Engineered thermal substrates

Heat, not silicon,
now sets the limit.

Xtherma makes a passive thermal layer that moves heat away from the chip faster than any metal, inside the package and cooler you already have. An OEM measured it on their own production hardware: 11 °C cooler at the junction, with nothing else changed.

Or write to contact@xthermacore.com

Cross-section WithoutWith Xtherma

Schematic, not to scale. The measured results are in the proof section below.

More usable performance
from the silicon you already paid for.
No redesign
Same package, same cooler, same silicon. The layer goes where the paste or spreader already is.
Measured, not modelled
by an OEM, on its own production hardware, with its own instruments.

The problem

Compute density is outrunning thermal architecture.

Chips are rated for more than their enclosures can cool. The result is silent: derating, throttling, and interfaces that degrade under load. The bottleneck sits in the first few hundred microns above the die.

01

Rated for more than the box can cool

Inside sealed enclosures, modules routinely run well below their rating because the heat has nowhere to go. The silicon is paid for. The performance is not used.

02

The interface gives way first

Thermal paste dries out and pumps out under sustained heat. Its resistance climbs with load, exactly when it matters most.

03

Cooling arrives too late

Fans, pumps, cold plates and chillers act after heat has already concentrated. The cheapest degree is the one removed at the junction, passively, before any of them.

The layer

Move the fix to where the heat starts.

Not a heat sink. Not a paste. A passive, engineered substrate between the die and the spreader, or built into the lid, that spreads heat before the rest of the cooling chain sees it. Same package, same cooler, no silicon redesign.

>1,000 W/mK
Moves heat about 2.5 times faster than copper, at under half the weight.
Passive
No fluids, no pumps, no power, no moving parts. Nothing to fail.
Stable under load
Unlike paste, its resistance barely moves as the heat rises. No pump-out, no dry-out.

Formats, cross-section and full engineering detail on the Technology page

Proof

Measured on production hardware, by the OEM.

An industrial OEM's hardware team put an Xtherma pad under the processor of a production module in a sealed, fanless enclosure. Nothing else changed. Their engineers ran it, on their instruments, and wrote the report.

Measured result
11°C cooler

at the junction at the highest load tested. The gain grew as the load rose, and the rest of the cooling chain was shown to be untouched.

Junction temperature rise above room

Kelvin, steady state, same hardware, same cooler

Light load-5.2 K
Medium load-10.4 K
High load-11.3 K
Paste referenceXtherma pad

The thermal path also became more predictable. With Xtherma, junction temperature tracked power in a straight line; with paste, it drifted upward as the interface heated, the signature of a bond line drying out. That predictability is what a reliability engineer designs around.

Resistance breakdown, linearity, thermal margin and a four-material comparison

Value

Cooler silicon is worth more than comfort.

Every degree removed at the junction becomes something you can sell or stop paying for. Which one depends on your system, so we quantify it on your hardware, in your units.

Headroom

More from the same part

Run the silicon you already bought at a higher sustained power, or hold the rating with margin to spare.

Reliability

Fewer trips and derates

A predictable thermal path means fewer throttling events, and guardbands you can actually use.

Cooling

Less hardware downstream

Heat spread early is cheaper to remove: smaller fans, plainer heat sinks, lower flow, less energy.

Lifetime

An interface that lasts

No paste to dry out or pump out, so the thermal path you qualify is the one you ship for years.

Every evaluation on your hardware ends with a value estimate for your module: watts of headroom, amps, kilowatts per rack, and what that is worth to you. Built on your numbers, not ours.

Quantify it on my hardware

Integration

From sample to qualified supply, without a redesign.

The layer goes where the paste or the spreader already is. The cooler stays. The silicon stays. The path from first sample to qualified supply is built around your qualification clock, not ours.

1

Evaluation sample

Standard formats with mounting guidance, so you can run the first comparison in your own rig within days.

2

Evaluation on your hardware

A scoped programme: reference build versus Xtherma build, your load profile, your sensors, protocol agreed up front. You keep the report.

3

Design-in programme

Lid-integrated or architected formats developed with your packaging team, qualification legs agreed with your reliability engineers.

4

Supply

Small-batch supply for qualified modules first, framework agreements as volumes grow.

AI and HPC accelerators Edge AI and embedded modules Power electronics and magnetics Aerospace, defence and space Lasers, RF and photonics

Proven so far on embedded modules and power magnetics. Accelerator-class packages are in design-in evaluation now.

Request an evaluation Tell us the package, the power and the hotspot area. You will hear back from the founder.