Apheros: Shaping The Future Of Data Center Cooling

As chips run hotter and system architectures become more complex, cooling has become a key constraint. Cooling alone is expected to consume up to 5% of global electricity by 2030 and already accounts for as much as 45% of a data center’s energy use, often adding millions of dollars in annual operating costs per facility. 

Power densities in data centers are rising fast, and conventional cooling systems are reaching their limits, putting performance, reliability, and energy efficiency at risk. 

To address this challenge, the industry is increasingly turning to advanced methods like two-phase and immersion cooling. However, fully unlocking their potential will require new materials that enable efficient heat transfer, support rapid vapor dynamics, and integrate with existing cooling infrastructure.

Apheros was founded in 2023 and spun out of ETH Zurich by Julia Carpenter (CEO) and Gaëlle Andreatta (CTO). The company develops high-performance cooling systems based on proprietary metal foams designed to improve thermal management across a range of advanced data center applications.

In August 2024, Apheros raised $1.85M in a pre-seed round led by Founderful.

Learn more about the future of data center cooling from our interview with Apheros co-founder and CEO, Julia Carpenter:

What Inspired You to Start Apheros?

I have a technical background and a PhD in Materials Science. During my PhD, I was fortunate to be part of a research group where four startups were founded while I was there. That was my first real exposure to entrepreneurship. I hadn’t planned to start a company. I thought I’d finish my PhD, become a manager, earn a good salary, and that would be my path. But watching people build not just companies but also company cultures really inspired me.

During my research, I developed a scalable process that resulted in a material with outstanding thermal properties. That was incentive enough to build a startup immediately after completing my PhD.

How Is Apheros Rethinking Thermal Management for Data Centers?

When I started, I thought I would just sell the material. But soon after, we pivoted. Today, we develop cooling solutions based on our metal foams, which combine ultra-high surface area, fully open porosity, and low density. We developed both the manufacturing process and the internal structure of the foam to achieve these properties. Our products include on-chip cooling components, cold plates, and boiler plates. These are compatible with different system-level cooling approaches, such as direct-to-chip liquid cooling and full immersion, depending on the specific architecture.

Our materials are engineered for easy integration and can serve as drop-in replacements for traditional thermal management solutions. Wherever there is a liquid involved, whether it stays liquid or evaporates, our solution makes sense. Current applications focus on high-performance electronics in data centers, where cooling is becoming a critical bottleneck. Currently, we are building MVPs with companies and aim to sell our cooling systems as integrated solutions. 

Editor’s note: On-chip cooling refers to thermal management techniques built directly into or onto the silicon die. They aim to solve extreme localized heat at the transistor level or die hotspots.

Editor’s note: Direct-to-chip cooling uses cold plates or similar devices placed on the chip package to transfer heat into a circulating liquid. The goal is to handle overall system heat more effectively without requiring exotic chip modifications.

Editor’s note: Immersion cooling submerges entire systems in a non-conductive liquid, allowing heat to dissipate passively or through phase change. It is ideal for densely packed electronics and environments where noise, space, or energy efficiency are critical.

How Does Your Material Perform in Two-Phase Cooling Systems?

Cooling today is mostly single-phase, which is simpler because there is no vapor involved. But as power densities continue to rise, two-phase cooling becomes more efficient, and a broader shift toward it is inevitable. We are preparing for that transition. 

Our foams offer strong flow and thermal performance, and they also promote fast bubble nucleation and release, which is essential in two-phase systems. It is crucial that vapor bubbles do not linger and form insulating gas layers. That is where our materials perform particularly well.

Editor’s note: Single-phase cooling uses a circulating liquid that remains in the same physical state throughout the cooling process. It is simpler to manage than two-phase systems and is widely used in today’s data center infrastructure.

Editor’s note: Two-phase cooling involves a liquid that evaporates as it absorbs heat, then condenses elsewhere in the system. This phase change allows more efficient heat transfer but requires careful control of vapor behavior to avoid thermal resistance.

What Is the Current Alternative to the Material You’re Developing?

Our material is designed for heat exchange. Currently, this function is typically handled by fin structures made of aluminum or copper, which are widely used due to their good thermal conductivity and ease of manufacturing. These fins increase the surface area, and liquid flows between them to remove heat. However, the geometry is limited by the fin size and density, especially under pressure drop constraints.

In our case, the liquid flows through a metal foam instead. This gives us a much higher internal surface area and a more uniform heat exchange, which results in improved thermal performance.

What’s the Biggest Challenge for You Right Now?

One of the main challenges is understanding the needs and wants of the different players in the industry. It varies from server designer to server designer, from chip to chip, and even from one data center to another. Everyone has slightly different preferences for how they want to build their cooling systems, including the pressures they use in the cooling circuit, the rack setup, and other design details. Getting clarity on all of this as a young startup has been one of the biggest hurdles, and we are working hard to navigate it.

How Well Can Your Solution Be Integrated Into Current Data Center Infrastructure?

Our cooling solution is a one-to-one replacement for existing systems. It is a fully integrated cold plate that connects to standard liquid cooling loops and can be installed without any changes to the surrounding hardware. It works with any service provider and in any data center environment. This makes it compatible with both current infrastructure and systems that are being planned.

What Milestone Would Validate Your Model at Scale?

We aim to deploy a number of our cooling solutions in data center racks and run them under continuous operation for several months. This will allow us to demonstrate long-term performance, reliability, and thermal stability under realistic conditions. The data center industry tends to be conservative, so extended runtime testing is essential. Reaching that milestone is a key step for us.

What Are Applications Where Your Material Offers the Greatest Performance Advantage?

The main use case is thermal management for high-power systems such as chips, power modules, and batteries, particularly in environments that already use or are transitioning to liquid cooling, as power densities continue to rise. Our foams are especially effective in data centers and high-performance computing, where managing heat reliably is critical. Their internal structure supports efficient heat transfer, helps extend component lifetime, and can reduce overall energy consumption.

We also see strong potential in electrolyzers, including systems that produce green hydrogen or convert CO₂ into other chemicals such as ethylene. Our materials have shown excellent results in this context. Compared to conventional porous components like meshes, felts, or sintered powders, our foams offer better flow characteristics and more effective reaction surfaces, which can help improve the overall efficiency of electrochemical systems.

What Areas of Data Centre Cooling Will See the Biggest Advancements Over the Next 5 to 10 Years?

I think we will see a significant shift toward immersion cooling, not just for CPUs and GPUs but for entire systems, which are generating increasingly high levels of heat. Immersion allows for more uniform and efficient cooling compared to conventional methods. However, it requires different infrastructure and workflows, so adoption will take time. Still, I believe this is the direction the industry is heading.

Another important shift I hope to see is better reuse of waste heat. That could mean using it for district heating or, in Switzerland, even for cheese production. Capturing and redirecting thermal energy that would otherwise be lost could bring real environmental and economic benefits. Designing systems with heat reuse in mind from the beginning would be a significant step forward, although progress in this area will depend on political support and regulation, which can be slow.

What Advice Would You Give to Fellow Deep Tech Founders?

Two things come to mind. First, treat it like a marathon, not a sprint. It is important to take care of yourself along the way. Take time off, go on holidays, and make sure to end your workday at a reasonable hour. If you work through every weekend for five years straight, you will burn out, and that is not sustainable for you or your company.

Second, try not to listen too closely to the naysayers. You know your technology better than anyone else. Product-market fit often takes time in deep tech, but most founders have a strong sense of the value their technology can deliver. Trust that understanding, stay focused, and keep building.