Innventure (INV): The Only Public Play on the Next Generation of Datacenter Cooling
Article # 11
Summary
Unique Commercialization Model: INV partners with large multinational corporations (MNC’s) to commercialize their most promising innovations, creating standalone operating companies with built-in strategic support.
De-Risked by design: These technologies typically come with years of prior R&D, validation, and market research, and often the sponsoring multinational is an early customer or channel partner.
Direct Exposure to Next Gen AI Datacenter Cooling: INV’s most valuable near-term asset is Accelsius, a dual-phase, direct-to-chip cooling company addressing the rapidly escalating thermal and energy constraints of AI datacenters.
Strong Industry Validation and Traction: Accelsius has raised nearly $110M from blue-chip strategic investors, built a $1B+ pipeline, and been selected as a key contributor to the U.S. Department of Energy’s ARPA-E COOLERCHIPS program in the last 24 months.
Repeatable, Capital-Efficient Flywheel: INV reinvests cash flows from operating subsidiaries into new incubations, creating a scalable, venture-style upside model supported by permanent public capital.
Intro
In the current AI and datacenter buildout, one of the key emerging constraints is the dissipation of heat from newer-generation chips and increasingly dense rack designs. Several technologies are being developed to address this rising heat load. One of the most promising is liquid cooling, which comes in multiple forms, including immersion cooling, single-phase cooling, and a newer approach known as dual-phase cooling. To our knowledge, INV is the only publicly traded company that provides direct exposure to this next generation of datacenter cooling technology. In addition, investors gain a majority stake in two other innovative companies, as well as partnerships with several major multinational corporations aimed at commercializing some of their most compelling innovations.
Source: Innventure Analyst Day 2024
Innventure management has executed this model multiple times before, most notably and quite publicly, with Procter & Gamble. In that case, they helped commercialize and scale PureCycle ($PCT) from a promising idea into a fully operating business, ultimately taking it public. Today, PureCycle has a market capitalization of nearly $2 billion. Remarkably, Innventure invested only a relatively modest amount, likely just a few 10’s of millions, of initial capital to bring PureCycle to the public markets at an approximate $1 billion valuation. This outcome highlights the potential of the venture-style model they employ.
Source: 2026 Analyst Day
While INV is no longer involved with PureCycle, the experience serves as a strong example of management’s ability to execute on commercialization-focused partnerships. Building on that success, management concluded they could improve the model further by creating a company that develops multinational, sponsor-backed technologies into operating businesses using permanent capital raised in the public markets. This structure allows them to combine that capital with the future cash flows of operating subsidiaries to scale innovative, and, in their view, relatively de-risked technologies that often remain buried within the bureaucracy of large corporations.
In effect, as shown below, INV has created a hybrid model that blends elements of traditional venture capital, the Danaher-style operating plan but on early-stage businesses, and an operating conglomerate. This strategy enables management to make long-term decisions aligned with the best interests of the operating companies and, by extension, shareholders. Investments are monetized either by running subsidiaries as cash-generating businesses or, when opportunities arise, exiting at attractive multiples and redeploying the proceeds into the next generation of technologies within the portfolio.
Source: Innventure Analyst Day 2024
Management argues that this approach offers small investors access to venture-style opportunities without many of the drawbacks typically associated with venture capital funds, namely high fees, illiquidity, limited transparency, and forced exits driven by finite fund lives. At least in theory, we believe this rationale is sound, though, as with most strategies, success depends on execution.
As noted in the opening paragraph, the most compelling subsidiary in the near term is Accelsius, which has developed a novel cooling solution for the datacenter market. We believe Accelsius will be the primary driver of the stock in the near term, and as a result, much of this rest of this article will focus on that business. That said, INV also owns stakes in AeroFlexx, an innovative liquid packaging company, and a majority stake in Refinity, a platform focused on recycling plastic back into petrochemical feedstocks, both of which we will briefly touch on later.
Datacenters are heating up figuratively and literally
As many readers are likely aware, there has been a lot in the news recently about datacenters becoming too hot and the new NVDA chips overheating when using air cooled datacenter designs. This is becoming a major bottleneck for the industry as we move to more and more powerful AI chips.
Datacenters are the central nervous system of the digital world, and overheating continues to be one of the biggest threats to their operation. This is a problem that continues to evolve and magnify as rack densities increase, chips become smaller and power use goes up. Overheating can lead to downtime, hardware failure, and significant financial impacts that can be up to $500k/hour of downtime or more in a modern datacenter, according to Uptime Institute.
In a nutshell, overheating in datacenters is primarily a result of increasing rack power density, with each newer server creating more heat, and being more closely packed together makes cooling more challenging. Currently, the rate of heat increase is outstripping the capability of traditional air-cooled datacenters. This heat dissipation challenge of traditional systems can be exacerbated by inefficient or poorly designed air circulation and tighter rack spacing, a trend we are seeing broadly.
According to Josh Clamen, CEO of Accelsius, “80% of datacenters in North America are still air cooled, by the way. So, the transition [to liquid cooling] is happening. It seems like it’s happening at sort of speed of a rocket. But there’s still a lot of datacenters around that are air cooled…But basically, if you have a liquid cooled datacenter, whatever the cooling architecture actually is in the white space or within the racks..[you have]…what’s called a facility water loop or primary water loop, and that primary water loop [in the gray space] is cooled using one of several mechanisms, but typically, it’s cooled by using chillers…within a chiller, there’s a compressor, just like in your home air conditioning. And those are really power hungry devices. So, the less you can rely on chillers or rely on chillers with their compressors turned on…the more you can save money.”
Cooling datacenters takes an enormous amount of electricity, by some estimates as much as 30-40% of the operating expenses of a datacenter. This is because they rely on chillers that are similar to giant versions of the air conditioner in your home. These chillers require compressors, and compressors tend to be one of the most power-hungry devices in a datacenter, just like your air conditioner at home.
“The datacenter industry is facing a critical inflection point where thermal management is no longer just a facility concern, but a strategic imperative for IT performance,” said Brian DiBella, President and CEO of Legrand, North and Central America. Legrand is a publicly traded company in Europe, and widely regarded as a leader in electrical and digital building infrastructures. With that background, this quote carries significant weight, there is a strategic imperative to address cooling concerns of modern datacenters.
The problem is fundamental, if you push electrons through a resistor of any type, it will heat up. This is a problem for all electronics manufacturers, and they have come up with various solutions to manage and dissipate excess heat. Most datacenters to date use air to manage excess heat, but air is only so good as a thermal conductor. The historical solution was, chill the air, hence the need for massive amounts of air chillers in datacenters, the more advanced the more cooling required.
Free Subscribers can unlock 1 article free - Unlock the rest of this article (or any other of your choice) by adding your email above!
As alluded to above we are now seeing datacenters implement liquid cooling, which uses water, or another dielectric, non-conductive liquid to cool datacenters. As we all know water is somewhat of a no-no when working with expensive electronics, despite this some datacenters have implemented water cooled systems. Water is a better thermal conductor than air, but it does have a fundamental problem, when, not if, it leaks it is going to be costly. This has led to many liquid cooled datacenters to use dielectric liquids, or goo, as we like to call it, which when it leaks doesn’t ruin electronics.
