Energy Consumption as a Critical Bottleneck
Scaling artificial intelligence faces a challenge that goes beyond pure technical performance: massive energy consumption. As more enterprises seek to deploy AI projects to production, computational and electricity costs have become one of the primary economic obstacles. Naveen Rao, who led AI efforts at Databricks, identified in this problem an opportunity to reimagine computing architecture from the ground up.
Unconventional AI and Oscillator-Based Architecture
Through his company Unconventional AI, Rao is developing a radical solution: replacing conventional chip architecture with an oscillator-based design. In June 2026, the company unveiled Un-0, its first image-generation model serving as proof of concept. "This is the 'hello world' of a new kind of computer," Rao told TechCrunch.
The significance lies not only in the output—Un-0 generates images with quality comparable to Stable Diffusion or GPT Image 1—but in how it achieves this. The model was built on a software simulation of Unconventional's oscillator-based chips, an architecture completely different from conventional systems and traditional LLMs. According to the company's research team, this approach could reduce power consumption up to 1,000 times.
From Software to Real Infrastructure
While the current version of Un-0 runs on a software simulation of Unconventional's chips, the team plans to release schematics for actual hardware soon. The ultimate goal is to build a complete inference stack from scratch, allowing the company to provide compute capacity like any other cloud provider, but with unprecedented efficiency.
"We will build a new kind of system composed of our chips," Rao explained. "We will run AI models there, and we will have a network cable where prompts come in and inferences go out, but it'll be done at 1/1000 of power."
This is particularly ambitious considering Unconventional AI has fewer than 50 employees. However, given the scale of current AI deployments and anticipated costs to meet growing inference demand, few projects are positioned to address a problem of this magnitude.
Implications for Technology Leaders
For technology teams evaluating AI solutions, the direction signaled by initiatives like Un-0 indicates a structural shift ahead. Energy is not just another operational cost: it is the physical limit that will define what is feasible and what is not in the coming years. A drastic reduction in consumption could transform the economics of projects that today seem prohibitively expensive, democratizing capabilities currently reserved for large cloud providers.
Rao is clear about the scale of the challenge: "AI scaling is hard because of energy. It's going to be the fundamental limit in the next few years. You just can't go past it. It's going to be an energy-limited problem, at the end of the day." For enterprises seeking to adopt AI at scale, staying alert to innovations in hardware architecture and energy efficiency is not optional—it is strategic.