Graphics processing units, the silicon on which most artificial-intelligence models run, are hogging energy. Goldman Sachs, citing the accelerating pace of incorporating GPUs in data centers, thinks AI will drive electricity demand up 160% by 2030.
It's unsustainable," argues Vishal Sarin, analog and memory circuit designer. Sarin spent over a decade in the chip industry and recently founded Sagence AI (previously known as Analog Inference) to design energy-efficient analog counterparts to GPUs.
"The applications that could make practical AI computing truly pervasive are limited because the devices and systems processing the data cannot meet the necessary performance," Sarin said. "Our mission is to break the performance and economics limitations, and in an environmentally responsible way."
Sagence designs chips and systems that implement AI models; the company also produces software to run on those chips. Although many companies build custom AI hardware, Sagence's chips are distinct from others in that they operate analogically, rather than digitally.
While all chips, and indeed even GPUs, store information digitally, as sets of binary strings of ones and zeros, analog chips represent data using a range of different values.
Analog chips aren't anything new. They had their heyday from about 1935 to 1980, helping model the North American electrical grid, among other engineering feats. But the drawbacks of digital chips are making analog attractive once again.
But digital chips must use hundreds of components to perform calculations that analog chips can accomplish in just a few modules. Moreover, digital chips usually cannot store enough data for processing close to the processors themselves, which can cause bottlenecks.
"All the leading legacy suppliers of AI silicon use this old architectural approach, and this is blocking the progress of AI adoption," Sarin said.
Analog chips like Sagence’s, which are “in-memory” chips, don’t transfer data from memory to processors, potentially enabling them to complete tasks faster. And, thanks to 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 one, analog chips tend to be more difficult to get to high precision because they are more sensitive to manufacturing imperfections. They also prove more challenging to program.
Still, Sarin sees Sagence's chips filling out a new spot on the datasheet: not displacing digital chips, perhaps, but speeding specialized applications in 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 introduce its chips to the market in 2025, is already working with "multiple" customers as it looks to compete with other AI analog chip ventures like EnCharge and Mythic, Sarin said. "We're currently packaging our core technology into system-level products and ensuring that 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, which have generated a total of $58 million for the startup in the six years since its founding.
The startup is now planning to raise capital again to expand its 75-person team.
“Our cost structure is favorable because we’re not chasing the performance goals by migrating to the newest [manufacturing processes] for our chips,” Sarin said. “That’s a big factor for us.”
Working in Sagence's favor might be the timing. Funding to semiconductor startups seems to be bouncing back after a lackluster 2023, according to Crunchbase. VC-backed chip startups raised nearly $5.3 billion from January through July, which is well ahead of last year when such firms saw less than $8.8 billion raised in total.
This is expensive — not only because of international sanctions and tariffs promised by the new Trump administration — but because it's hard to win customers who have become "locked in" to ecosystems like Nvidia's. Last year, AI chipmaker Graphcore, which raised almost $700 million and was once valued at close to $3 billion, filed for insolvency after struggling to gain a strong foothold in the market.
Sagence must demonstrate that indeed, its chips draw dramatically less power and deliver more efficiency than alternatives: and gather enough venture funding to fabricate at scale if it is to have any hope of success.