“We’re in a business where we can make money right now. While it is important to develop advanced and advanced technologies, it is also important to preempt markets where there is demand. For the first time in fabless, we’re going to show you meaningful sales right away.”
When it comes to ‘fabless’ in the domestic semiconductor market, there is a preconceived atmosphere of looking at it with some preconceived notions. It is true that some fabless companies, which were evaluated as having a promising future, suffered an earnings shock after going public, and an implicit scarlet letter was carved in the market.
Soteria almost became a victim of “fabless trauma.” Recently, however, such an environment has been seen as a good opportunity to further differentiate the company’s capabilities.
Large-scale sales are expected to start this year. It will be the first to break the preconceived notion that ‘domestic semiconductor fabless companies cannot make sales’. Right now, sales that are expected to be recognized from the second quarter of this year to the first half of next year are worth 200 billion won.
Soteria CEO Kim Jong-man (pictured) focused on the business that will generate immediate sales based on the technical philosophy (pragmatism) he has inherited since he was a professor at Georgia Tech (Georgia Institute of Technology).
CEO Kim said, “Innovation that is biased towards advanced technology and only pursues dreams is only ‘innovation in the laboratory’. I felt that commercialization was another area.” He said, “Entrepreneurship is not just about technology. We had to find the point of contact between the market and the customer and make a product that considered the mass production stage, but it was actually difficult in the school.”
This is why Mr. Kim left his job as a university professor and plunged into the ruthless world of entrepreneurship. “One of TSMC’s policies is to ‘lower the bar for innovation,’ which means that only really practicable and practical innovations can serve customers,” he said. I wanted to get out of this and challenge myself in an area where I could actually do mass production and commercialization, and that was the starting point of my start-up.”
His philosophy leads directly to Soteria’s main business. CEO Kim said, “Isn’t our business ASICs (semiconductors on demand)? “We are deep tech, but the reason for our business is to make a lot of sales through sales. It’s about understanding customer needs, marketing well, and thinking about pain points, but customer engagement is the key to that.”
Soteria’s flagship MIK-100 chip (project name: Artemis) is the first result. It is an ultra-low-power, high-end computing (HPC) accelerator. It has a Near Threshold Voltage (NTV) design that can drive the accelerator at 0.3V. It also supports the shortest path technology for heat dissipation and the Bad Core Management System for core management during immersion cooling. The 4nm single tape-out (switching to the production stage after chipset design) through Samsung Electronics’ foundry is the first case of fabless in Korea.
The MK-100, which was created with the market and customer needs in mind from the beginning, has led to contracts with global data center customers from the development stage. Sales recognition according to the 200 billion won letter of intent (LOI) received last year will be recognized sequentially from the second quarter at the earliest.
Kim is already looking at the next stage. We started the development of the ‘LLM Accelerator’. After all, it is an item that can be mass-supplied to the market as soon as the development stage is completed. It is a technology derived from discussions with global customers from the initial stage.
Kim said, “NVIDIA’s AI· It is a chip that is installed in the GPU infrastructure and maximizes data processing efficiency while assisting,” he said, adding, “In short, it plays a role in increasing the cost-effectiveness of the GPU at a low cost. It can eliminate the shortcomings of existing GPUs that can appear in a more advanced AI environment, such as power consumption and bottleneck.”
He added, “In the early days of AI, high-spec GPUs are making a splash, but now that AI is becoming commonplace and popular, cost-effective systems will be considered,” adding, “It is a system that can solve the frustration of a market that is in short supply with technology that complements NVIDIA’s low-end products, and it is the most effective system that can respond to cost-effective AI semiconductor infrastructure in the era of AI popularization.”
Chairman Kim’s mid- to long-term goals are also in line with “pragmatism.” “We don’t dream far into the future,” he said, adding, “Our market is already a big market with a lot of competitors, but if you think about it the other way around, it means that you can make sales right away. We’re going to be a money-making fabless.”
Soteria is a startup founded in 2018 by former Georgia Institute of Technology professor Kim Jong-man and researchers from Samsung Electronics, SK Hynix, Intel, and Apple. Along with the surge in corporate value, it is also preparing to go public. NH Investment & Securities and Hana Securities will be selected as co-sponsors and plans to submit a request for preliminary examination for listing in the first half of next year at the latest.
https://www.thebell.co.kr/free/content/ArticleView.asp?key=202502181954372480101139
