Discovered Materials secures $9 million to cool AI semiconductor chips

The immediate implication is a potential reduction in the massive power and cooling overhead that currently accounts for up to 40% of data center electricity usage. As streaming providers increasingly integrate AI for real-time transcoding and recommendation engines, the underlying hardware is hitting thermal limits that air cooling can no longer manage. By accelerating the discovery of thermally conductive dielectrics, Discovered Materials could enable higher density in 3D-stacked chips, directly lowering the cost-per-stream for B2B infrastructure. Watch for the first commercial licensing agreement with a major GPU vendor within the next 12 months as a signal of manufacturing viability.

The thermal challenge identified by Discovered Materials is underscored by the deployment of NVIDIA Blackwell GPUs, which operate at significantly higher power densities than previous generations. Per TechPlusTrends in April 2026, a single Blackwell rack can consume between 120 kW and 150 kW, producing thermal loads that physically exceed the capacity of air cooling. This shift has forced a massive industry transition toward liquid cooling infrastructure, with sustainable AI in marketing frameworks highlighting that data center electricity consumption is projected to reach 565 TWh this year, a 26% year-over-year increase driven primarily by AI-optimized servers.

Beyond hardware efficiency, the environmental impact of these facilities is facing increased regulatory and public scrutiny. According to reporting from World Resources Institute in February 2026, large AI data centers can consume up to 5 million gallons of water daily for cooling, comparable to the usage of a small town. In response, legislative efforts like Washington state’s House Bill 2515 have emerged to force transparency around energy and water consumption. For streaming professionals, these infrastructure costs are no longer secondary concerns; they represent a primary operational bottleneck as AI-driven inference begins to account for a larger share of total data center power demand, which the IEA projects will double by 2030.