Product1 publisher3 min readPublished
Molex buys into CAEPlus on a retrofit thesis, and discloses almost no numbers
The investment targets active liquid cooling meant to fit existing halls rather than new builds. Missing: any figure for heat removed, any price, any ship date.
The Product Desk · Product desk
Drafted by a language model from the sources cited here and checked against its claim ledger before publication. How we use AISend a correction

What happened
- Molex, an electronics and connectivity company, is investing in CAEPlus, a thermal management startup, to advance an active liquid cooling platform designed to handle rising heat from next-generation AI and high-performance computing systems.
- The funding will help CAEPlus develop BoundaryCool, an active liquid cooling system.
- BoundaryCool pulls heat away from GPUs, CPUs and TPUs using active liquid cooling; unlike passive cold plates it actively moves heat away from the chips while being designed to work with existing data center infrastructure.
- BoundaryCool takes a different approach from conventional passive cold plates by using an active cooling architecture to enhance heat removal at the chip level.
- The technology is being developed for data centers where increasingly powerful processors are pushing thermal management systems to their limits.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
Molex has taken a strategic investment in CAEPlus, a thermal management startup, to fund an active liquid cooling platform called BoundaryCool aimed at AI and high-performance computing hardware [1][2]. The claim that matters commercially is not the cooling physics but the installed-base pitch: Molex says the platform is being designed to stay compatible with older data center architectures, so operators could upgrade cooling without rebuilding entire facilities [7].
That is a bet on retrofit rather than new construction. A cold plate loop that assumes a purpose-built hall sells to whoever is pouring concrete; a system that drops into what already exists sells to everyone else, including sites whose useful life otherwise ends when rack density outruns the plant. Molex frames BoundaryCool as being for data centers where increasingly powerful processors are pushing thermal management to its limits [5], and the platform is described as pulling heat from GPUs, CPUs and TPUs [3].
The technical differentiation, as stated, is against passive cold plates. Molex says BoundaryCool uses an active architecture to improve heat removal at the chip level rather than relying on a passive plate [4], and that it is being developed to deliver significantly higher heat-transfer performance and greater processor temperature reductions than passive cold plate solutions [6]. Note the tense throughout: this is a platform being developed and validated, with the funding earmarked for exactly that work [8]. The source material contains no watts, no delta-T, no coolant temperature, no rack-level figure, and no third-party test result. There is also no explanation of what makes the architecture active beyond the assertion that it moves heat away from the chips [3]. Financial terms were not disclosed [11].
The second half of the deal is the part an operator should read carefully. Molex retains exclusive licensing rights to apply CAEPlus technology to its own portfolio of pluggable I/O solutions [10]. Molex sells connectors and copper interconnect; the executive quoted on the deal, Jairo Guerrero, runs its Copper Solutions business [9]. Optical and high-speed pluggable modules are their own thermal problem inside a dense rack, and an exclusive license there is a way to attach cooling to a product line Molex already ships, independent of whether BoundaryCool ever becomes a cold plate replacement at scale. For CAEPlus, founded to address changing data center cooling requirements as processor power and heat density rise [12], the investment supplies cash plus Molex engineering and hardware-scaling capability as it moves toward commercialization [8][13]. Founder and CEO Milad Bashirzadeh describes the goal as enabling a new class of active liquid cooling solutions [14].
The honest summary: an interconnect vendor has bought optionality on a cooling startup, taken an exclusive license in the segment it already owns, and made a compatibility claim that is the whole value proposition and is currently unquantified. Retrofit compatibility is easy to assert in a press release and hard to deliver, because the constraint is rarely the cold plate. It is the loop, the CDU, the floor loading and the facility water temperature the existing hall can actually supply.