Buying SU-8 masters and PDMS chip fabrication: the two numbers in every quote, and the contract term that doubles your second order

Every custom PDMS quote decomposes into two numbers: non-recurring engineering, paid once, and per-part cost, paid every time. Buyers routinely optimise the wrong one. At a quantity of one the NRE is about 95 percent of the bill and the per-part price is nearly irrelevant. At a thousand the NRE has effectively vanished and the per-part price is the whole conversation. Negotiating hard on a per-chip figure for a ten-chip order is wasted effort; negotiating hard on setup for a thousand-chip order is the same mistake in reverse.

What makes this category unusual, and worth a buying guide, is that a small number of fabricators publish rate cards. That means the arithmetic is checkable before anyone opens a sales conversation, which is not true of almost anything else in custom biology hardware.

The published rate cards

All figures captured 2026-09-01 from the vendor pages cited. These are published list rates for each fabricator’s stated standard process, not quotations for your part.

Penn QNF’s rates are academic internal rates. Check eligibility before building a budget on them, this is the same trap that runs through institutional core pricing across the sector, where the cheapest published number is frequently one an external buyer cannot access at all.

Soft lithography does not scale, and the numbers prove it

Run HiComp’s published structure, $1,500 NRE plus $80 per chip, across four quantities.

MODELLED, this is arithmetic on published figures, not a quotation.

The shape of that curve is the single most important thing to understand before commissioning custom PDMS. Per-chip cost falls 94 percent between one and a hundred parts, and then essentially stops, from $95.00 to $81.50 between 100 and 1,000, a further 14 percent across a tenfold increase in volume.

The reason is structural rather than commercial: every PDMS chip is an individual casting operation. Pour, degas, cure, demould, punch, bond. There is no economy of scale in a process where the marginal unit costs the same amount of hands-on time as the first one. This is why injection moulding exists, and why Protolabs’ aluminium tooling from $1,495 becomes the interesting comparison somewhere in the high hundreds of parts, you are buying a process that actually amortises, at the cost of a tooling commitment and a material change.

uFluidix’s published example of $60 per chip at 100 chips sits below HiComp’s $95.00 effective figure at the same quantity, but the two are not directly comparable: uFluidix does not separately state NRE, so the $60 may or may not include it. When you compare two fabricators, convert both to a total for your actual quantity before concluding anything. A quote with a low per-part price and an unstated setup fee is not a low quote.

The master is cheap; casting from it is not

A second structural fact hides inside the Penn QNF rate card, and it is the one that explains why academic cleanroom fabrication frequently costs more than a commercial quote despite the lower headline rate.

The published master price is $1,060 for single-layer. The published staff support rate is $55 per hour. What the master price excludes is labour, your labour or theirs. At an hour and a half of hands-on casting time per batch, that is $82.50 of staff time per batch at the published rate, which dominates mould amortisation at every quantity above about ten parts. The mould is a one-time $1,060. The casting is forever.

This is why the correct question to a cleanroom is not “what does a master cost” but “what does a finished, bonded, inspected chip cost me, including the time I have to spend in your facility.” The answers diverge by more than the mould price.

What actually drives a quote up

Three variables, all visible in the published data:

Layer count. Penn QNF prices a single-layer master at $1,060 and a two-layer at $3,295. That is a 3.1× multiplier for one alignment step. Alignment is the expensive operation in photolithography, and every additional layer is another alignment. If your design has two layers because the geometry genuinely requires it, that is the price. If it has two layers because the first draft was drawn that way, redraw it.

Channel height. Blue BioDevice tiers on height alone: $790 at 5 µm and above, rising through $890 and $990 to $1,150 at 300 µm and above. Tall SU-8 features need thicker resist, longer exposure and more careful development, and the price tracks that. A design that specifies 300 µm because it seemed like a round number, when 150 µm would work, costs $160 more per master and carries a harder process.

Aspect ratio. Both Penn QNF and Blue BioDevice cap aspect ratio at roughly 1:4 width to height as standard. Above that you are outside the standard process, which means quote-only, longer lead times and a higher failure rate on the master. eNUVIO advertises high-aspect-ratio moulding as a differentiating capability precisely because it is hard, it is quote-only there too.

And the one that is not a fabrication variable at all: iteration count. Every design revision means a new master at $790 to $3,295. Three revisions on a two-layer design at Penn QNF rates is $9,885 in masters before you have a working device. Iteration is the largest hidden cost in any custom fabrication programme, and the way to control it is to spend more time on the first design review and less on the second and third masters.

The RFQ line almost nobody includes

Here it is, and it is worth more than any price negotiation in this category:

Is the master mould delivered to the buyer, or retained by the fabricator?

That is a contractual term, not a price, and it determines whether your second order is a per-part order or a full NRE again. Two quotes that look nearly identical at n=100 diverge completely on the repeat order if one fabricator sells you the tooling and the other keeps it.

Blue BioDevice sells the master to the customer, so the buyer owns the tooling. That is stated on their published page. For every other fabricator, ask, and get the answer in writing alongside a second question: where the master is retained, what does a repeat order cost? A fabricator who will state a repeat-order price is giving you the amortisation you paid for. One who will not is reserving the right to re-quote from scratch.

The comparator every custom quote must beat

Before commissioning anything, price the finished catalogue device that already exists. eNUVIO publishes live public prices for its OMEGA range, captured 2026-09-01:

At $35.63 to $55.00 per experiment, sterile, in stock, with a published specification, that is the bar. A custom programme at HiComp’s structure costs $95.00 per chip at a hundred parts, and the chip is not yet bonded, surface-treated, sterilised or inspected. The custom route has to buy something the catalogue part cannot deliver, a compartment that does not exist, a footprint that fits an instrument you already own, a force geometry nobody makes. “It would be nicer” does not clear that bar.

What the rate cards do not cover

Every published figure above is for a master or for a bare cast part. A usable culture device also needs bonding, surface treatment, sterilisation and inspection, and no fabricator surveyed publishes a price for any of them. We do not estimate them. What we can say about each:

• Bonding. To glass, to a second PDMS layer, or supplied unbonded. Quote-only everywhere.
• Surface treatment. PDMS is hydrophobic and reverts toward hydrophobicity after plasma treatment, so the shelf life of the treatment matters as much as the treatment itself. If the part ships treated, ask how long the treatment lasts and how it is stored.
• Sterilisation. The published PDMS fabrication protocol work by Masterton and colleagues states that autoclaving, dry heat and gamma are effective but can alter PDMS mechanical properties, and prefers ethanol immersion, UV irradiation and oxygen plasma as methods that do not affect mechanical integrity. The same work gives Sylgard 184 modulus as roughly 1 to 4 MPa depending on cure temperature between 25 °C and 200 °C, a four-fold range set by a process parameter, which is worth knowing if your device’s stiffness is part of the biology.
• Inspection and documentation. Dimensional inspection against the agreed drawing, and first-article inspection before batch release, are the two lines that make a batch defensible.

The full quote anatomy, line by line, is on what a real PDMS quote includes.

The material caveat that should be on the front page

PDMS is the default for research culture devices and the wrong default for anything that touches hydrophobic small molecules. Kemas and colleagues found that more than 99 percent of chlorpromazine partitioned into the PDMS bulk within 24 hours, with absorption differing more than 1000-fold across eight polymers and PMMA and thiol-ene epoxy roughly 1500-fold lower.

If your device will be dosed with lipophilic compounds, the material decision comes before the fabricator decision, and it may well take you out of soft lithography entirely. That is a cheaper realisation now than after the masters are cut.

What to put in the RFQ

1. File format, ideally GDS. DXF-to-GDS conversion is billable at the academic cleanroom rate.
2. Minimum feature width, and the tolerance genuinely needed rather than preferred.
3. Channel height and tolerance, since height drives the master price tier directly.
4. Number of layers, and whether inter-layer alignment is required. One alignment step is a 3.1× multiplier at published academic rates.
5. Aspect ratio, against a standard cap of roughly 1:4.
6. Material: PDMS, thermoplastic or glass. This changes the method, not just the price.
7. Substrate and bonding requirement: bonded to glass, to another PDMS layer, or unbonded.
8. Surface treatment or coating requirement, and its stated shelf life.
9. Sterilisation method the finished device must survive.
10. Quantity now and realistic quantity within twelve months, quoted at both.
11. Whether the master is delivered to the buyer or retained by the fabricator, and if retained, the price of a repeat order.

How we can help

We put the same specification to fabricators who publish rate cards and to those who do not, and we convert every response to a total at your actual quantity so the comparison means something. We also make sure line 11 is in the document, because it is the one that decides what your second order costs and it is the one that gets left out.

The paired datasheet, with the full specification field list and supplier evidence notes, is at SU-8 masters and priced PDMS chip fabrication. Send us the drawing.