Demand is growing worldwide to make artificial intelligence (AI) lighter and more efficient. This trend is showing up in global acquisition battles for AI lightweighting platform corporations. Nvidia acquired OmniML in 2023 and last year bought Deci AI for $300 million (about 442.5 billion won). Qualcomm acquired Edge Impulse in Mar. The acquired corporations are all AI lightweighting and optimization corporations.
There is also a corporation in Korea competing with them. The main player is Nota AI, led by CEO Chae Myeong-su, 36. Meeting with ChosunBiz at the Nota AI office in Samseong-dong, Seoul, on the 12th, Chae said, "There was an acquisition offer from a giant corporation, but we turned it down because our goal is not to serve just one company but to make AI run on every electronic device in the world."
In 2022, Nota AI developed and launched Netspresso, an AI lightweighting and optimization platform that reduces AI models to as little as one-tenth the size and delivers fast inference speeds. That's why Netspresso is nicknamed an AI "diet pill." Nota AI's business is largely divided into two based on Netspresso. The platform business provides SDKs to global semiconductor companies to optimize and lighten AI models or semiconductor chips. A prime example is last month's deal for Nota AI to provide optimization technology for Samsung Electronics' latest application processor (AP), Exynos 2500.
The other pillar, the solutions business, aims to supply customized AI solutions to existing B2B (business-to-business) markets such as transportation, security, and healthcare. An example is Nota AI's contract in Apr. to supply a Generative AI-based intelligent transportation system (ITS) solution to the Dubai Roads and Transport Authority. Through the video control solution provided by Nota AI, when a traffic accident occurs, the authority can easily receive an automated accident investigation report via CCTV.
A month after listing on the KOSDAQ market, Chae said, "These days it's hard to handle all the proposals coming into the company," adding, "Interest in the platform institutional sector has grown since the contract with Samsung Electronics, and in the solutions institutional sector as well, after the contract with the Dubai Roads and Transport Authority, we've been getting a lot of inquiries from Southeast Asia, the United States, and even local governments."
Nota AI was founded in Apr. 2015 by four people, including Chief Technology Officer (CTO) Kim Tae-ho. It began as a company offering a solution to reduce keyboard typos. The company name also came from "no typos." But in the process of developing a typo-reduction solution, privacy issues arose first, followed by heat generation and processing speed problems. Chae recognized that there would be similar demand for optimization and lightweighting technology when running AI on smartphones. So in 2018 the company pivoted into an AI optimization corporation. The workforce grew from four at the time of founding to about 140 now. The following is a Q&A with Chae.
◇ "Trim AI models 50–60% with Netspresso"
─Why didn't you found a startup yourself?
"I met Kim, the CTO, in 2017, after graduating from Soongsil University's Department of Industrial and Information Systems Engineering and earning a master's in knowledge service engineering at KAIST. While performing alternative military service at the KAIST AI Research Institute, my service overlapped with Kim's by a year. Our research chemistry was good. We're also both born in 1989. In Sept. 2018, I was preparing to study for a Ph.D. in Germany. I had a three-month gap, and Kim said Nota AI's business wasn't going well and the three founders had left, and asked me to take over Nota AI. When I joined Nota AI, I was the only employee, and the office was a space of about 13 square meters on the KAIST Daejeon campus. I connected the solution to reduce typos to AI lightweighting and optimization as a way to solve its problems and thought there would be strong demand. Honestly, I had no work experience, so I didn't even think about starting a company myself. In research, my role was to stand at the whiteboard and propose ideas, and Kim's role was to quickly implement my ideas on the computer; that's still how we work today."
─The name "Netspresso" is intriguing.
"Launched in 2022, Netspresso is a technology that reduces AI model size and computation so high-performance AI can run on low-spec devices. Named through an in-house contest, Netspresso means compressing AI models as if extracting only the essence from coffee beans in a neural network. Through Netspresso, corporations can reduce AI model size to under a maximum of 10%. In this case, power consumption drops 30–40%. Typically, AI models are lightened by about 50–60%."
─There are many AI lightweighting corporations. What sets Nota AI apart?
"In AI optimization and lightweighting, know-how is more important than a 'golden rule' about how to write code and achieve results. Like master craftsmen whose work a newcomer can't copy overnight, you need accumulated experience. We are, so to speak, craftsmen who sculpt AI models. Nota AI has experience working with hundreds of models across 37 semiconductors. Our key customers and partners include Samsung Electronics, Qualcomm, ARM, Nvidia, Sony, and Renesas."
◇ Responsibility has grown after the listing… aiming to turn profitable in 2027
─The stock, which started with an offering price of 9,100 won, is nearing 40,000 won.
"Listing on the KOSDAQ has made us more responsible and more cautious in decision-making. Fortunately, the stock has performed well since the listing. When Jensen Huang, Nvidia's CEO, visited Korea this year, at a chicken restaurant in Samseong-dong, Seoul, CTO Kim Tae-ho brought an Nvidia DGX Spark and Thor (THOR) and, in the capacity of a partner, offered congratulations on our listing and got his signature on the products."
─Sales are growing, but operating profit is still in the red.
"Sales have been increasing since the end of 2023 as we began expanding the business in earnest. Costs inevitably rose to build products, but now, aside from personnel expenses, there aren't many major expenses. I think operating losses will improve significantly next year, and we will turn profitable in 2027."
─What are your short- and long-term plans?
"Right now, we discuss AI models individually, such as large language models (LLMs), lightweighting, and optimization. But we expect there will soon be demand for lightweighting and optimization at the AI system level, and we plan to develop related technologies. For example, if we talk about an AI agent, multiple AI models—such as voice and video—will be bundled and operate together. If on-device AI needs lightweighting because of device capacity limits, the cloud will need lightweighting to cut costs. We intend to focus on that."