According to an application, CITIC SEC released a research report stating that with the large-scale ramp-up of 800G optical modules and 1.6T entering its first year of volume growth, the power consumption per module and the demand for temperature control precision are rising significantly. Optical module thermal management is being upgraded from a "supporting link" to a "performance-determining link." Optical module thermal management has formed a three-tier system: chip-level precise temperature control (Micro-TEC plus aluminum nitride ceramic substrate and tungsten-copper/molybdenum-copper heat sinks), interface thermal conduction (TIM thermal gels and graphene pads), and module-level heat spreading (VC vapor chambers and cold plates). The thermal management value per module systematically rises with each speed iteration; combined with CPO/NPO architectures where the optical engine and switch chip share the same thermal domain, the industry is experiencing both volume and price growth. Regarding the competitive landscape, core segments such as TEC and ceramic substrates have long been dominated by Japanese manufacturers, while domestic Chinese manufacturers are accelerating breakthroughs, with ample room for import substitution. The report recommends focusing on three investment sub-themes: TEC and ceramic substrates, TIM and graphene thermal materials, and VC vapor chambers and liquid cooling structural components.

Main views of CITIC SEC are as follows

The "thermostat system" of optical modules: a three-layer heat dissipation chain that is closely interlinked. The core contradiction in optical module heat dissipation is not the absolute power consumption value, but rather that heat is too concentrated鈥攄evices such as DSPs, lasers, and TIAs are integrated into an extremely small space, resulting in extremely high local heat flux density. Moreover, the laser wavelength drifts approximately 0.1nm per degree Celsius with temperature changes, while the channel spacing in WDM systems is only 0.8nm. Uncontrolled temperature will directly lead to channel crosstalk, increased bit error rates, and device lifespan degradation. Optical module thermal management thus forms a three-tier system: (1) chip-level precise temperature control鈥擬icro-TEC semiconductor coolers provide laser temperature stability at the +/-0.1 degree Celsius level, supported by aluminum nitride ceramic substrates and tungsten-copper/molybdenum-copper heat sinks for thermal conduction and thermal expansion matching; (2) interface thermal conduction鈥擳IM thermal interface materials fill microscopic air gaps between chips and the housing or VC (air has extremely poor thermal conductivity), with 800G mainstream solutions using thermal gels with thermal conductivity of 8-15 W/m.K; (3) module-level heat spreading鈥擵C vapor chambers rapidly spread local hot spots. There are currently two growth logics for optical module liquid cooling.

Growth logic one: AI computing power drives simultaneous volume and price increases for optical modules, with the thermal management segment expanding in tandem. AI cluster construction is driving explosive demand for high-speed optical modules, and by 2027, 1.6T is expected to further rise to a scale of over 70 million units. Thermal management is a standard component for every high-speed optical module, and demand for materials such as TEC, ceramic substrates, TIM, and VC will expand in sync with module shipments. The industry scale has entered a high-speed expansion channel. At the 1.6T stage, with material specification upgrades and increased usage, the thermal management value per module will further rise, and the growth rate of the thermal management segment is expected to continue outperforming the optical module industry itself.

Growth logic two: High speeds and CPO/NPO architectures push up heat dissipation density and precision requirements, systematically raising value. In 1.6T modules, the heat flux density of DSPs, driver chips, and lasers increases significantly, with some solutions' power consumption entering the 20-30W range. The heat dissipation approach is shifting from "passive heat dissipation" to "active thermal management": internal TIM is being upgraded from 8-15 W/m.K gel to 15-20 W/m.K (some with diamond powder added), and is expected to further transition to graphene thermal pads. VC vapor chamber penetration is increasing notably, and customized solutions such as integrated vapor chambers (with housing and heat spreader designed as one piece) bring additional value premiums. TEC shows a "volume growth" logic while also seeing value increases. Under CPO/NPO architectures, the optical engine and switch ASIC share the same constrained thermal space, making high-end heat dissipation shift from "optional" to "mandatory": thermal pads or phase-change materials are needed between the optical engine and the shared heat sink; the laser side requires ceramic substrates, heat sinks, and TEC; and the system side adopts VC or liquid cooling cold plates鈥攖he technical barriers and per-unit value of the thermal management segment are rising simultaneously.

Overseas manufacturers' dominance is loosening, and domestic Chinese manufacturers have great potential. The TEC segment has long been dominated by Japan's Yamato Thermoelectric and KELK, with domestic localization rate below 5% in 2023. However, domestic Chinese manufacturers have now achieved batch supply of 400G/800G Micro-TEC, with 1.6T products entering small-batch validation and capacity ramping rapidly, raising the domestic localization rate to approximately 15%. Ceramic substrates and upstream AlN powder have long been monopolized by Japan's Tokuyama and Kyocera, but domestic substrate manufacturers have now entered the supply chains of leading optical module makers and become main suppliers, with some powder manufacturers also accelerating sample validation. In the TIM and graphene thermal pad fields, mainland Chinese manufacturers lead in technology, with mass-produced products achieving thermal conductivity of 130 W/m.K level, entering overseas major customer supply chains and advancing application introductions in 1.6T optical modules. Following the import substitution path of other optical communication segments, domestic manufacturers in the thermal management segment, leveraging cost, response speed, and capacity advantages, have high certainty and large elasticity for share gains.

Investment strategy: Optical module thermal management is a scarce link in the AI computing chain that combines "certain supporting role, rising value, and import substitution" three-fold logic. It is recommended to focus on three investment sub-themes: (1) TEC and ceramic substrates (core chip-level precise temperature control link, with the highest Japanese manufacturer share and greatest import substitution elasticity); (2) TIM and graphene thermal materials (fastest material upgrade iteration driven by 1.6T/CPO/NPO, with mainland Chinese manufacturers leading in technology); (3) VC vapor chambers and liquid cooling structural components (penetration rising with certainty as power consumption increases, extending to high-value solutions such as integrated vapor chambers and liquid cooling cages).

Risk factors: 1) AI computing power demand falling short of expectations; 2) 1.6T and CPO/NPO industrialization progress falling short of expectations; 3) Changes in heat dissipation technology routes (such as penetration of non-cooling solutions exceeding expectations); 4) Domestic manufacturers' customer certification and mass production progress falling short of expectations; 5) Intensified industry competition putting pressure on prices and gross margins.