German Startup Semron Introduces Low-Cost, 3D-Scaled Chips for AI on Mobile Devices

Semron's 3D-scaled chips with memcapacitors aim to revolutionize mobile AI.

Semron, a German startup, is making waves with its ground-breaking 3D-scaled chips, which are set to transform artificial intelligence (AI) on mobile devices.

Co-founded by Kai-Uwe Demasius and Aron Kirschen, both from the Dresden University of Technology, Semron's chips boast a unique approach that could reshape how AI models run locally on smartphones, earbuds, VR headsets, and other mobile device (via TechCrunch).

3D-Scaled AI Chips

The magic behind Semron's innovation lies in the utilization of "3D-scaled" chips, which employ electrical fields for calculations instead of the conventional electrical currents found in standard processors.

Aron Kirschen, one of Semron's co-founders, highlights the significance of this shift, emphasizing higher energy efficiency while keeping production costs remarkably low. This could be a game-changer for companies relying on AI computing resources.

"The unique features of our technology will enable us to hit the price point of today's chips for consumer electronics devices even though our chips are capable of running advanced AI, which others are not," Kirschen tells TechCrunch.

What sets Semron apart is its use of a somewhat unconventional component called a "memcapacitor," a capacitor with memory, for running computations. Most computer chips rely on transistors, but Semron's memcapacitors, made of conventional semiconductor materials, tap into the principle of charge shielding.

Energy-Efficient AI Model Training

This unique approach allows for the manipulation of electric fields, controlled by the chip's memory storing the different "weights" of an AI model. The outcome? A significant reduction in energy usage and heat generation at the chip level.

Semron plans to leverage this advantage by stacking as many as hundreds of layers of memcapacitors on a single chip, vastly increasing compute capacity. It is akin to having hundreds of chips seamlessly integrated into one powerful package.

In a 2021 study published in the journal Nature Electronics, Semron, in collaboration with the Max Planck Institute of Microstructure Physics, demonstrated the potential of its technology.

The team achieved energy efficiencies of over 3,500 TOPS/W, a metric indicating a remarkable 35 to 300 times improvement over existing techniques. This suggests that Semron's memcapacitors could usher in a new era of energy-efficient AI model training.

However, Semron is still in the early stages, described by Kirschen as a "pre-product" with minimal revenue. The challenges ahead include the daunting tasks of mass manufacturing and building a substantial customer base.

Despite stiff competition, Semron has attracted attention and investment. With an 11-person workforce poised to grow by approximately 25 individuals by year-end, the startup has secured 10 million euros in funding from investors such as Join Capital, SquareOne, OTB Ventures, and Onsight Ventures.

Stay posted here at Tech Times.

(Photo : Tech Times Writer John Lopez)

ⓒ 2024 TECHTIMES.com All rights reserved. Do not reproduce without permission.
Join the Discussion
Real Time Analytics