GPUs, essential for running AI models, consume large amounts of energy, which will lead to a 160% increase in electricity demand by 2030, according to Goldman Sachs.Vishal Sarin, a designer of analog circuits and memory, founded a company, Sagence AI, to create more efficient alternatives to this problem, as shown by TechCrunch.
The startup develops analog chips, which, unlike digital chips, can represent data with a range of values and do not require data transport between memory and processors, reducing energy consumption and improving speed.
Although analog chips were popular from 1935 to 1980, their energy efficiency is being reconsidered due to the limitations of digital chips, such as the need for multiple components and memory bottlenecks.
read more
Chip War: Japan has a plan to turn the tide shutterstock/riopatuca The difference between Sagence chips Sagence chips, called “in-memory chips”, can perform calculations with fewer modules and greater data density, providing excellent performance and lower latency for AI applications offer . Despite their advantages, analog chips are more difficult to manufacture and program with high precision. Sarin believes Sagence's chips complement digital chips, which accelerate specific applications on servers and mobile devices. The company, which intends to launch its chips in 2025, has already secured US$58 million in investments and is preparing to expand its team. With the growth of semiconductor startups and the growing demand for AI solutions, Sagence could benefit, as long as it proves the superiority of its chips in terms of energy consumption and efficiency.
Sagence chips are expected to be launched in 2025 – Image: 3dartists / Shutterstock.com