Remember to star this public account ⭐️ to receive updates promptly.
Source：ContentTranslated from Nikkei.
A Japanese startup is designing cost-effective artificial intelligence chips to challenge Nvidia. The company’s CEO stated that they plan to significantly increase sales next year and be “IPO-ready” by 2028.
Experts warn that computing power, transmission capacity, and enormous energy demands will become major bottlenecks for global AI applications. Tokyo-based EdgeCortix believes they can address all these issues with efficient chips that can process AI workloads directly on devices like satellites and smartphones, without relying on data centers powered by Nvidia’s powerful but expensive chips.
“Nvidia does not dominate the entire ‘edge’ market,” said Sakyasingha Dasgupta, co-founder and CEO of EdgeCortix, in an interview with Nikkei Asia. “There is no single winner in this market. Nvidia is not the winner. So it is a completely open competitive environment.”
“Edge” refers to devices that are relatively constrained in size and power, such as portable devices, robots, or satellites. Today’s AI applications typically require massive computing power, which only data centers can provide.
Although American chip giant Nvidia is the preferred designer of chips for heavy AI workloads, experts point out that its bulky and expensive designs may pose challenges for companies looking to introduce this technology in harsh environments or at affordable prices.
Dasgupta illustrated how EdgeCortix’s approach could be applied to satellites. Instead of sending large amounts of data back to Earth for processing, it is more energy-efficient to use AI chips that allow satellites to execute certain applications on board, thereby reducing data transmission and energy consumption.
EdgeCortix is relatively unknown outside the semiconductor industry but recently surprised the industry by announcing a prototype contract with the U.S. Department of Defense’s Defense Innovation Unit. Other companies, including Lockheed Martin Aerospace and Google, have also received similar contracts aimed at developing and integrating military and commercial technologies for better interoperability of land, air, and sea military equipment. The company stated that this is the first time a Japanese semiconductor company has joined this project. The contract amount has not been publicly disclosed.
EdgeCortix has also entered into a strategic partnership with Japanese lunar mission startup ispace to explore low-power AI-guided lunar landing and rover navigation. The chip company is also in talks with “a large aerospace company in the U.S.” that wants to test EdgeCortix chips for running AI on commercial aircraft, indicating that its demand is not limited to defense applications.
Dasgupta declined to disclose customer details but stated that the company expects to expand its business by 2026, focusing on robotics and aerospace applications. Previously, the demand for chips in these areas was low and primarily for research purposes.
The CEO added that with anticipated sales growth, EdgeCortix will be “IPO-ready” within two to three years. He mentioned the idea of listing in the U.S. due to greater liquidity there compared to Japan, but did not provide any details.
EdgeCortix is a fabless chip design company, similar to Nvidia but on a much smaller scale. The Tokyo-based company has raised nearly $100 million through equity and government funding, including investments from Japanese chip manufacturer Renesas Electronics and the investment arm of Japanese financial group SBI Holdings.
Thanks to its efficient design, its chips stand out among traditional AI chips. Its first mass-produced AI chip, Sakura-II, has four times the energy efficiency of Nvidia’s flagship H100 chip at comparable computing power. However, Sakura-II is more focused on serving more specific AI deployment processes, while the H100 is more versatile.
The design of the Sakura chip aims to minimize memory access during data processing. Unlike typical AI chips that frequently access memory to fetch new tasks or results—leading to higher energy consumption and the so-called “memory wall” technical issues—this chip, designed by a former senior researcher at IBM Research, can handle tasks more efficiently, reducing the need for frequent memory access.
“The fundamental reason for our existence is to make AI more sustainable,” Dasgupta said. “Our goal has always been to significantly improve energy efficiency and cost efficiency while minimizing performance loss.”
Reference link
https://asia.nikkei.com/business/technology/japanese-startup-brings-ai-chips-to-space-in-bid-to-rival-nvidia
*Disclaimer: This article is original by the author. The content reflects the author’s personal views, and Semiconductor Industry Observation reproduces it only to convey a different perspective, which does not represent Semiconductor Industry Observation’s endorsement or support of this view. If there are any objections, please contact Semiconductor Industry Observation.
END
This is the 4147th issue shared by Semiconductor Industry Observation, welcome to follow.
Recommended Reading
★A Chip That Changed the World
★U.S. Secretary of Commerce: Huawei’s Chips Are Not That Advanced
★“ASML’s New Lithography Machine, Too Expensive!”
★The Quiet Rise of Nvidia’s New Competitors
★Chip Crash, All Blame Trump
★New Solutions Announced to Replace EUV Lithography!
★Semiconductor Equipment Giants, Salaries Soar by 40%
★Foreign Media: The U.S. Will Propose Banning Software and Hardware Made in China for Cars
Star this account ⭐️ to receive updates promptly, small account to prevent loss
Seeking likes
Seeking shares
Seeking recommendations
