Tracking AI Data Center Cooling Companies

Tracking AI Data Center Cooling Companies

Global League (G-L.Ventures ) is committed to a data-driven top-down research approach to analyze market trends, landscape and dynamics for investors and startups. Our thesis currently prioritizes the intersection of AI infrastructure, energy transition, and national security. (Learn more about AI Infrastructure thesis-based deal tracking)


Quick Overview of Data Center Cooling Tech

The data center cooling industry is undergoing significant changes as the surge in AI workloads drives significant demand for advanced cooling technologies. Compared with other segments related to the recent AI hype, the investment in this segment is “cooler,” so there are lots of opportunities for investors. Presently, 80 percent of data centers employ air cooling, which constitutes 40 percent of the total energy consumption within data centers.?

Beyond high power usage, water consumption is another operational concern. For instance, an average Google data center consumes 450,000 gallons of water daily . Other major tech companies, such as Meta report comparable water usage scales. New cooling solutions must address AI data centers' CapEx and OpEx considerations and reduce computing's environmental impact and square footage! Keywords or key metrics include power density, power usage effectiveness (PUE) , and water usage effectiveness (WUE) .

There are room cooling, in-row cooling, rack cooling, free cooling and more for different designs. Air cooling has an upper limit of effectiveness at approximately 20kW per rack. Beyond this threshold, liquid cooling, whether through direct-to-chip or immersion methods, becomes imperative. Major players in the data center space have all made announcements around liquid cooling adoption (King of Hill in data center cooling , by TDK Ventures) Some liquid cooling solutions use cold plates and internal pumps to dissipate heat on certain components. Most new servers introduced into data centers today are preconfigured for deployment and more often sold with liquid cooling connectors.?

While liquid cooling only requires retrofitting servers and data centers, immersion cooling usually requires reconfiguring the racks to submerge them in a liquid that doesn't conduct electricity (dielectric fluids). Some startups are addressing older systems to enhance cooling without a retrofit of old data centers. Rack-mountable immersive cooling is available. Both liquid and immersion cooling as well as hybrid environments that combine air and liquid cooling are being adopted.?Vertiv, the leading company in this space, gives a diagram of the evolution.

Credit: Vertiv

There is a wide range of innovations for cooling hotspots on chips at different granularity levels, including integration into chip packages. As part of ARPA-E's COOLERCHIPS program , Nvidia and HP are jointly working on advancing embedded microfluidic cooling technology at the silicon wafer level. Chips integrated with microfluidics have been under development by some organizations for years. Below is an image from imec, the top semiconductor research institute.

Credit: imec

There have also been developments in new data center infrastructure designs, ecological integration with environments, and recycling of computing heat into electricity or heating water. Some have been doing the same for bitcoin mining. On the other end, some specialize in the construction, deployment, and maintenance of subsea data centers and claim that by locating data centers on the seafloor, enterprises can reduce their data center cooling costs by up to 90%.?

New materials or physics mechanisms for better cooling including waterless liquid cooling and better dielectric materials for semiconductor chips are being developed and backed by VC/CVC investors. Thermal management is a more holistic approach than just cooling. Many companies seek to reimagine a data center industry with no heat wasted through partnerships or new technologies. Let’s take a look at some companies in this segment.?


Examples of Tracked Companies

Large players: Vertiv Holdings Co. (US), Schneider Electric SE (France), Munters (Swedish), STULZ GmbH (Germany), Coolcentric (US), Rittal GmbH & Co. KG (Germany), Nortek Air Solutions, LLC (US).

High-growth startups to watch (some examples with growth signals; total funding numbers are as of September 2024) -

Nautilus Data Technologies (United States)

Nautilus provides modularized prefabricated compact data center infrastructure and sustainable cooling designs to reduce build time from two years to only six months. Their green data centers achieve Power Usage Effectiveness (PUE) of 1.15 or less, 100kW per rack, heat rejection over 8000 watts per square meter, 50%+ savings in energy usage, 30% CO2 emission reduction and up to 30% cost reduction. No water consumption, no refrigerants, no chemicals, no chillers, and no cooling towers. Server heat can be captured by traditional hot aisle, rear?door cooling, direct?to?chip, or immersion cooling. (performance comparison )

Investors: OIC, Ireland Strategic Investment Fund, Emerson Collective

Total funding / Total equity funding (rounded to million in USD): $136M / $36M

Iceotope (United Kingdom)

Iceotope provides Precision Liquid Cooling through the precise delivery of dielectric fluid across the entire IT stack using the same rack-based architecture as air-cooled systems. It combines the best attributes of tank immersion and direct-to-chip cooling and reduces energy use by up to 40% and water consumption by up to 100%. It also offers edge computing solutions and micro data centers. Its partners include Schneider Electric, SK Telecom, SK Enmove, Intel, Hewlett Packard Enterprise, and Mellanox. They have set up its US HQ, and also snags a Microsoft data center executive as their advisor.

Top investors: ABC Impact, Aster, Ombu Group, nVent, Salica

Total funding / Total equity funding (rounded to million in USD): $73M / $73M

Jetcool Technologies (United States)

JetCool’s patented Microconvective cooling? liquid cooling is a platform technology that uses arrays of fluid jets to cool the industry’s highest-power devices. Unlike typical heat sinks or traditional cold plates that pass fluid over a surface, their cooling jets route fluid directly at the surface, creating an order-of-magnitude improvement in heat transfer. The foundry-friendly solutions are added during chip fabrication, eliminating the need for thermal pastes or metal heat sinks. (more on performance benchmark )

Top investors: Bosch Ventures, In-Q-Tel (IQT), Raptor Group, Schooner Capital, MassVentures, ARPA-E

Total funding / Total equity funding (rounded to million in USD): $37M / $35M

Qarnot (France)

Qarnot recycles computing waste heat by constructing data centers in buildings with closed heating circuits, keeping buildings and shower water warm year-round. The company's software services cater to various sectors, including finance, 3D animation, fluid dynamics, big data, AI, and medical research. Qarnot also provides hardware solutions for IT waste heat recovery and ecological heating solutions. Additionally, the company offers a carbon accounting tool, supplying detailed reports to manage energy consumption and reduce environmental impact.

Top investors: Société Générale Ventures, ADEME Investissement, Demeter, la Banque des Territoires, Colam Impact, Avolta Partners

Total funding / Total equity funding (in USD): $52M / $25.5M

Green Revolution Cooling / GRC (United States)

GRC develops fluid submersion cooling systems including a white mineral oil formulation for cooling servers and other data center hardware, rack installations, and pump modules. Their offering eliminates the need for chillers, air conditioners, etc. It has partnered with Dell Technologies, Intel, and Vertiv. Shell, which offers its own cooling fluid, has deployed some clusters from Penguin Solutions in Texas data centers that use AMD processors within GRC Immersion Cooling Systems. (News )

Top investor: SK Lubricants

Total funding / Total equity funding (rounded to million in USD): $35M / $35M

LiquidStack (United States)

LiquidStack was incubated by and spun out of Blockchain infrastructure provider Bitfury. It provides direct-to-chip coolant distribution unit (CDU), single-phase and two-phase immersion cooling. Having pioneered advanced liquid cooling solutions since 2012, it has served hyperscale, colocation, enterprise, edge, and blockchain data centers. (performance benchmark and white paper ) Back in 2015, Bitfury acquired Hong Kong-based Allied Control, which specializes in immersion cooling using Novec Engineered Fluid created by 3M. At that time, Allied Control’s client reported a PUE of 1.02.?

Top investors: Tiger Global, Trane Technologies, Wiwynn (strategic, whose product offerings include two-phase immersion cooling systems claiming PUE < 1.01)

Total funding / Total equity funding (rounded to million in USD): $40M / $40M

ResetData (Australia)

ResetData provides “Liquid Cooling as a Service,” cloud services that include bare metal servers, virtual machines and GPU cloud. It is committed to data sovereignty and security. Operating across the whole value chain, its capabilities cover a new engineering and fit-out format in the data hall, a systems integration capability, including server OEM, a cloud services capability to enable customer utilization. 40% lower TCO. It also reduces carbon footprint by 45% and eliminates water waste entirely. Australasian real estate fund manager Centuria Capital Group has acquired 50% stake in ResetData for up to AUD 21 Million to establish liquid immersion cooled edge data centers in Australia and New Zealand.

Top investor: Centuria Capital Group?

Total funding / Total equity funding (Buyout,? in AUD): 21M / 21M

GemaTEG (United States)

GemaTEG pivots the industry from cooling to Thermal Management, fine-tuning silicon temperatures to boost CPU and GPU performance beyond just cooling chips. It has partnered with Leonardo, a global aerospace, defense, and security leader, to enhance the efficiency and sustainability of AI infrastructures. By implementing the DaTEG solution, it aims to boost the performance of next-generation chips within Leonardo's data centers. DaTEG? 1.0 supports chips consuming up to 500 watts, with future enhancements to accommodate chip designs of up to 1000 watts. Rather than waiting to upgrade to more powerful yet costly servers, DaTEG enhances the efficiency and capacity of current systems.

Top investors: Flywheel Investment Conference, RoundX

Total funding / Total equity funding (in USD): $4.1M / $2.2M

Asperitas (The Netherlands)

Asperitas launched its first data center cooling solution in 2017 and its new Direct Forced Convection technology in June 2024 for immersion cooling solutions across multiple workloads, including HPC and AI. It partners with Stulz to offer a compact, modular end-to-end data center solution with integrated immersion cooling for indoor and outdoor installation. Also, it has built an alliance of partners including Shell, Intel, Dell Technologies, Kiturami Bumyang, Gigabytes, Siemens, Kingston, Exellyn, Penguin Computing, Exellyn, etc. (white paper and spec. )

Top investor: PDENH, EASME - EU Executive Agency for SMEs

Total funding / Total equity funding (in USD): $81.9K(grant) / unknown

Corintis (Switzerland)

Corintis specializes in silicon microfluidic cooling technology with a monolithic, 3-dimensional network of microscopic channels that enable 2,000 W/cm of heat extraction, extremely low-pressure drop, and reduced coolant consumption. This also enables very low operating temperatures and is a key enabler for future high-density and vertically integrated IC designs.?

Top investors: Venture Kick, Intel Ignite, Acequia Capital (AceCap), Rand Hindi, Prix Strategis

Total funding / Total equity funding (in USD): $205.8K / unknown


These are just some of the companies in this AI data center cooling technology segment, where we are tracking about 80 companies from pre-seed to IPO stages. A large array of interesting pre-seed and seed-stage challengers, some with promising signals, are on our radar. If you would like to get data on all the companies we’re tracking and receive regular updates, join our club . The goals of our AI Infrastructure thesis-based deal tracking are:

  • Tracking big pictures of market landscapes, competitions, dynamics, signals, overinvested spaces, and underinvested opportunities.?
  • Tracking companies and investors in focused spaces. Who are the top investors? Who are the mature, early-stage, and bootstrapped companies to watch?
  • Tracking deals and investing opportunities for promising companies proactively.
  • Knowing what works and what does not in these spaces, in the past and now.
  • Learning about which public or private companies in the value chain are strong competitors, strategic partners, or investors for a company.
  • Doing deep dives on specific companies, competitions, or investors on demand.

AI computing infrastructure and data centers are important for national security. Countries like Europe and Australia are securing data sovereignty and security by building and operating their own hyperscale data centers and cloud computing. Aligning with clean energy sources or building new ones to power AI computing is in strong demand.?

The US needs three times its current nuclear energy capacity to meet AI's growing power needs, the US Department of Energy (DOE) has published an updated version of its Pathways to Commercial Liftoff Advanced Nuclear report, warning that AI and data center load growth will require tripling nuclear capacity by 2050; from 100GW to 300GW. We’ll share trends and companies working on these fronts in the following newsletters.

Our data sources: Netzeroi, CB Insights, Pitchbook, Crunchbase, Data Center Knowledge, Data Center Dynamics, DOE, and news media across the Internet.


Recent News to Review

S&P 500 Earnings Report from FactSet (published on Sep 27, 2024) -

For Q3 2024, the estimated (year-over-year) earnings growth rate for the S&P 500 is 4.6%.

The Information Technology sector is expected to report the highest (year-over-year) earnings growth rate of all eleven sectors at 15.6%. At the industry level, 5 of the 6 industries in the sector are projected to report year-over-year earnings growth: Semiconductors & Semiconductor Equipment (37%), Technology Hardware, Storage, & Peripherals (14%), Software (7%), Electronic Equipment, Instruments, & Components (5%), and IT Services (2%).

At the company level, NVIDIA ($0.74 vs. $0.40) is expected to be the largest contributor to earnings growth for the sector. If this company were excluded, the estimated (year-over-year) earnings growth rate for the Information Technology sector would fall to 8.2% from 15.6%.?

AI Accelerated Servers Fuel Growth in Data Center Spending

Data centers increased capital expenditures (capex) by nearly 50% in the second quarter of 2024, driven by the appetite for AI-ready hardware, according to a new report from Dell’Oro Group. Dell’Oro forecasts a robust 35% growth in worldwide data center capex overall for 2024. Insight from the California-based research firm indicates that data center capex grew by 46% in Q2 of 2024, compared to the same period a year earlier.

Some related news we collected:

  • Microsoft spent $55.7 billion on capital expenditures (capex) during its fiscal 2024 year (ended June 30), most of which went toward AI data center infrastructure and chips. Microsoft’s AI bet will lead to over $100 billion in data center finance leases. In July, Microsoft told investors in a footnote of its annual report that finance leases that had not yet begun had soared to $108.4 billion.
  • Amazon spent $30.5 billion on capex during the first half of 2024, mostly related to AI.
  • Alphabet spent $25 billion on AI capex in the first half of 2024.
  • Meta Platforms plans to spend between $37 billion and $40 billion on AI capex for the whole of 2024.
  • Oracle spent $6.9 billion on AI capex in its fiscal 2024 year (ended May 31).
  • Tesla and xAI are projected to allocate $20 billion to AI initiatives by the end of 2024, according to a Morgan Stanley analysis.

Nvidia has picked up another startup (at a steep discount): OctoAI

OctoAI specialized in optimizing and deploying AI models. Nvidia bought it for somewhere in the $165M-$250M range, a steep discount from its prior valuation of $900M and a poor outcome for most of its investors (it had raised $132M in total).

Nvidia also acquired Run:ai for $700M (raised $118M in total) and Deci AI (raised $55M in total) earlier this year. Both aim to optimize AI workloads and make utilizing Nvidia’s pricey chips more efficient. (CB Inisghts had a relevant analysis )

Cerebras Systems Announces Filing of Registration Statement for Proposed Initial Public Offering

Cerebras intends to list its Class A common stock on the Nasdaq Global Market under the ticker symbol “CBRS.”

(Reuters ) Cerebras' revenue surged to $78.7 million in the 12 months ended Dec. 31 versus $24.6 million a year earlier. Loss narrowed to $2.92 per share compared with $4.28 apiece. The company said 83% of its total annual revenue came from AI tech group G42, which is partly owned by Abu Dhabi's sovereign investor Mubadala.?

Submer Secures $55.5M Round for Immersion Cooling

Submer announced an investment round of $55.5 million USD, led by M&G, an investment company with £346.1 billion assets under management as of June 2024.

The investment comes from its £5 billion purpose-led Catalyst strategy, with Barclays Bank Ireland PLC acting as the placement agent to Submer. Among existing investors, Planet First Partners and Norrsken VC reinvested alongside M&G, and together with Mundi Ventures, reaffirmed their support as strong backers of Submer’s strong financial position and growth potential.


Apply to Join Global League Club


Our previous newsletters can be read here on LinkedIn .

要查看或添加评论,请登录

社区洞察

其他会员也浏览了