Graphics processing units (GPUs), central to AI models, are high-energy consumers. With increasing GPU use, AI is projected to boost electricity demand by 160% by 2030, according to Goldman Sachs. Vishal Sarin, an experienced circuit designer, sees this as unsustainable and founded Sagence AI to create more energy-efficient alternatives. Sagence focuses on developing analog chips, distinct from the digital norm like GPUs. Unlike digital chips that store data in binary form, analog chips can represent data with various values, allowing for potentially higher data density and faster processing by not needing to transfer data between memory and processors.

These "in-memory" chips can thus complete tasks with less power and latency, offering a viable solution to AI hardware's current limitations. However, analog chips face challenges, such as precision issues due to complex manufacturing and programming difficulties. Sarin views Sagence’s analog chips as complementing digital chips, not replacing them, aiming to enhance server and mobile AI applications. Sagence plans to release their chips by 2025 and is engaging with multiple customers, competing against other AI analog chip firms like EnCharge and Mythic. The company, backed by investors like Vinod Khosla and TDK Ventures, has raised $58 million since its founding and plans another funding round to expand its 75-person team. Despite challenges like existing customer loyalty to companies like Nvidia and broader geopolitical trade barriers, Sagence must demonstrate lower power consumption and higher efficiency in its chips to succeed. This comes as semiconductor startup funding is trending upward, suggesting a promising market environment.