420 to $630 per million cells on our model, against catalogue prices of $350 to $650 per million. Making them in-house beats buying single-million vials, and loses to buying five-million vials.

That is the whole finding, and it is not the one most buy-versus-build discussions reach. The decision is dominated not by whether you differentiate, but by what pack size you would buy instead.

What this model is and is not

Every unit price below was read off a named supplier page or a published institutional rate card on 2026-09-01. Every quantity, hour and yield figure is our assumption, stated explicitly. You should replace our assumptions with yours; the point of publishing the model is that you can.

We have deliberately excluded every line item whose unit price we could not verify. That biases the result in favour of building, because the excluded items are all in-house costs. We say so in the limitations and we return to it below.

The batch we are modelling

Line 1: consumables, per batch

All prices are STEMCELL Technologies published list prices, captured 2026-09-01. Where the page shows a “From” price across multiple pack sizes, we use the “From” figure and name the pack it corresponds to.

Two things stand out. First, medium and supplement are 68 percent of the consumable bill — $1,050 of $1,567 for BrainPhys plus SM1 alone. Second, growth factors are the next largest block at $440 and they are bought as whole packs regardless of batch size, which means the consumable cost per million falls sharply if you scale the batch up.

Line 2: labour

We could not find a published commercial technician rate for cell culture work, so we anchor on three real published hourly rates from institutional rate cards:

We carry the model at two rates and report both columns throughout: $55/hour as the low anchor and $143/hour as the high anchor. At 24 hours per batch that is $1,320 and $3,432 respectively.

If your institution charges you nothing for technician time because the person is on a grant line, the low column is closer to your reality. If you are a company paying a loaded salary, the high column is closer, and probably still low.

Line 3: facility

We use the Salk Institute’s published external non-profit biosafety cabinet rate of $49.00/hour as the facility charge, applied to 24 hours of occupancy per batch: $1,176.00.

This is the least satisfying number in the model. It is a real published rate for a real piece of equipment, but it is a bookable-hood rate rather than an incubator-occupancy rate, and no institution we found publishes an incubator hourly charge. If your lab already owns the incubator and the hood and treats them as sunk, set this line to zero and the model still works — the breakeven table below shows what happens when you do.

Line 4: per-batch QC

One assay. That is the minimum defensible per-batch release test and it is the only one we have carried, which is another way the model favours building. A catalogue vial arrives with a certificate of analysis covering identity, viability, purity against named markers, sterility and mycoplasma. Reproducing that per batch in-house would add several hundred dollars — see the QC testing cost page for what each assay costs.

Line 5: the variable cost per batch

Line 6: the fixed costs

These are one-time and must be amortised across however many batches you will actually run.

The two gene-editing anchors are for constructs that are analogous to, but not identical to, a doxycycline-inducible NGN2 cassette at AAVS1. We are using them as order-of-magnitude anchors for what a targeted knock-in costs at a core facility, not as a quote for that specific edit. If you buy a ready-made inducible line instead, substitute its purchase price here.

The comparator: what buying costs

The BrainXell figures were captured on 2026-09-01; a subsequent automated re-fetch was blocked, so treat them as accurate at capture rather than continuously verified.

The breakeven

Breakeven is the number of batches at which the cumulative in-house cost drops below the cumulative catalogue cost:

breakeven batches = fixed cost ÷ (catalogue cost per batch − in-house variable cost per batch)

Reading the table

Against single-million vials, building wins quickly if your labour is cheap. At the $55/hour anchor, Route A pays back in under two batches. Even the expensive engineered route pays back inside ten batches. A lab running one differentiation a month recovers the setup cost within the first year.

Against single-million vials, building barely wins at all if your labour is expensive. At $143/hour the per-batch saving collapses to $120.40. Route B does not pay back within any realistic project horizon — 91 to 179 batches. At that labour rate you are not saving money by differentiating; you are converting a purchasing decision into a staffing decision and calling it a saving.

Against five-million vials, building never wins. At $350 per million, the catalogue price is below our in-house variable cost in every labour scenario. There is no batch count at which the arithmetic turns over, because the per-batch saving is negative before fixed costs are even considered.

This last row is the most important one on the page, and it is the one that gets left out of most buy-versus-build discussions. Before asking “should we differentiate our own?”, ask “have we priced the largest pack size the supplier sells?” In the data above, moving from a one-million to a five-million vial cuts the per-million price by 46 percent — larger than the entire margin the in-house route is competing for.

Sensitivity: what happens when batches fail

We could not find a published failure rate for iPSC neuron differentiation from any commercial or institutional source, so we do not assert one. Instead, here is what the model does at four failure rates. A failed batch consumes its full variable cost and yields nothing, so the effective cost per successful batch is variable cost ÷ (1 − failure rate).

The high-labour column is brittle. A single failed batch in ten wipes out the entire advantage and puts in-house behind the catalogue. If you are paying loaded staff costs and your protocol is not yet reliable, the model says buy.

The low-labour column tolerates roughly a one-in-three failure rate before parity with single-million vials. That is a substantial cushion, and it is the honest reason academic labs differentiate their own neurons: their marginal labour cost is close to zero and their tolerance for a lost batch is high.

What this model excludes, and why it matters

Every one of these is a real cost of building in-house that we could not price from a named source, and every one of them makes the in-house column an underestimate:

• Coating substrate. Laminin, poly-D-lysine, poly-ornithine or a basement membrane extract. Required. Not priced here.
• Plasticware. Plates, flasks, pipette tips, tubes, cryovials. Required. Not priced here.
• Cryopreservation. DMSO-based freezing medium, controlled-rate freezer time, liquid nitrogen. Required if the output is cryopreservable vials, which is what the catalogue comparator is.
• Per-batch characterisation beyond mycoplasma. Identity, purity against neuronal markers, post-thaw viability. A catalogue vial arrives with these.
• Protocol establishment. The first several batches of any new differentiation are learning, not production. The model treats batch one as if it works.
• Commercial-use licence. Research use of a purchased iPSC line is normally granted. The right to sell a product built on a differentiated derivative usually is not. Nobody publishes this fee. It is the largest unpriced number in the whole comparison.

Add these back and the in-house column rises. We have not added them because we will not invent a figure, but you should assume the true in-house cost is meaningfully above $419.76 per million.

The decision rule this model supports

1. Price the largest pack size first. If a supplier sells a five-million vial and you need five million cells, that is very likely the cheapest route available to you and no in-house model will beat it.
2. If you need single-million quantities repeatedly and your labour is inexpensive, and you already own the incubator, building is defensible and pays back fast.
3. If your labour is loaded at commercial rates, building is a staffing decision dressed as a cost decision. Do it because you need protocol control, a specific genotype, or a modification nobody sells — not because it saves money, because at $143/hour it does not.
4. If you need a genotype, reporter or timepoint nobody catalogues, the model is irrelevant. You are building because there is no alternative, and the right question is what the engineering step costs, not what the neurons cost.
5. If you intend to sell anything downstream, settle the licence question before you settle the cost question. It dominates.

Reproducing this model with your own numbers

The three inputs that move the answer most, in order:

1. Catalogue pack size. Moves the comparator by up to 46 percent.
2. Labour rate. Moves the in-house variable cost by 50 percent across our two anchors.
3. Failure rate. Moves the effective in-house cost by 50 percent between 0 and 33 percent.

Everything else — reagent brand, exact media volumes, facility charge — moves the answer by less than either of the first two. If you are going to spend an afternoon refining one number, refine your labour rate, not your media consumption.

How we can help

We built this model because clients kept asking us the question and we did not want to answer it with an opinion. If you send us your batch size, your cadence and your labour rate, we will run the same arithmetic against current quotes for your actual specification — including volume pricing that is not on any public page.

Sometimes the answer is that you should differentiate your own and not buy anything from anyone. We will tell you when that is the case; the model above is what we use to work it out.

Send us a specification.