Google engineer awarded four US patents for AI data center power resilience in 2026

The patents cover distributed power transfer, dynamic anomaly protection, energy control systems, and UPS timing, all designed to keep AI infrastructure running as demand surges.

Google engineering director Krishnanjan Gubba Ravikumar has been granted four US patents in 2026 focused on making AI data center power systems more resilient. Photo Credit: Krishnanjan Gubba Ravikumar

A Google engineering director has been granted four US patents so far in 2026, each focused on making the power systems that underpin AI data centers more resilient, efficient, and fault-tolerant.

Krishnanjan Gubba Ravikumar, Principal Engineer and Director at Google, shared the news on LinkedIn. He leads teams responsible for power and energy systems, controls, microgrids, and cybersecurity across Google's data center infrastructure, supporting both cloud and AI/ML workloads including TPU and GPU compute.

The patents address a set of related problems: how to keep power flowing reliably to AI infrastructure as the scale and density of compute demand continues to accelerate.

Four patents target different failure points in data center power

The first patent covers distributed power transfer, introducing a control scheme for Main-Tie-Tie-Main (MTTM) configurations that uses distributed relay logic to improve reliability during power transitions.

The second addresses dynamic anomaly mitigation, with systems designed to protect AI infrastructure by dynamically adapting thresholds to manage transient events in real time.

The third relates to energy control systems, using integrated sensors to create a feedback loop between data center operational states and real-time energy management.

The fourth covers timing for UPS power transfer, a system that uses real-time modeling of rectifier DC holdup voltages to determine the optimal moment to switch between power sources.

Gubba Ravikumar framed the work in the context of the current moment for power engineering: "The rapid evolution of AI presents unique challenges for power infrastructure, making it an incredibly exciting time to be a power and electrical engineer."

Patents build on a career spanning 30-plus inventions

Gubba Ravikumar holds a PhD in Electrical and Computer Engineering from Washington State University and a Mini-MBA from Rutgers Business School. Before joining Google in late 2020, he spent 12 years at Schweitzer Engineering Laboratories, where he rose from associate power engineer to principal engineer, leading teams of up to 60 engineers, project managers, and technicians on mission-critical protection and control systems.

He holds more than 30 patents, publications, and awards across electric grid applications and data center technologies. His co-inventors on the four 2026 patents are Hariharan Subramanian, Electrical Engineering Manager at Google; Kei Hao; Sangsun Kim; and Dr. Hammad Khan, a senior electrical engineer with a background spanning mining, renewable energy, and mission-critical systems.

As he put it on LinkedIn: "These grants represent another step toward making large-scale critical infrastructure more resilient through distributed intelligence." With AI compute demand driving unprecedented power density requirements across the data center industry, the patents reflect a growing recognition that the engineering keeping the lights on is becoming as strategically important as the models running on top of it.