SiliconAngle covered Iceotope's $26 million Series B on May 14 as a scaling story rather than an invention story. Its framing leans on CEO Simon Jesenko's line that the company has spent years building products purpose-built for AI infrastructure and is now ready to scale, set against a SemiAnalysis projection that liquid-cooled AI accelerator capacity grows from roughly 3 gigawatts to more than 40 gigawatts inside two years.
The product line is the argument. KUL AI is the rack-scale system for cloud and hyperscale floors. KUL BOX is a compact unit aimed at single servers and edge sites. Both use a non-flammable dielectric fluid in a sealed chassis rather than facility water at the chip. Iceotope calls the approach direct-to-everything.
The phrase is doing real work. Cold plate cooling removes heat from the hottest components and leaves the rest of the board on air, which keeps a residual air-handling burden in the room. Direct-to-everything puts the whole board in contact with dielectric fluid, which removes that residual and the failure surface that comes with mixed air and liquid in the same rack. The thermal case is strong. The reason immersion keeps losing to cold plates has been integration friction and serviceability, not heat removal. Iceotope's sealed-chassis design exists specifically to answer that objection, and the scale-up money is a bet that the answer now holds.
KUL BOX targets telecom, media, and enterprise sites running AI inference close to where data is produced. That is a different buyer than the hyperscale floor, with no raised-floor water plant and no thermal engineering staff. A sealed dielectric unit that ships and installs without facility water is a cleaner fit for that environment than a cold plate loop that assumes a CDU and a manifold. The market position of single-phase immersion in 2026 depends heavily on whether edge AI becomes a real volume segment or stays a pilot category. Iceotope is funding as if it becomes real.
The honest read of the SiliconAngle framing is that Iceotope's risk has moved from physics to operations. The technology works. What $26 million has to buy is production throughput, qualified supply, and field service coverage at a pace that matches a market doubling in under two years. That is the same wall every challenger in the liquid cooling supply chain race hits. The vendors that clear it convert a design win into a multi-site standard. The ones that do not become a footnote with good patents. Iceotope has the patents. The next eighteen months decide whether it has the line.