November 19, 2024
The majority of AI models are powered by graphics processing units (GPUs), which are notoriously energy-intensive. According to Goldman Sachs, by 2030, the increasing use of GPUs in data centers would result in a 160% rise in the need for electricity. Vishal Sarin, a seasoned designer of analog and memory circuits, established Sagence AI, originally known as Analog Inference, after seeing that this tendency was unsustainable. The business focuses on developing GPU substitutes that consume less energy, tackling issues with both performance and the environment.
Sagence creates systems and analog chips for AI applications, as well as the software needed to program them. Analog chips use a range of values to represent information, in contrast to standard digital chips that store data as binary values. Because of this distinction, analog circuits may process data directly in memory and complete some computations using fewer parts, removing bottlenecks brought on by data transfers between processors and memory. Furthermore, compared to their digital counterparts, analog processors provide a higher data density.
Despite being available for decades, analog technology is making a comeback because digital processors can't keep up with the demands of the modern world. However, there are drawbacks to analog chips as well, namely the need for exact manufacture and the complexity of programming. Sagence markets its chips as an adjunct to digital technologies, focusing on certain server and mobile device applications to solve problems like latency, cost, and power consumption.
Sagence, which plans to commercialize its chips in 2025, is already collaborating with clients and facing competition from other analog chipmakers that specialize in AI, such as EnCharge and Mythic. Investors such as Vinod Khosla, TDK Ventures, and Aramco Ventures have contributed $58 million to the company, and it intends to seek more money to grow its staff of 75 employees. Sagence's cost-effective approach, which stays away from the newest production techniques, increases its allure in a cutthroat industry.
Potential opportunities for Sagence are indicated by recent trends. The amount of venture capital money raised for semiconductor startups is increasing; in the first half of 2024 alone, $5.3 billion was raised. There are still obstacles to overcome, though, such as rivalry from well-known firms like Nvidia, expensive fabrication, and the problem of entering markets where established technologies predominate. The hazards are highlighted by the case of Graphcore, an AI chipmaker that declared bankruptcy in spite of substantial funding.
Sagence's success depends on its ability to deliver energy-efficient, high-performance chips while efficiently scaling production. The company aims to uniquely position itself in the progressing AI hardware landscape by addressing both economic and environmental challenges.
Drive innovation through data with Code Labs Academy’s Data Science & AI Bootcamp.