Sapeon Unveils AI Semiconductor 'X330' Four Times Faster... Targeting Data Centers

4x Computing Performance and 2x Power Efficiency Compared to Previous Model
Mass Production to Begin in First Half of Next Year

Global AI semiconductor design company (fabless) Sapeon has launched the data center AI semiconductor 'X330,' which delivers four times the computing speed of its predecessor and supports large language models (LLM).

Sapeon unveiled the X330 on the 16th at the SK Tech Summit held at COEX in Gangnam-gu, Seoul.

Ryu Su-jeong, CEO of Sapeon, is introducing the X330.
[Photo by Sapeon]

AI semiconductors are chips that execute AI learning and inference at ultra-high speed and low power consumption. They can be considered the brain of AI. While most companies implement AI services using graphics processing units (GPUs), these are expensive and consume a lot of power, making AI semiconductors an emerging solution. The X330 is an inference neural processing unit (NPU) that offers four times the computing performance and more than twice the power efficiency compared to its predecessor, the X220. According to the company, it outperforms competitor products released this year by about two times in computing performance and 1.3 times in power efficiency. Sapeon CEO Ryu Su-jeong said, "Simply put, the computations that previously required four X220 units can now be handled by a single X330," adding, "Its high performance per watt reduces the chronic issues of power costs and carbon emissions in data centers."

Support for LLM has been added, demonstrating excellent performance on transformer-based LLMs, the core technology behind conversational AI like ChatGPT. Additionally, the application range has been significantly expanded compared to the previous model, enabling use across various industries.

Sapeon plans to accelerate its targeting of AI service model development companies and data centers through the cost-efficient X330. After conducting prototype testing and reliability verification with key customers, mass production will begin in the first half of next year.

Sapeon X330 [Photo by Sapeon]

Sapeon is particularly focusing on the inference market. The global AI data server market is growing at an average annual rate of 31%. Within this, the inference market is growing at an average annual rate of 32%, which is 5 percentage points higher than the training market. CEO Ryu stated, "We are confident that we will establish ourselves as a leader who changes the trend in a market currently dominated by GPUs and CPUs."

Furthermore, alongside the newly released X330, Sapeon plans to introduce AI NPUs for high-performance edge devices such as autonomous vehicles and CCTV. The autonomous driving specialized model X340 is scheduled for release next year. In collaboration with SKT, Sapeon plans to provide voice phishing defense and spam/smishing prevention systems through AI and big data analysis using AI semiconductors.

Sapeon is an AI semiconductor fabless company spun off from SKT, aiming at the global market. The X330 is the first product released since the spin-off from SKT last year. CEO Ryu Su-jeong said, "We continue to enhance our competitiveness by figuring out how to operate and optimize the numerous AI solutions owned by the SK Group on Sapeon chips," adding, "We will create a virtuous cycle where partner companies’ solutions running on Sapeon chips can achieve lower operating costs and higher competitiveness."