By Bill Haskell, CEO of Innventure
With an average of 1 billion daily queries, the server farms powering ChatGPT consume about 2.9 million kWh daily, equivalent to 100,000 U.S. households (based on their average daily consumption of 29 kWh)—and that's just one AI application in a market set to triple by 2027 (Forbes, 2024).
The Breaking Point
Behind the closed doors of data centers worldwide, a crisis is unfolding that threatens to bottleneck the entire AI revolution: heat.
"The hyperscalers alone—just the four largest—are spending over $250 billion this year on data center development," says Josh Claman, CEO of Accelsius. "That's $5 billion every week. Mind-numbing scale that's grown five-fold since 2022."This isn't a gradual evolution. It's a revolution happening at breakneck speed."
"Liquid cooling as a pragmatic solution has always been niche. But we're seeing the dynamic rapidly changing," Claman explains. "Mainstream CPUs are becoming much hotter after years of incremental increases in wattage—suddenly, we're seeing a J-curve in terms of the heat these chips produce."
The numbers tell the story: 48% of enterprise data centers now deploy racks exceeding 20kW of power, up dramatically from 31% just two years ago (451 Research, 2024). Modern AI workloads demand 30-50kW per rack or higher—with some projections showing future requirements reaching 250kW per rack (Vincent, 2025). Traditional air cooling, which struggles beyond 10- 15 kW per rack, has hit a wall.
Data centers have evolved from obscure server warehouses to what Emil Sayegh calls in Forbes, "essential infrastructure as highways and airports, connecting us to the digital economy that fuels our lives" (Sayegh, 2024).
But this digital backbone is overheating, and the consequences ripple through every industry dependent on computing power.
The Foresight Before the Storm
Rewind to summer 2022. As our team at Innventure (NASDAQ: INV) was finalizing the launch of Accelsius, most of the world had never heard of ChatGPT. OpenAI hadn't even released it yet.
We weren't clairvoyant—we were just connecting dots that others missed between physics, market economics, and computing's inevitable trajectory.
"The industry was beginning to recognize that air had reached its limits for data center cooling. CPU and GPU wattages were growing at a rate never expected, and new approaches were needed," Claman recalls of that pivotal moment.
While others saw an emerging challenge, we saw an existential threat forming. The world's data centers were approaching what engineers call "the thermal wall"—a physical limitation that traditional air cooling simply cannot overcome. Every new generation of processors was generating more heat in tighter spaces. Something had to give.
At that critical moment, Innventure's systematic DownSelect® rigorous analysis, which advances only 2% of evaluated opportunities— identified breakthrough IP that delivered a two-phase, direct-to-chip solution capable of revolutionizing the market. We purchased the IP and brought in Josh Claman to build Accelsius and commercialize the technology into an enterprise-class solution.
What we couldn't predict was how quickly generative AI would accelerate this crisis from "emerging challenge" to "existential threat" for the entire digital economy. The market we anticipated for years four and five is here today.
The Perfect Storm Accelerates
Three converging forces have created what industry analysts describe as a "perfect storm" accelerating the transition to liquid cooling solutions:
The Density Dilemma Explodes
"We've never seen a market evolve this quickly," Claman notes. "Every few weeks, the conversation changes, becomes more sophisticated. While the AI market projections started at $3B by 2027, they've rapidly evolved to $5B, then $7B—and continue to grow."
The U.S. data center market has nearly doubled in size since 2020, growing at a rate of 21% annually, with vacancy rates approaching zero (Sayegh, 2024). The global data center industry is estimated to be worth $242.72 billion and is projected to more than double, to over $584 billion, by 2032 (Fortune Business Insights, 2025). Tech giants like Amazon, Google, Microsoft, and Meta are leading the charge, with the number of hyperscale data centers doubling roughly every five years (Forbes, 2024)
These facilities handle over 95% of the world's internet traffic, ensuring real-time connectivity for billions of users.
The Sustainability Imperative
"What surprises people is that about 1.2 percent of US energy is focused on cooling data centers—not on processing data, but on cooling data centers," says Claman. "We think we can take that down by 80 percent and dramatically reduce carbon emissions."
The numbers are staggering: Data centers now consume 1-3% of global electricity, with projections showing this could reach 4-5% by 2030 without efficiency measures (International Energy Agency, 2023). By decade's end, U.S. electricity consumption for AI-driven data processing is expected to exceed that used for the production of aluminum, steel, cement, and chemicals combined (Petroni, 2025).
With cooling representing 35-40% of a data center's energy consumption (Uptime Institute, 2024), advanced liquid cooling technologies can reduce cooling energy consumption by 56% compared to traditional air cooling (Lawrence Berkeley National Laboratory, 2023)—with some implementations achieving energy savings of up to 95% (Association for Computing Machinery, 2024).
The Financial Equation
"We talk about making every watt count," says Claman. "If you can only secure X megawatts of power, you want revenue-generating compute, not overhead cooling. By improving power usage efficiency, we dramatically increase revenue for each square foot of data center space."
Beyond the technical advantages, the economic case for liquid cooling has become too compelling to ignore:
- Dramatically reduced energy and operational costs
- Higher compute density maximizing real estate value
- Extended hardware lifespan reducing replacement costs
- New potential revenue opportunities through heat repurposing
- Essential future-proofing for AI workloads
The reality: Traditional air cooling is no longer sufficient or cost-effective. The clear economic advantages of liquid cooling have made its adoption not just beneficial but necessary.
The Two-Phase Solution
When we identified this market opportunity at Innventure, we knew it wasn't about creating an incremental improvement. It was about developing a breakthrough solution that would transform how data centers manage heat.
"What differentiates Accelsius? We're future-facing," explains Claman. "Our two-phase direct-to-chip cooling can handle multiple generations of future chips because of the phase change we orchestrate,” he continued. “Single-phase water simply can't address the heat flux demands of next-gen processors."
While McKinsey's analysis identifies three primary liquid-based cooling approaches—rear-door heat exchangers (40-60kW racks), direct-to-chip cooling (60-120kW racks), and immersion cooling (100kW+ racks) (Srivathsan et al., 2024)—Accelsius found a unique technological edge with two-phase direct-to-chip cooling.
"The dielectric, non-conductive fluid is no longer just a talking point—it's a fundamental advantage," says Claman. "Water leaks destroying $5 million server racks isn't theoretical FUD anymore. It's happening, and it's part of the ROI calculations of data center operators."
Built for Scale, Designed for Tomorrow
"NeuCool™ technology cools 1500+ watts of power per socket—which is unmatched among competitors," notes Claman. "NVIDIA H100 chips use 700W, but we're already prepared for next-generation processors that will generate significantly more heat."
The science is straightforward yet powerful: liquid can absorb more than 3,000 times more heat than air for the same volume, enabling Accelsius' technology to remove heat with up to 25x the efficiency of traditional air cooling.
Critically, ease of integration has been central to the design philosophy.
"Liquid cooling offers tremendous heat absorption benefits, but many competing products require a data center retrofit or an entirely new facility," explains Claman. "NeuCool™ fits into traditional racks without special mounting."
"Our current evaporative plates already test successfully with chips 2-3 generations beyond today's processors," Claman shares. "With our innovation pipeline in additive manufacturing, materials science, and micro-channeling, we're prepared for the computing demands of 2030 and beyond."
The Market Affirms the Vision
The data is clear: we've reached the tipping point for liquid cooling adoption.
Goldman Sachs estimates that the server cooling market will reach $10.6 billion in 2026 from $4.1 billion this year. Within that market, liquid cooling's penetration rate will reach 57% among AI servers in 2026, compared with just 23% this year (Wong, 2024). The broader data center liquid cooling market is projected to grow from $4.3 billion in 2023 to $11.2 billion by 2028, at a CAGR of 21.1% (MarketsandMarkets Research, 2024).
"Accelsius has gone from concept to commercial traction in under three years with Innventure's backing," Claman notes. "We're seeing early innovators convert to POs, joint innovation centers, and proof-of-concept deployments—exactly tracking analyst projections for market adoption."
Even efficiency breakthroughs like DeepSeek's AI models are accelerating rather than slowing this growth.
As John Medina from Moody's Ratings explains: "It lowers the cost of computing and potentially increases the number of new companies and new applications that can be created. It actually could lead to more usage than less usage" (IndustryDive, 2025).
Mitchell Osborne, director of MEP at Adolfson & Peterson Construction, reinforces this view: "While more efficient AI models will reduce the amount of power consumed, we are still in AI and quantum computing's infancy. I do not foresee data centers or power infrastructure construction slowing down" (IndustryDive, 2025).
The market is coming to Accelsius faster than anticipated. Original projections for years four and five are tracking closely to today's reality—a testament to both the technology and Innventure's strategic vision.
Riding the Wave
"Our four-segment market approach ensures we're riding the wave, not buried beneath it: penetrating hyperscalers through OEM partnerships, innovating with colocation providers, leveraging OEM channels, and serving AI-as-a-service providers with the most power-dense solutions," explains Claman.
Accelsius is designed to meet the scale of this opportunity.
This wasn't hope—it's preparation.
"When a Microsoft, AWS, or Google starts buying from us, the volume becomes unimaginable,” says Claman. “Our strategy: core internal manufacturing paired with strategic partners like Celestica, who can rapidly scale production lines globally."
This diversified approach is backed by a manufacturing strategy designed to handle explosive growth. Rather than being caught unprepared when orders start pouring in, Accelsius has built redundancy into its supply chain by prioritizing North American suppliers for faster response times and maintaining control over critical manufacturing processes.
Essential Infrastructure for the Digital Future
Hard tech innovation requires patient capital.
Innventure's “Conglomerate Model”—built around identifying and commercializing transformative technologies that we believe have billion-dollar enterprise value potential—provides exactly that foundation (Bain & Company, 2024).
Unlike software startups, Accelsius needs investment in electrical engineers, mechanical engineers, thermal experts, PhD scientists, and supply chain professionals. Where traditional VCs write checks and offer advice, Innventure delivers operational expertise, market access, and the entrepreneurial agility needed to transform technically validated solutions into potentially market-defining enterprises (Harvard Business Review, 2023).
We don't just fund companies—we found, operate, and scale them with a full-spectrum approach that provides both the strategic resources necessary for rapid commercialization.
This isn't just another tech trend. The strategic importance of cooling infrastructure for enabling the AI revolution is fundamental and non-negotiable.
What makes liquid cooling particularly compelling is that it isn't speculative technology—it's essential infrastructure for continued digital expansion.
The industry has moved from asking "if" to asking "when" and "how" to implement liquid cooling (Allan, 2025). For data centers, the transition is complete—liquid cooling is no longer optional but essential for continued growth.
The Infrastructure of Tomorrow, Built Today
When you're solving a problem that's both technically unavoidable and environmentally imperative, you're not just building a business—you're creating essential infrastructure for the digital future.
"The industry was beginning to recognize that air had reached its limits for data center cooling. CPU and GPU wattages were growing at a rate never expected, and new approaches were needed," explains Claman (Raghunathan, 2024).
This foresight positioned Accelsius to develop solutions precisely calibrated to the changing requirements of next-generation computing infrastructure.
The question isn't whether liquid cooling becomes essential infrastructure, but who will capture the value as this technological imperative transforms the digital economy. Those who build the cooling foundations for tomorrow's computing demands won't just participate in the market—they'll define it.
The calculus is straightforward: as computing becomes more powerful and more ubiquitous, the companies that solve these fundamental infrastructure challenges become as essential to our digital society as highways and airports were to the physical economy.
The tipping point isn't approaching—it’s already here. The infrastructure for tomorrow's AI economy is being built today. The winners aren't those who saw it coming, but those who built for it before others realized it was already here.
At Innventure, we play with the long game in mind—by creating opportunities we believe can achieve a billion-dollar enterprise value or more, and then we intend to hold onto our companies indefinitely.
Innventure's prescient identification of this market opportunity—and our decision to build Accelsius before the market’s tipping point—positions the company at the forefront of this essential infrastructure transformation.
The market has spoken. The physics are undeniable. The economics are compelling. And the solution—built with foresight before the world knew it needed it—is ready to scale.
This article represents industry research and analysis based on publicly available information and does not constitute investment advice. Readers should conduct their own due diligence regarding specific investment opportunities.
Bill Haskell is the CEO of Innventure, bringing over 30 years of experience in building and scaling technology companies across diverse sectors. A seasoned entrepreneur and executive, Bill has served as CEO, COO, and strategic planning executive for numerous ventures, and has been a director for more than a dozen public and private companies globally. His expertise spans company building, financial modeling, M&A advisory, and taking companies public through IPOs. Bill's unique approach combines deep industry knowledge with a talent for assembling high-performing teams, making him a sought-after leader in the world of corporate innovation and technology commercialization.
This article was originally posted as an original Innventure Thought Leadership Article.
References
451 Research. (2024). Data Center Rack Density Report.
Allan, K. (2025, January 25). Why liquid-cooled infrastructure has become mainstream. ITPro.
Association for Computing Machinery. (2024). Energy efficiency in data center cooling operations. Communications of the ACM.
Bain & Company. (2024). How Tech Leaders Commercialize Innovation.
Geary, B. (2024, March 6). Revolutionizing data center sustainability: The power of liquid immersion cooling technology. Forbes Technology Council.
Green Grid Association. (2024). Water Efficiency Metrics for Data Centers.
IndustryDive. (2025, January). DeepSeek and the future of data center demand. IndustryDive.
International Energy Agency. (2023). Data Centers and Data Transmission Networks.
Lawrence Berkeley National Laboratory. (2023). Energy Efficiency in Data Centers.
MarketsandMarkets Research. (2024). Data Center Liquid Cooling Market Report.
McKinsey & Company. (2023). Generative AI Market Analysis.
McKinsey & Company. (2024). Total Cost of Ownership Analysis: Liquid Cooling vs. Traditional Cooling.
Petroni, G. (2025, January 10). AI boom to fuel surge in data center energy needs, IEA says. The Wall Street Journal.
Raghunathan, P. (2024). How Accelsius, led by CEO Josh Claman, is revolutionizing data center cooling for AI. Pulse 2.0.
ResearchAndMarkets.com. (2024). Data Center Market Landscape Report 2024: A $398.8 Billion Industry by 2029.
Sayegh, E. (2024, September 25). Data centers at the heart of the AI and digital economy surge. Forbes.
Schneider Electric. (2024). Total Cost of Ownership: Liquid Cooling vs. Traditional Cooling in Data Centers.
Uptime Institute. (2024). Global Data Center Survey.
Vincent, M. (2025, February 5). How AI's transformative impact on data centers is driving unprecedented industry growth, innovation, and global expansion. Data Center Frontier.
Wong, J. (2024, May 28). Cooling for AI is a hot stock market trade—for now. The Wall Street Journal.
© 2025 Innventure. All rights reserved.
The forward-looking statements are based on the current expectations of the Innventure, Inc’s (the “Company’s”) management and are inherently subject to uncertainties and changes in circumstances and their potential effects and speak only as of the date of this news letter. There can be no assurance that future developments at the Company or any of its subsidiaries will be those that have been anticipated. These forward-looking statements involve a number of risks, uncertainties (some of which are beyond the control of the parties) or other assumptions that may cause actual results or performance to be materially different from those expressed or implied by these forward-looking statements. These risks and uncertainties include, but are not limited to, those factors discussed and identified in other public filings made with the Securities and Exchange Commission by the Company and the following: (a) expectations regarding the Company’s and its subsidiaries’ strategies and future financial performance, including their future business plans, expansion and acquisition plans or objectives, prospective performance and opportunities and competitors, revenues, products and services, pricing, operating expenses, market trends, liquidity, cash flows and uses of cash, capital expenditures, and the Company’s and its subsidiaries’ ability to invest in growth initiatives; (b) the implementation, market acceptance and success of the Company’s and its subsidiaries’ business models and growth strategies; (c) the Company’s and its subsidiaries’ future capital requirements and sources and uses of cash; (d) the Company’s ability to meet the various conditions, including performance targets, and access installments draws under its line of credit with Western Technology Investors; ; (e) the Company’s ability to meet the various conditions and satisfy the various limitations under the Standby Equity Purchase Agreement (the “SEPA”) with YA II PN, Ltd., including exchange caps, issuances and subscriptions based on trading volumes, to access the funds available under the SEPA; (f) that the Company will have sufficient capital to operate as anticipated; (g) the Company’s ability to obtain funding for its operations and future growth; (h) developments and projections relating to the Company’s and its subsidiaries’ competitors and industry; (i) its subsidiaries’ ability to meet, and to continue to meet, applicable regulatory requirements for the use of their products and the numerous regulatory requirements generally applicable to their products and facilities; (j) the outcome of any legal proceedings that may be instituted against the Company in connection with its recently completed business combination; (k) the Company’s ability to find future opportunities to license or acquire breakthrough technology solutions from multinational corporations (“MNCs”) and to satisfy the requirements imposed by or to avoid disagreements with its current and future MNC partners; (l) the risk that the Company may be deemed an investment company under the Investment Company Act, which would impose burdensome compliance requirements and restrictions on its activities; (m) the Company’s ability to sufficiently protect the intellectual property rights of itself and its subsidiaries, and to avoid or resolve in a timely and cost-effective manner any disputes that may arise relating to its use of the intellectual property of third parties; (n) the risk of a cyber-attack or a failure of the Company’s or its subsidiaries’ information technology and data security infrastructure; (o) the ability to recognize the anticipated benefits of its recently completed business combination; (p) geopolitical risk and changes in applicable laws or regulations; (q) potential adverse effects of other economic, business, and/or competitive factors; and (r) operational risks related to the Company and its subsidiaries.
