Theses›Datacentre & AI compute›CNT thermal interface

CNT thermal interface

Carbon nanotube thermal interface materials reach ≥$1B in annual revenue by 2030, up from <$50M in 2024, driven by AI accelerator and HPC qualification wins.

  • Bear
  • Early
  • $18B by 2035

The callRestraints win: carbon nanotube thermal layers have been 5 years away since 2010 because nobody has published the aging data an accelerator maker needs.

Disclosure: TrimTabs, a company in the carbon nanotube market, pays me to write a newsletter for it.

D1D2R1R2R3C1foragainst D1D2R1R2R3C1foragainst
Fig. 1The balance. Drivers and catalysts against restraints, each block sized by its weight; catalysts hang dashed because they have not happened yet, sized by their odds. It tips toward the restraints.
  1. D1Accelerator packages are past 1kW with hot dies next to heat-sensitive HBM, and greases and silicone pads built for 300W CPUs pump out and dry out at that load.
  2. D2Aligned nanotube arrays conduct roughly 5-10x a premium silicone pad and stay compliant as the package warps.
  3. R1No CNT product has published 5,000-hour GR-468 aging data, which is why the material has been 5 years away since 2010.
  4. R2Henkel and the Japanese chemical majors hold the qualifications and the relationships, and the slice open to newcomers by 2030 is near $120M.
  5. R3Graphene, diamond, liquid metal and composite hybrids may take the high-end accelerator slots first.
  6. C1NVIDIA or AMD names a CNT die-level thermal layer in a shipping accelerator with multi-thousand-hour aging data, the design slot that proves the material.end-2027
Rests onthe beliefs underneath
Supported by · confidence
What would prove itdated tests, open and resolved
20272028202920302031today12345predictionswatch 20272028202920302031today12345predictionswatch
Fig. 2Each ring is a dated test: dashed while open, struck through when overdue, ticked or crossed once resolved.
  1. 1At least 2 CNT TIM products reach Telcordia GR-468-qualified production by 2027.Dec 2027 · open
  2. 2CNT TIMs win at least one Tier-1 hyperscaler AI accelerator design slot by 2027.Dec 2027 · open
  3. 3Two CNT products in GR-468 or AEC-Q200 qualified production, and a tier-1 accelerator design slot.Dec 2027 · open
  4. 4Carbon nanotube TIM revenue exceeds $1B globally by 2030.Dec 2030 · open
  5. 5CNT thermal revenue above $1B; below it the material stayed a niche.Dec 2030 · open
Expressionslisted, private, technologies
  • 6501.THitachi
    • full
  • HEN3.DEHenkel (Loctite, Bergquist)
    • expensive
     Sells silicone and silicone-free gap fillers, phase-change materials and thermal pads to electronics, server and automotive makers.
  • DOWDow
    • not screened
     Sells DOWSIL silicone thermal greases, gels and gap fillers to electronics and automotive makers.
  • 4063.TShin-Etsu Chemical
    • full
     Sells silicone thermal greases, pads and phase-change sheets to CPU, GPU and electronics makers.
  • PHParker Hannifin (Chomerics)
    • not screened
     Sells THERM-A-GAP gap fillers, thermal pads and dispensable compounds to electronics and telecom makers.
  • DDDuPont (Laird Performance Materials)
    • not screened
     Sells Laird thermal pads, gap fillers and phase-change materials alongside EMI shielding to electronics makers.
  • MMM3M
    • not screened
     Sells thermally conductive tapes, pads and gap fillers to electronics and battery makers.
  • 4004.TResonac (ex-Showa Denko and Hitachi Chemical)
    • expensive
     Sells thermal and packaging materials, including die-bonding films, to semiconductor and power-module makers.
Private · 8 private companies, 4 of them in the US.

TrimTabs · Carbice · NAWAH · Boston Materials · Hitachi Chemical · JX Nippon Mining · NovoLINC · Arieca

8 more private companies tracked, not shown

Technologies
Who has already bet on itprivate rounds, from public sources

$20M raised by 4 companies, $14M of it in the last three years. Early: Little money in so far; a lead investor would set the pace.

Backed by a strategic investor: NovoLINC.

What changedthe record, newest first
  • 31 Jul 2026The carbon-nanotube thermal-interface story does not currently have an analyst forecast behind its billion-dollar headline — the whole next-generation die-level TIM category is projected in the low hundreds of millions by the early 2030s, so the CNT bet is a bet on beating the analyst curve, not riding it.
  • 31 Jul 2026A foundry packaging voice added a new hurdle for vertically aligned CNT interfaces — great through-plane conductivity but degraded lateral spreading, just as leading packages go lidless and lose their heat spreader — while confirming from a fourth independent seat that serviceability, not raw performance, is what data-centre TIM buyers pay for.
  • 3 Jul 2026The thermal-interface layer is being attacked from two sides at once — in-die microfluidic cooling is maturing in real deployments, and new in-stack diamond spreaders are appearing — but neither removes the die-level interface itself, which is why premium TIM demand looks durable through the decade.
1 earlier
  • 3 Jul 2026A competing materials vendor conceded carbon-nanotube interfaces win on lifetime, reworkability and vacuum survival while arguing they lose on raw conductivity, which is the reliability-led premium story CNT vendors should be selling anyway.
Reading
Show the reading