Accelsius Expands Austin Lab for Megawatt AI Cooling Tests

AI liquid-cooling test equipment at an Austin data center cooling facility at sunset

Accelsius said Monday that it has upgraded its Springdale facility in Austin, Texas, to multi-megawatt power, giving the company a larger test bed for two-phase, direct-to-chip cooling systems aimed at future AI racks.

According to the company’s October 12 announcement, the lab can test a single rack at megawatt-plus loads or multiple racks at today’s highest densities. Accelsius also plans to add 18,000 square feet to the site in January 2027.

The expansion arrives as AI infrastructure vendors push toward much denser systems. BitcoinVersus.Tech recently covered Delta’s 3 MW cooling platform, 1 MW-class AI rack power, and liquid-cooled AI data center expansion.

Accelsius says two-phase direct-to-chip cooling can reduce cooling energy versus single-phase systems.

Accelsius says its NeuCool platform uses a two-phase dielectric refrigerant loop to remove heat directly from chips. The company also disclosed a forthcoming 375 kW-capable multi-rack unit beyond its existing MR250 product.

The company cited July benchmark results showing NeuCool produced NVIDIA B200 junction temperatures 9°C lower than single-phase direct-to-chip cooling under warm-water conditions. That performance claim comes from Accelsius and has not been independently retested by BitcoinVersus.Tech.

Accelsius CTO Richard Bonner explains the company’s two-phase direct-to-chip cooling approach.

One response to “Accelsius Expands Austin Lab for Megawatt AI Cooling Tests”

  1. […] launch connects with BitcoinVersus.Tech coverage of data-center liquid cooling, megawatt AI cooling tests and server drive […]

    Like

Leave a Reply