A Hidden Gas Threatens Chipmaking's Tiniest Transistors

SnowPure research finds dissolved oxygen can damage advanced chips at levels once considered harmless.

15 Feb 2026

For decades, engineers treated dissolved oxygen in ultrapure water as a corrosion nuisance, something to manage at the margins rather than obsess over. New research from SnowPure suggests that era is ending. At the two nanometer node, the company argues, dissolved oxygen has become a functional device contaminant capable of degrading transistor performance directly.

The numbers behind that claim are striking. SnowPure found that concentrations above one hundred parts per trillion can meaningfully affect transistors in advanced gate all around architectures. That threshold sits far below what older water treatment specifications ever targeted, meaning systems built for yesterday's chips may already be unfit for today's.

Why does this matter so much now? Gate all around transistors pack far more surface area into a smaller space than their predecessors, giving oxygen more opportunity to interfere with delicate structures. What once dissipated harmlessly now leaves a measurable fingerprint on device behavior.

None of this is happening in isolation. It sits alongside related SnowPure work on silica and boron removal in electrodeionization, the polishing technology used downstream of reverse osmosis to reach the purity levels advanced fabs demand.

Silica exists mostly as an uncharged species in neutral water, so its removal happens in a specific polishing zone within the module, where the compound finally becomes ionized. As node geometries shrink, required silica levels have fallen from the low parts per billion range toward one hundred parts per trillion at the most advanced nodes. The margin for error keeps shrinking alongside the transistors.

Boron poses a different puzzle entirely. Unlike silica, it does not polymerize, which makes it harder to remove through reverse osmosis, ion exchange, or electrodeionization alike. Engineers cannot simply apply the silica playbook and expect it to work.

Together, these findings point to the same conclusion from different angles. Instrumentation and treatment specifications calibrated for previous generations of chips may no longer catch the contaminants that matter most. Fabs chasing the next node will need water systems built around thresholds nobody thought to measure a few years ago.