Microfluidic fabricators compared: method, feature size, tolerance, MOQ and NRE

The short answer: two fabricators publish a rate card, and the decision is usually made by method rather than by vendor. HiComp publishes $1,500 of one-time engineering plus $80 per chip for PDMS soft lithography. Blue BioDevice publishes SU-8 master moulds from $790 to $1,150 by channel height. Parallel Fluidics publishes no price but publishes the most complete design rules and tolerances in the market, in production thermoplastics rather than PDMS. Everyone else is quote-only.

The second finding, and the one that saves the most money: for cell culture applications, an off-the-shelf device usually wins. eNUVIO’s OMEGA-MP is $440 for four devices covering eight experiments — $55 per experiment with no tooling cost at all. Against $1,580 for one custom part, that is 28 experiments before a custom route has produced anything.

The fabricator comparison

Method decides more than vendor

The single most consequential choice on this page is not which company you use. It is which fabrication method your device needs, because the cost structure differs by orders of magnitude and the crossover points are predictable.

Soft lithography in PDMS: cheap to start, never cheap at volume

Taking HiComp’s published $1,500 NRE plus $80 per chip and amortising it:

Per-part cost falls by 95 percent between one and a hundred parts, then essentially stops falling. Between 100 and 1,000 the per-chip cost moves from $95 to $81.50 — a 14 percent improvement for a ten-fold increase in commitment. Soft lithography does not scale, because every chip is an individual casting operation. There is no volume at which it becomes cheap.

Note also that HiComp’s $80 rate is explicitly stated for initial batches of 1 to 20 units with volume discounts available as orders scale, so the table above is a conservative reading of their own structure at the high end.

The crossover between fabricators is real and worth discovering. uFluidix publishes 100 chips at $60 each — $6,000 with tooling apparently included — against $9,500 on HiComp’s published structure at the same quantity. That is a 37 percent spread at n = 100 and it costs nothing to find out.

Buying a master and casting yourself: the labour trap

Master and per-chip rates below are the published rate cards captured 2026-09-01.

This looks unbeatable and it is not, because the excluded costs are most of the true per-part cost. A PDMS casting cycle — weighing, mixing, degassing, pouring, curing, peeling, punching, cleaning, plasma treating, aligning, bonding — is several hours elapsed with roughly one to two hours hands-on per batch. At the $55 per hour academic staff rate published by the same facility that sells the $1,060 master, ninety minutes of hands-on time is $82.50 per batch, which dominates the mould amortisation at every tier above ten parts.

The honest reading: buying a master is cheap, casting from it is not, and the cost is labour rather than materials. If your technician’s time has any real cost, the outsourced $80-per-chip route is competitive far earlier than the mould price suggests.

Yield makes this worse and nobody publishes a figure. Bonding failures, leaks at inlet ports and alignment errors are the common modes. We could not find a published failure rate from any fabricator in this survey, so we do not assert one — but a 20 percent loss rate raises effective per-part cost by 25 percent, and that alone can move the buy-versus-make answer.

Thermoplastic moulding: the PDMS problem, solved at a different price

Parallel Fluidics publishes no price, and publishes the most useful technical document in this comparison. Their design guide gives hard numbers where everyone else gives adjectives:

• Minimum moulded microfeature: 50 µm width, 50 µm height. Height must not exceed 50 percent of the fluidic layer thickness.
• Aspect ratio: 3:1 maximum height to width, 1:100 minimum when capped, with a stated preference for staying close to 1:1 for consistent results.
• Corner radius: 125 µm minimum, with a 10:1 maximum height-to-radius ratio.
• Feature separation: 300 µm minimum at the narrowest point; 1.5 mm minimum from the outer perimeter on a capped device.
• Tolerances: ±20 µm or ±10 percent on microfeatures, whichever is greater; ±125 µm on macrofeatures and machined features; 100 µm flatness on the cap surface.
• Stated three-day turnaround on custom device iterations, and a 24-hour quote from an uploaded 3D CAD model.

They also publish a solvent compatibility and physical property table across COC, COP, PMMA and PC that is genuinely useful regardless of who you buy from. The material findings that matter for culture work: PMMA and PC are incompatible with acetone, DMSO and polar organic solvents; COC and COP are resistant to all three, and no listed material resists chlorinated solvents such as dichloromethane. COC and COP have moisture absorption of 0.01 percent against 0.2 to 0.3 percent for PMMA and PC, and COP is described as having low surface adsorption for the most demanding applications.

That last point is the substantive reason to consider thermoplastic over PDMS for anything involving small molecules. Toepke and Beebe documented PDMS absorption of small molecules and the consequences in microfluidic applications, and the effect silently changes effective dose. If you are dosing a compound and reading a response, the wetted material is a data-quality question, not a procurement preference.

Laser, CNC and hot embossing: prototype-flexible, no published pricing

Potomac Photonics structures the work in three explicit phases and the reasoning is sound. During process development they favour direct machining and laser methods over mould-based fabrication precisely because design parameters change frequently and direct methods allow rapid manipulation of feature dimensions, including fabricating partial chip designs to screen difficult features such as microchannels and pillars quickly. Scale-up then moves to hot embossing or injection moulding, with the accompanying post-replication operations of bonding, surface treatment and interfacing.

That phasing is the right mental model for any programme, whoever fabricates. Match the method to the maturity of the design, not to the eventual volume.

We found no published per-part pricing for laser ablation, micro-CNC, hot embossing or 3D-printed microfluidics from any named fabricator in this survey. That is the state of the market, not an omission here.

What actually drives a custom quote up

From the published design rules and constraints of the fabricators above:

• Layer count. Penn QNF prices a single-layer master at $1,060 and a two-layer master at $3,295 — a 3.1× multiplier for one extra layer. Alignment is a step change, not an increment. HiComp’s quote form bins designs by single height, two heights and more than two, and flags multi-layer as outside the standard package.
• Channel height. Blue BioDevice tiers on height alone: $790 at 5 µm and above rising to $1,150 at 300 µm and above, a 46 percent increase across the range.
• Feature size and tolerance. Penn QNF’s standard process covers widths above 20 µm within ±10 percent and heights to 100 µm within ±20 percent; anything tighter is explicitly subject to additional review and fee. Parallel Fluidics guarantees ±20 µm or ±10 percent as standard.
• Aspect ratio. Both soft lithography fabricators cap around 1:4 width to height as standard, and Parallel Fluidics caps moulded features at 3:1. Tall narrow structures break and delaminate. eNUVIO’s patented high-aspect-ratio moulding method is the specific exception in this comparison, with a shipping product carrying a 0.5 mm wide feature over 12 mm tall.
• Bonding complexity. HiComp’s quote form separates PDMS-to-PDMS, PDMS-to-thin-film, PDMS-to-glass, three-or-more-layer bonding and PDMS-to-other-substrate, and separates alignment accuracy into no alignment, ≥0.2 mm and <0.2 mm. Each step up is a cost step.
• Design maturity. Penn QNF bills design assistance and any consultation beyond the first 30-minute meeting at $55 per hour, charges the same rate for DXF-to-GDS conversion, and states that changes after fabrication begins incur charges for processing already executed. Blue BioDevice charges a technical fee of zero when a simple design arrives as GDSII and up to roughly $220 when drawing work is required. Arriving with a finished GDS file is worth real money at every fabricator here.
• Iteration count. The largest hidden cost in every custom programme. Each design revision in a soft lithography workflow means a new master. At $790 to $3,295 per master, five iterations is $4,000 to $16,500 before a single usable device exists. This is precisely the problem Parallel Fluidics’ three-day turnaround and single-click ramp is positioned against.

The off-the-shelf comparison, which usually wins for culture work

Before commissioning anything, price the catalogue. eNUVIO publishes live prices for its PDMS culture device range:

$35 to $141 per experiment with zero tooling. Against a custom route at $1,580 for the first part, an off-the-shelf device buys 11 to 44 experiments before you have spent what one custom part costs.

This is the price ceiling every custom quote must justify itself against. The only good reasons to go custom are geometry the catalogue does not offer, a feature size the catalogue does not hold, or a volume high enough that per-part cost dominates. “We would prefer our own design” is not one of them at n = 10.

Choosing

The RFQ line nobody includes

Specify all of the following and the quotes become comparable. The last one is the most commonly omitted and the most expensive to get wrong.

• File format, ideally GDS. DXF conversion is billable at multiple fabricators here.
• Minimum feature width and the tolerance you actually need, distinguished from the one you would like.
• Channel height and tolerance — height drives the master price tier directly at Blue BioDevice.
• Number of layers, and whether alignment between layers is required, with the accuracy needed.
• Material: PDMS, COC, COP, PMMA, PC or glass. This changes the method, not just the price.
• Substrate and bonding requirement — bonded to glass, to another layer, or supplied unbonded.
• Surface treatment or coating, if any.
• Sterilisation method the device must survive.
• Quantity now, and realistic quantity within twelve months. Quote both.
• Whether the master mould or tooling is delivered to you or retained by the fabricator.

That last point deserves emphasis. Blue BioDevice and Penn QNF sell you the master, so you own the tooling. HiComp states the photomask is yours to keep or store with them. Others quote a chip price with tooling retained. Two quotes that look similar at n = 100 can diverge completely at your second order, and the difference is a contractual term rather than a price.

Where to go next

What custom microfluidic fabrication costs has the full volume-tier arithmetic and the method comparison table. PDMS versus COP versus PMMA covers the material choice in depth, including the absorption problem. Custom microfluidic device fabrication covers the service-side scoping questions.

How we can help

Microfluidic quotes are hard to compare because fabricators bundle non-recurring engineering differently, some sell you the tooling and some keep it, and almost none publish per-part pricing at volume.

We put one specification to several fabricators, normalise the responses into the NRE-plus-per-part structure used above, and check who retains the master. Where an off-the-shelf device would do the job — and for culture applications it frequently would — we will tell you that rather than run a custom quote cycle.