Graphics processing units (GPUs), the chips on which most AI models run, are power-hungry beasts. As GPUs are increasingly incorporated into data centers, artificial intelligence will increase electricity demand by 160% by 2030, Goldman Sachs estimates.

This trend just isn’t sustainable, says Vishal Sarin, an analog circuit and memory designer. After greater than a decade within the chip industry, Sarin launched Sagence AI (previously called Analog reasoning) to design energy-efficient alternatives to GPUs.

“Applications that could make practical AI computing truly ubiquitous are limited because data-processing devices and systems cannot achieve the required performance,” Sarin said. “Our mission is to break through the constraints of efficiency and economics in an environmentally friendly way.”

Sagence develops chips and systems to run AI models, in addition to software to program those chips. While there isn’t a shortage of corporations creating custom AI hardware, Sagence is somewhat unique in that its chips are analog, not digital.

Most chips, including graphics processors, store information digitally as binary strings of zeros and ones. In contrast, analog chips can represent data using a variety of various values.

Analog chips usually are not a brand new concept. Their heyday was from 1935 to 1980, helping, amongst other things, to model the North American electrical grid. However, the shortcomings of digital chips make analog solutions attractive again.

First, digital chips to require a whole lot of components to perform certain calculations that analog circuits can perform with just a number of modules. Digital chips typically need to transfer data forwards and backwards from memory to processors, which causes bottlenecks.

“All of the leading legacy AI silicon vendors use this old architectural approach, which is blocking progress in AI implementation,” Sarin said.

Analog chips like Sagence, that are “in-memory” chips, don’t transfer data from memory to processors, potentially allowing them to perform tasks faster. And by having the ability to use a variety of values ​​to store data, analog chips can provide higher data density than their digital counterparts.

However, analog technology has its drawbacks. For example, achieving high precision with analog chips will be harder because they require more precise manufacturing. They are also normally harder to program.

However, Sarin believes Sagence’s chips complement, not replace, digital chips, for instance to speed up specialized applications in servers and mobile devices.

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

Sagence, which plans to bring its chips to market in 2025, is working with “multiple” customers because it looks to compete with other analog AI chip corporations akin to EnCharge and Mythic, Sarin said. “We are now packaging our core technology into system-level products and making sure we fit into existing infrastructure and deployment scenarios,” he added.

Sagence has secured investments from backers including Vinod Khosla, TDK Ventures, Cambium Capital, Blue Ivy Ventures, Aramco Ventures and New Science Ventures, raising a complete of $58 million within the six years since founding.

Now the startup plans to raise capital again to expand its 75-person team.

“Our cost structure is favorable because we do not seek to achieve performance goals by migrating to the latest (manufacturing processes) of our chips,” Sarin said. “This is an important factor for us.”

The timing may be in Sagence’s favor. For Crunch BaseFunding for semiconductor startups appears to be returning after a weak 2023. From January to July, VC-backed chip startups raised nearly $5.3 billion — significantly greater than last yr, when such corporations reported a complete of slightly below $8.8 billion.

In this environment, chip production is an expensive proposition, made even harder by international sanctions and tariffs promised by the incoming Trump administration. Acquiring customers who’re “stuck” in ecosystems like Nvidia is one other uphill climb. Last yr, AI chipmaker Graphcore, which raised nearly $700 million and was once valued at nearly $3 billion, filed for bankruptcy after struggling to gain a powerful foothold out there.

To have any probability of success, Sagence will need to prove that its chips actually devour significantly less power and supply higher performance than alternatives, and that it may well raise enough enterprise capital funding to have the ability to produce at scale.