Co-founded by a 25-year cleantech scaling veteran, Marc Ottolini, and Bernard Cooper, a PhD specialist in quantum cryogenic cooling, Infiniflux is bringing a novel approach to solve the AI chip overheating crisis. The UK-based deeptech startup and winner of the IC Taiwan Grand Challenge Award 2026 delivers a breakthrough cooling solution designed for next-generation hardware architectures.
While power consumption per chip continues to climb toward multi-kilowatt levels as artificial intelligence workloads skyrocket, current cooling architectures are struggling to remove heat fast enough without ballooning hardware footprints or energy bills.
In an exclusive interview with TechSoda, Infiniflux CEO Marc Ottolini detailed how their breakthrough technology—EPiC2—delivers a giant leap in data center efficiency and chip longevity.
“A few companies are trying to go even further by putting liquid cooling inside the dies themselves. We call that a bridge too far,” said Ottolini. “The energy they need to force liquid through the die is many times higher than what EPiC2 needs. That doesn’t make data centers happy because they are looking to bring their energy bills down, not up.”
A System-Level Platform, Not a Standalone Component
Crucially, Infiniflux does not view next-generation thermal management as a mere bolt-on part. “A cooling system is not a component. It’s a system that cuts across the whole chain and needs to be embedded inside the chip package,” explains Marc Ottolini.
This philosophy was the driving force behind the company’s decision to name the system EPiC2. Positioned as an open infrastructure platform rather than a simple hardware component, EPiC2 requires close co-design across multiple industries. Because the solution touches everything from package assembly to server chassis liquid loops, Infiniflux plans to license its IP directly to OSATs for package integration while providing turnkey loop hardware and controls to server ODMs.
What is EPiC2?
Infiniflux has officially rolled out its core technology platform name: EPIC2 (Embedded Phase Change in Chip Cooling).
EPiC2 integrates a specialized silicon cold lid directly inside the chip packaging, placing the two-phase cooling medium in immediate contact with the silicon die’s primary hotspots. By bypassing intermediate heat spreaders and traditional thermal interface materials (TIMs), EPiC2 achieves a 75% reduction in thermal resistance. This allows processors to operate at 15 to 20 degrees Celsius cooler junction temperatures even under intense multi-kilowatt power demands.
The system transports six times more heat per watt of cooling energy compared to single-phase liquid systems. Most importantly, EPiC2 delivers precise die temperature uniformity while consuming minimal pumping energy.
To understand why EPiC2 represents a step-change in semiconductor thermal management, one must look at how traditional liquid cooling operates. Ottolini explained in his IC Taiwan Grand Challenge Award Ceremony speech that standard Direct Liquid Cooling (DLC) relies on sensible heat transfer—forcing water with glycol through a copper cold plate mounted on top of the chip. As heat moves from the silicon through a Thermal Interface Material (TIM), across a heavy copper lid, through a second TIM layer, and finally into the copper cold plate, huge thermal bottlenecks build up. Furthermore, as the coolant absorbs heat while traveling across the plate, its temperature rises, creating uneven cooling across the die and forcing pumps to work at high pressures.
EPiC2 replaces sensible heat transfer with latent heat of vaporization through Direct2Die two‑phase cooling. Instead of simply heating a liquid, the die’s thermal energy boils a specialized non‑aqueous refrigerant directly at the hotspots. The phase change absorbs large amounts of heat without increasing fluid temperature, keeping junction temperatures extremely low and uniform while delivering very high heat‑transfer coefficients. In contrast, single‑phase cooling must rely on high pumping power to approach similar performance.
ROI for Data Centers: Near-Term and Long-Term Value
Infiniflux’s technology offers compelling economic returns across both immediate deployments and next-generation chip architectures.
Immediate Benefits for Existing Chips (2.5D Architectures)
For current 2.5D chip deployments, EPiC2 unlocks roughly 25% “free compute” by preventing thermal throttling and allowing processors to sustain higher clock frequencies continuously. It cuts operational expenditure by over 20% through reduced cooling energy demands and lower maintenance overhead. As EPiC2 works at high inlet temperatures, it also enables dry cooling, thus eliminating the huge water consumption of data centres. Furthermore, by eliminating extreme temperature gradients over the die surface, it more than doubles hardware lifespan. For a typical 100MW data centre representing a $3.5 billion IT capital investment, EPiC2 delivers between $1 billion and $1.5 billion in savings over three years, yielding a payback period measured in months.
Strategic Freedom for Next-Gen Silicon (including 3D ICs & CPO)
Beyond immediate operational returns, Infiniflux provides chip architects with 20 degrees Celsius of extra thermal headroom to reimagine future chip designs. This headroom allows chip suppliers to shrink overall silicon footprints without risking thermal runaway. EPiC2 enables dense stacking of scorching CPUs and GPUs right alongside heat-sensitive High Bandwidth Memory (HBM). Additionally, the company is also developing a cooling platform for vertical 3D IC architectures. In such chip configurations, their cooling technology will also play a role in so-called co-packaged optics (CPO), where optical interconnects suffer from temperature-induced signal drift; EPiC2’s two-phase mechanism maintains constant temperatures to guarantee optical reliability.
Zero Silicon Redesign & The Taiwan Ecosystem Advantage
Ottolini emphasized that EPiC2 requires zero modifications to the silicon die or foundry process itself. By integrating exclusively at the packaging stage, Infiniflux avoids adding years of re-engineering or expensive foundry qualifications.
To execute this fabless vision, Infiniflux relies heavily on Taiwan’s semiconductor ecosystem. The silicon cold lids are manufactured using standard MEMS foundry processes widely available across Taiwan. Moreover, Taiwan offers a unique geographic concentration encompassing chip design houses, OSATs, server ODMs, and hyperscaler partners within proximity.
What’s Next for Infiniflux?
The company is finalizing a significant institutional funding round with a Middle Eastern investor, linked to setting up an advanced AI infrastructure test facility. Concurrently, Infiniflux is leveraging SEMICON Taiwan to deepen relationships with local Taiwanese venture capital and strategic partners to build a permanent local presence to follow up strong engagement after winning the IC Taiwan Grand Challenge Award at COMPUTEX 2026.
The startup is assembling a launching coalition of key OSATs, ODMs, and chip designers to run parallel qualification tracks. By advancing these efforts simultaneously, Infiniflux aims to establish EPiC2 as an open packaging infrastructure standard that helps the global semiconductor industry conquer the thermal wall.