[ad_1]
Graphics processing units (GPUs), the chips that most AI models run on, are power-hungry beasts. As a result of the accelerated integration of graphics processing units into data centers, artificial intelligence will contribute to a 160% increase in electricity demand by 2030, according to Goldman Sachs. Estimates.
This trend is unsustainable, says Vishal Sareen, an analog memory circuit designer. After working in the chip industry for more than a decade, Sarin launched Sagence AI (formerly Analogical reasoning) to design energy-efficient alternatives to GPUs.
“The applications that could make practical AI computing widespread are already limited because the devices and systems that process the data cannot achieve the required performance,” Sarin said. “Our mission is to break the constraints of performance and economics, in an environmentally responsible way.”
Sagence develops chips and systems to run artificial intelligence models, as well as software to program these chips. Although there is no shortage of companies creating dedicated AI hardware, Sagence is somewhat unique in that its chips are analog rather than digital.
Most chips, including graphics processing units, store information digitally, as binary strings of ones and zeros. In contrast, analog chips can represent data using a range of different values.
Analog chips are not a new concept. They reached their peak from about 1935 to 1980, helping to model the North American electrical grid, among other engineering work. But the drawbacks of digital chips make analog attractive again.
First, digital chips requires Hundreds of components to perform certain calculations that analog chips can accomplish using only a few modules. Digital chips usually also have to transfer data back and forth from memory to processors, causing bottlenecks.
“All the leading legacy suppliers of AI silicon are using this outdated architectural approach, and this is holding back progress in AI adoption,” Sarin said.
Analog chips like Sagence chips, which are “in-memory” chips, do not transfer data from memory to processors, potentially enabling them to complete tasks faster. Thanks to their ability to use a range of values to store data, analog chips can have higher data density than their digital counterparts.
However, analog technology has its downsides. For example, it may be more difficult to achieve high accuracy with analog chips because they require more precise manufacturing. They also tend to be more rigorous in programming.
But Sarin sees Sagens chips as complementing — not replacing — digital chips, for example, to speed up 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 bring its chips to market in 2025, is taking on “multiple” customers as it looks to compete with other AI-powered analog chip projects such as EnCharge and Mythic, Sarin said. “We are currently packaging our core technology into system-level products and ensuring they 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 total of $58 million in the six years since its founding.
Now, the startup plans to raise capital again to expand its 75-person team.
“Our cost structure is appropriate because we are not seeking to achieve performance targets by moving to newer (manufacturing processes) for our chips,” Sarin said. “That’s a big factor for us.”
The timing may work in Saginis’ favor. per CrunchbaseSemiconductor startup funding appears to be picking up again after a lackluster 2023. From January to July, venture capital-backed chip companies raised nearly $5.3 billion — a figure that far exceeds last year, when these companies were seen raising less than $8.8 billion in total.
In this case, making chips is an expensive proposition — made more difficult by international sanctions and tariffs promised by the incoming Trump administration. Winning over customers who have become “locked in” in ecosystems like Nvidia’s is another tall order. Last year, AI chip maker Graphcore, which had raised nearly $700 million and was once valued at nearly $3 billion, filed for bankruptcy after struggling to gain a strong foothold in the market.
To have any chance of success, Saginis will have to prove that its chips actually consume significantly less power and offer higher efficiency than alternatives — and raise enough project funding to manufacture them at scale.
[ad_2]