The domestic system semiconductor startup ecosystem is dreaming of a leap forward thanks to the increase in semiconductor-based demand and the AI craze. At the System Semiconductor Symbiosis Forum held for the second time this year, many deep tech companies participated and exchanged technology and vision.

On the 17th, the 2nd Sangsaeng Forum 'Deep Tech Open Innovation Day' of 2023 was held at the SKY 31 Convention in Lotte World Tower, Jamsil. This event, hosted by Seoul National University’s System Semiconductor Industry Promotion Center (SIPC), brought together deep-tech startups that will lead future new growth fields.

△Semiconductor fabless companies and various cutting-edge IT solution startups in the fields of system semiconductors, future mobility, biohealth, robots, and eco-friendly energy gathered to promote convergence between different technologies and announce the latest technologies.

Professor Choi Chang-ki of Seoul National University said, “In order for early startups to create new value and open up new markets, startups with early market technologies should come together and seek to expand the market,” and added, “This Symbiosis Forum is an event designed to create opportunities in new markets.”

Center Director Lee Hyuk-jae said, “Seoul National University SIPC has been supporting startups in the system semiconductor field,” and “This time, we have created a place where startups from five fields can come together to introduce and merge different technologies.”

The event featured many companies with AI technology capabilities, and among them, AI solutions and computer vision solutions suitable for edge environments were introduced.

■ MobilinTr, “Preparing for the upcoming edge AI market”
Experts and industry insiders predict that at present, AI semiconductors are taking up a large share of the market with AI server chips for data centers, but after the big cycle of AI chips from big tech companies and cloud service providers, demand for AI chips for AI inference or edge computing will follow.

Shin Dong-joo, CEO of Mobilint, said, “We focused on the fact that legacy solutions including CPUs and GPUs are not yet ready for edge AI,” and Mobilint was spurring the development of cost-effective AI chip solutions targeting edge AI in the growing AI semiconductor market.

Mobilint is a fabless startup that provides AI solutions that can be processed in embedded environments such as robots, CCTVs, and smart cities. It focuses on advancing edge AI, which most companies are still at the demonstration level, to a level where it can be mass-produced as a commercial product.

Mobilint's AI accelerator is preparing to enter the market with a cost-effective chip that can improve the trade-off relationship between performance, power efficiency, and price.

CEO Shin said, “We have developed an edge AI chip that is 3.3 times more powerful and consumes less than 1/5 the power at 1/3 the price of NVIDIA’s 3 million won GPU,” adding, “We believe that we need to demonstrate a competitive edge that is significantly superior to NVIDIA’s GPUs in order to enter the market.”all.

The Mobilint Edge AI chip has already been produced as a prototype, and verification tests are scheduled to take place in Anyang, Pangyo, and other locations in the second half of this year, replacing Nvidia GPUs. It comes in two form factors: a card-shaped chip and a stand-alone AI box, and it was emphasized that it can process various algorithms targeting the edge field with high accuracy.

“The point to pay attention to is how much we can increase programmability while maintaining power efficiency,” said CEO Shin, emphasizing that it is important to increase the values of these factors that have a trade-off relationship.

■ DeltaX, “Computer Vision Camera Replacing Lidar”
DeltaX is an artificial intelligence solution company that researches and develops computer vision. It provides cognitive solutions for applications such as robots and autonomous driving through computer vision technology applied to cameras. DeltaX, which is comprised mostly of master's and doctoral degree holders majoring in computer engineering, has demonstrated its outstanding track record by winning first and second place in domestic and international artificial intelligence competitions and has proven its technological prowess through patents.

CEO Kim Soo-hoon said, “Since automobiles are applications that operate in edge environments, optimized and lightweight design for fast data processing at the edge is key,” and explained, “DeltaX mostly processes data through its own algorithms.”

Most autonomous vehicles map and recognize their surroundings using lidar, which is a factor driving up costs. “Even with a single camera, depth measurement is excellent, and data from about 29 to 30 can be processed in a 30fps video,” said CEO Kim, adding that a model that achieves 95% accuracy compared to lidar in a specific ROI area has also been implemented.

In this way, a camera solution that can replace LiDAR can provide advantages to manufacturers by not only recognizing the surroundings, but also mapping the surrounding space like LiDAR by reconstructing a three-dimensional structure.

DeltaX disclosed that it is conducting 12 projects with Hyundai Motor Company, Kia Motors, and Hyundai Mobis, and is currently building a system to monitor the safety of drivers and passengers in vehicles, with the goal of mass production in 2026.

CEO Kim said, “DeltaX, which has received seed investment so far, will proceed with additional investment from Hyundai Motors and other companies starting this year, as well as fund raising from SIFIs (Systemically Important Financial Institutions),” and predicted that DeltaX’s sales performance will also become visible when full-scale mass production is carried out by customers in 2026.