Databricks' former AI chief unveils 1,000x power-efficient AI architecture
Original: Databricks’ former AI chief thinks he can cut AI’s power bill by 1,000x
Why This Matters
Energy efficiency in AI inference is becoming a critical bottleneck for industry scaling and cost management.
Naveen Rao, former head of AI at Databricks, founded Unconventional AI to reduce AI inference power consumption by up to 1,000 times using oscillator-based computing architecture. The company released its first model, Un-0, an image-generation system demonstrating comparable performance to Stable Diffusion.
Unconventional AI, led by Naveen Rao, unveiled its first AI model Un-0 on June 25, 2026, showcasing a novel oscillator-based computer architecture designed to dramatically reduce power consumption in AI inference processing. The company released a research paper detailing how they built a fully functional image-generation model using software simulation of the new architecture, achieving performance equivalent to state-of-the-art diffusion models like Stable Diffusion and OpenAI's GPT Image 1. Rao stated the oscillator-based computing approach could ultimately reduce power use by as much as 1,000 times compared to conventional systems. Currently, Un-0 runs on a software simulation of Unconventional's oscillator chips. The company plans to release actual chip schematics soon and build a complete inference stack from the ground up, eventually offering compute capacity with 1/1000th of traditional power consumption. With fewer than 50 employees, Unconventional AI is pursuing an ambitious goal. Rao emphasized that energy constraints will become the fundamental limit for AI scaling in coming years, positioning power efficiency as a critical challenge the industry must address.