Sagence builds analog chips to run AI

Graphics processing units (GPUs), the chips on which most AI models run, are power-hungry beasts. As a result of accelerating the installation of GPUs in data centers, AI will drive a 160% increase in electricity demand by 2030, Goldman Sachs estimates.

This practice is not sustainable, says Vishal Sarin, an analog circuit and memory designer. After working in the chip industry for more than a decade, Sarin launched Sagence AI (formerly known as Analog Inference) to design power-saving alternatives to GPUs.

“The applications that can make AI computing truly widespread are limited because the devices and data processing systems cannot achieve the required functionality,” Sarin said. “Our goal is to break operational and economic limits, and in an environmentally responsible way.”

Sagence develops chips and systems to implement AI models, as well as software to program these chips. While there’s no shortage of companies building custom AI hardware, Sagence is somewhat unique because its chips are analog, not digital.

Many chips, including GPUs, store information digitally, such as their binary bits and zeros. In contrast, analog chips can represent data using a range of different values.

Analog chips are not a new concept. They had their heyday from 1935 to 1980, helping model the North American electric grid, among other engineering feats. But the limitations of digital chips make analog attractive again.

For one, digital chips require hundreds of components to perform certain calculations that analog chips can achieve with just a few modules. Digital chips often have to transfer data back and forth from memory to processors, which causes bottlenecks.

“All of the leading legacy AI providers are using this archaic architecture, and this is hindering the progress of AI adoption,” Sarin said.

Analog chips like Sagence’s, which are “memory” chips, don’t transfer data from memory to processors, which may be able to complete tasks faster. Also, because of their ability to use a range of values ​​to store data, analog chips can have higher data density than their digital counterparts.

Analog tech has its downsides, though. For example, it can be difficult to achieve high accuracy with analog chips because they require more precise manufacturing. They also tend to be difficult in plans.

But Sarin sees Sagence’s chips as a complement to — not a substitute for — digital chips, for example, to speed up specialized applications on servers and mobile devices.

“Sagence products are designed to eliminate the power, cost and latency issues inherent in GPU hardware, while delivering high performance for AI applications,” he said.

Sagence, which plans to bring its chips to market by 2025, is working with “a lot” of customers as it looks to compete with other AI chip companies such as EnCharge and Mythic, Sarin said. “We are currently packaging our core technology into system-level products and ensuring that we fit into existing infrastructure and deployment environments,” he added.

Sagence has received investment from backers including Vinod Khosla, TDK Ventures, Cambium Capital, Blue Ivy Ventures, Aramco Ventures and New Science Ventures, raising a total of $58 million in the six years since its founding. .

Now, the startup is planning to raise money again to expand their team of 75 people.

“Our costs are good because we don’t rush the goals of working with the migration of young people [manufacturing processes] with our chips,” said Sarin. “That’s a big factor for us.”

The timing may work in Sagence’s favor. Per Crunchbase, funding for semiconductor startups looks set to bounce back after the 2023 deficit. From January to July, VC-backed chips raised about 5.3 billion dollars – a figure ahead of last year, when such firms saw less than 8.8 billion dollars. value.

If so, chipmaking is an expensive proposition — made more difficult by international sanctions and tariffs promised by the incoming Trump administration. Winning customers “locked in” to ecosystems like Nvidia is another uphill battle. Last year, AI chipmaker Graphcore, which raised nearly $700 million and was once valued at $3 billion, filed for bankruptcy after struggling to gain a foothold in the market.

To have any chance of success, Sagence will have to prove that its chips, in fact, draw much less power and deliver higher efficiency than alternatives – and raise enough business funding to scale.