Luxor Energy and Bentaus Curtail AI GPU for ERCOT 4CP Signal

Luxor Energy and Bentaus announced in a press release that Bentaus's Ziani platform reduced a live AI inference GPU's power draw to 25% in under 500 milliseconds in response to a Texas 4CP curtailment signal. The companies said the test used signals from Luxor's Qualified Scheduling Entity and completed the curtailment in a closed loop without interrupting inference workloads.

The test was tied to the Four Coincident Peak program run by Electric Reliability Council of Texas (ERCOT). Under 4CP, transmission costs are allocated based on electricity use during the four highest 15 minute demand peaks from June through September.

Bentaus said Ziani can reduce GPU power use while inference workloads checkpoint, pause, and resume. Luxor Energy and Bentaus said they are now working to enroll the technology in additional grid programs, including ERCOT's Emergency Response Service, and to run it across multiple generations of AI GPUs.