Objection one: “Both quotes say inhibited glycol. Same thing, then.”

The word “inhibited” is where the comparison stops and where the difference begins.

An inhibitor package is a set of chemicals chosen for a set of metals. A package built around carbon steel protects carbon steel. Put that same package into a loop containing aluminium plates and the aluminium is not being protected — it is being left alone while the steel next to it is looked after.

Two quotes can both say “inhibited glycol” and describe fluids with different metal coverage. The difference is invisible on the line item and visible in year four.

Ask for the metal list. Not “suitable for common metals”. The actual list: copper, brass, stainless, and aluminium, named separately. If a supplier cannot name them, the fluid may still be fine, but the claim is unfalsifiable, and an unfalsifiable claim is not a basis for comparing two prices.

Objection two: “Ours tests well below the conductivity line. That covers it.”

It does not, and the reason is about sequence rather than magnitude.

The industry commonly references around 1500 μS/cm as a service red line for coolant conductivity, drawn from the clearance and creepage logic in equipment safety standards. A fluid that starts well below that line genuinely has a safety layer.

But conductivity is the last layer, not the first. By the time a rising number matters, liquid is already somewhere it should not be. The layer that prevents the event sits further upstream — corrosion control that keeps the plate wall from thinning, and seal compatibility that keeps the fitting from opening.

A low conductivity figure on a quote tells you what the fluid does after something has already gone wrong. It says nothing about the two defences in front of it.

Objection three: “Aluminium is a small part of the loop.”

In surface area, often true. In consequence, rarely.

Aluminium corrodes in the opposite direction from steel. An inhibitor tuned for one does not automatically suit the other. And aluminium tends to appear in exactly the components that are hardest to inspect — brazed or skived plates inside a rack, where the wall thickness is small by design because thin walls move heat.

A loop that is 90% steel by area can still fail at the 10% that is aluminium, because the failure mode is a hole in a plate and a hole anywhere counts the same.

Objection four: “Seals are the mechanical contractor’s problem.”

The seals are specified by the mechanical contractor. Whether the fluid degrades them is a fluid question.

A fluid that protects metal beautifully and swells or hardens EPDM, FFKM or PTFE has not removed a leak path. It has moved it. A moved leak is still a leak, and it usually presents at a fitting rather than at a plate, which is why it gets attributed to installation rather than chemistry.

This one is cheap to close. Name the seal materials in the enquiry and ask each supplier to confirm compatibility against them. It costs one line and it closes the second leak point at the same time the first one is closed.

Objection five: “This grade costs more per litre.”

The honest answer begins by agreeing. Yes, it usually does.

Then the comparison changes units. A secondary-side leak in a cold-plate loop is not a maintenance ticket. It is an event involving live hardware: a cabinet down, a board possibly lost, an investigation, and a replacement fill on top. The fluid price difference between two quotes is a rounding error against any of that.

Per litre is the wrong unit because the two numbers do not predict the same outcome. Total cost of ownership is the right unit, and on that axis the gap between a corrosion-engineered grade and a generic one usually inverts.

The five questions that end the argument

Send these to every supplier on the list. The answers make the quotes comparable, which is the thing price alone cannot do.

A quote that answers all five is comparable. One that answers none is not cheaper — it is just priced against a different, thinner question.

What a real quote should contain

None of these is a premium feature. They are what a secondary-side quote looks like when it has been written for a cold-plate loop rather than copied from a chilled-water schedule.

The arithmetic, in one table

The first row is the one procurement sees. The fourth row is the one that decides the total.

To run this properly you need three numbers you already have: the cabinet count, the fluid charge per loop, and what one hour of compute downtime costs your operation. Multiply the fluid price difference across the whole charge and you get a figure that looks large in isolation. Set it against a single leak event and it usually stops looking large.

Most buyers never do this multiplication. Nobody asks them to, and the three numbers live in three different departments.

There is a second version of the same arithmetic that reviewers forget. Sampling costs the same in both columns. The trend review costs the same. The only line that actually moves is the fluid, and it moves against a risk measured in cabinets rather than in litres.

What we send a buyer

Send us the wetted-material list and the loop volume. You get back a secondary-side grade matched to copper and aluminium together, a quote with the corrosion defence spelled out rather than summarised, and a baseline sample plan.

Where a question belongs to the cold-plate maker or the CDU maker, we say so on the quote. A supplier who claims everything is claiming something they cannot support.

FAQ

How do I choose a secondary coolant for a cold plate loop?

Check the corrosion defence: which metals the inhibitor protects together, seal compatibility, pH buffer, impurity control and conductivity redundancy — in that order, before price.

What should I ask a coolant supplier about corrosion?

Ask which metals the package protects, whether it was verified against your seal materials, and how pH and impurities are held stable. A one-line “inhibited” claim is not an answer.

Is low conductivity enough to keep a data centre safe?

No. Conductivity is the last layer. The plate has to stay closed through corrosion control and seal compatibility first; low conductivity is the redundancy behind those.

Why does aluminium need a specific grade?

Because aluminium can corrode the opposite way from steel, and an inhibitor tuned for steel does not automatically suit it. The grade should be selected for aluminium where plates include it.

Are seals really a leak risk in the secondary loop?

They are the second-most-common leak point. A fluid that swells or hardens EPDM, FFKM or PTFE seals trades one failure for another, which is why seal verification belongs in the purchase.

Does a corrosion-engineered grade cost much more?

Usually far less than the cost of one cold-plate leak and the outage it causes. Frame it as total cost of ownership, not price per litre, and the gap shrinks to nothing.

How often should the secondary loop be sampled?

Follow published equipment guidance: about every six months on high-density AI loads, about annually on general service, and after every top-up or repair. Baseline at commissioning.

Can one supplier cover both the secondary and primary sides?

Yes, through duty-matched grades. The point is matching grades to jobs — secondary-side corrosion control and primary-side freeze margin are different specifications, and one drum store rarely serves both well.