- Apheros has developed a material that can help boost the cooling power of cold plates
- The company isn't interested in systems design, but wants to be a materials supplier for cooling vendors large and small
- Analyst Lucas Beran said its tech is intriguing, but it still has plenty of hoops to jump through
When it comes to liquid cooling, the more surface generally means more cooling power. That’s why it’s potentially a big deal that Swiss startup Apheros claims to have developed a material that can increase the surface area inside cold plates by as much as 1,000x.
The company calls its material “metal foam,” two words that to the lay person seem to mean opposite things. But as cofounder Gaëlle Andreatta explained, “foam” simply refers to the porous structure of its solution while “metal” refers to the material used. Under a microscope, Apheros’ solution looks a bit like a metal sponge, though that specific term means something different in metallurgy, cofounder Julia Carpenter noted.
The material is something Carpenter stumbled upon while working on her Ph.D.
“It was one of those things where you get lucky, and it works,” she told Fierce. “We really could immediately tell both that the properties were extraordinary and also that we could scale this technology, which is not always the case. Sometimes you have great properties but no scalability at all.”
Carpenter and Andreatta said Apheros isn’t looking to get into cooling systems' design. Rather it’s looking to work with cooling vendors like Boyd as a materials supplier.
Basically, Apheros’ metal foam can be inserted into single phase cold plates used in data center servers to increase their cooling capacity by up to 90% (depending on the existing system design). It’s a cheaper way than carving channels in the plate to increase surface area.
For Apheros, success looks like “integration with the base systems” and “getting up and running in a data center” to help with cooling, energy consumption and overall environmental goals, Andreatta said.
Asked what segment of the market Apheros is looking to serve, Carpenter said it is targeting 800-1,000 watt chips. So, not the top of the line, but still beefy power levels, which will soon become a large middle-tier segment.
Apheros as a company is only a year old, so its efforts to scale production and conduct testing and trials of prototypes are still underway. But Carpenter said by the end of 2025 or mid-2026 it expects to be ready for production at scale.
And why are the founders so confident they can scale the technology? Well, for starters, they just raised $1.85 million to help them do so. But also the manufacturing techniques they plan to use are based on well-established food production tech used for things like chocolate mousse and meringue. Go figure.
The founders added Apheros is also eyeing ways to use its material for two-phase direct-to-chip and both single and two-phase immersion cooling down the line.
Analyst take
Dell’Oro Group Research Director Lucas Beran told Fierce Apheros’ technology “certainly seems like an exciting opportunity” but noted “a lot of questions remain about how viable it can be.”
Beran pointed out that breaking into the data center market tends to be a “very long process” due to the critical nature of compute infrastructure. So, he sees a lot of testing, validation and proof-of-concept work on the horizon.
He added Apheros will also need to address questions about what kinds of flow rates the metal foam can handle as well as how use of the material might impact overall system design. By way of example on the latter point, he questioned whether existing liquid cooling filtration systems would be sufficient to avoid contaminants clogging the microscopic channels in the metal foam.
Even if it gets over those hurdles, he said timing will be a factor in Apheros’ success.
The 800-1,000-watt chip segment “is an underserved portion of the market, but that’s also because you can get away with air and do things that prevent needing cold plates on processors,” he said. He added that today, there are also relatively few processors in that range but he would expect that to change in the coming years as the power levels for mid-tier chips creep upward.
“It definitely makes sense for where things will go, but it’s certainly not where the market is today,” Beran concluded.