Cooling & Thermal
Cooling and thermal management for high-density AI compute spans liquid cooling, immersion cooling and industrial thermal/HVAC equipment supplied by established players such as Johnson Controls, Trane Technologies, Vertiv and Modine, alongside materials suppliers such as Solstice Advanced Materials, with Nvidia acting as a central ecosystem driver through its GPU platform thermal requirements and cooling-hardware certification programs.
Active institutional exposure flipped from a net outflow of $6.5 billion in Q1 2026 to a net inflow of $13.3 billion in Q2 2026. 53.3% of companies in this sub-theme saw increased active positions in Q2 2026.
Overview
This sub-theme covers the equipment, materials and infrastructure used to remove heat from high-density AI compute racks, ranging from liquid cooling and immersion cooling to the broader HVAC and heat-exchanger base that underpins data-center thermal management. Johnson Controls and Trane Technologies compete directly in HVAC equipment, security, and fire-detection/suppression markets, while Vertiv operates as a narrower, offering-specific competitor to Johnson Controls with a global but more limited product set that also overlaps with Eaton. Modine supplies heat exchangers, cooling modules and thermal-management products to a broad industrial and HVAC customer base, including Carrier, Caterpillar, Deere and Trane itself, positioning it as a component-level supplier into the thermal value chain. Solstice Advanced Materials supplies refrigerants and thermal-management materials, competing with Chemours in its Thermal & Specialized Solutions segment, and retains licensing and services ties to Honeywell following its spin-off. Nvidia sits structurally at the center of this sub-theme not as a thermal-equipment vendor itself, but because its accelerator platforms define the power density and thermal envelope that liquid- and immersion-cooling vendors must design around, and it partners directly with Vertiv on power and thermal infrastructure aligned to its AI and high-performance-computing architectures.
Investment Thesis
Investment thesis details are part of the Pro plan.
Bottleneck Tracker
Bottleneck details are part of the Pro plan.
Key Companies
Related Events
Brookfield and its Radiant neocloud platform plan a UK AI campus at Drax Power Station that could exceed £20 billion and 1 GW of capacity. Drax shares rose about 8.8% as investors priced in monetization of its grid-connected site. Phase 1 (up to 100 MW from Q1 2028, £2.5 billion) still needs planning, regulatory and network approvals, and a tenant, and no final investment decision has been made. The project tests whether the UK can host large-scale AI compute despite high power costs and grid connection delays, which pushed Microsoft out of the Nscale deal in October. If delivered, it would add sovereign-style UK compute capacity and demand for power, cooling, chips and construction, though execution risk remains high.
SoftBank's reported $50 billion plan would be one of the largest single AI infrastructure commitments in Southeast Asia, dwarfing Google's $2 billion, AWS's $6.2 billion and Microsoft's $2.2 billion pledges in Malaysia. It builds on a framework agreement with Grab and Petroleum Sarawak Berhad to explore AI and digital infrastructure in Sarawak. The plan reinforces Malaysia's position as a regional data center hub, which has already drawn about $94 billion in related investment since 2021. Details such as timeline, structure and financing are not confirmed, and SoftBank has not commented, so the headline figure carries execution risk. Suppliers of data center power, cooling, networking and compute stand to benefit if the spending materializes, while it also adds to SoftBank's heavy AI capital commitments.
TikTok USDS Joint Venture is exploring multi-gigawatt compute procurement, with one source citing a four-gigawatt target by the early 2030s. The scale adds to the global scramble for AI infrastructure and could create sizable demand for data center capacity, power and chips. Oracle, which is both a managing investor and the primary cloud provider, remains the likely anchor beneficiary, though sourcing from other providers is possible. Localized retraining of the recommendation algorithm on U.S. data and a planned B2B AI product point to rising inference and training needs. Sources are anonymous and the spokesperson only affirmed Oracle's primary role, so the size and timing of any deals remain unconfirmed.
Nvidia is moving further up the stack by selling a pre-integrated, containerized building block rather than chips or single racks. Shrinking data center deployment from six months to 14 days could speed customer capex conversion into revenue for Nvidia and its manufacturing partners. Nvidia's content per gigawatt has risen from about $18 billion to more than $40 billion, and the Pod reinforces that trend. Server ODMs and OEMs such as Foxconn, Quanta, Pegatron, Lenovo, HPE and Gigabyte gain volume but also face a tighter, Nvidia-defined reference design that may limit differentiation. Pre-integrated liquid cooling and cabling shift some value from on-site integrators and traditional data center construction toward factory-built systems, and the pods are pitched to European customers facing deployment constraints.
A pluggable optic at 12.8 Tbps would sharply raise bandwidth density for AI cluster cabling, favoring optical module, DSP and laser suppliers who can meet 400G-per-lane and 100 W thermal demands. Nvidia's participation, despite its move to co-packaged optics on its switches, shows pluggables remain strategic for NICs and flexible deployments. The open standard widens the addressable market beyond Cisco and Nvidia but raises thermal, reliability and blast-radius challenges for module and cage makers. Microsoft's use of Cisco's P200 for scale-across networking strengthens Cisco's position against Nvidia Spectrum-XGS and Broadcom Jericho 4. Commercial impact is gradual, tied to 204.8 Tbps switches and Vera Rubin Ultra-class racks later in the decade.