South Korean Semiconductor Firm Panmnesia Successfully Secures $12.5 Million, Achieving a Valuation of $81.4 Million

Panmnesia , a South Korean chip startup, has  secured The seed funding round of $12.5 million, led by Daekyo Investment, included significant financial backing from respected organizations such as SL Investment, Smilegate Investment, GNTech Venture Capital, Time Works Investment, Yuanta Investment, and Quantum Ventures Korea.

Focusing on the creation of intellectual property for CXL technology, the firm aims to revolutionize extensive data centers by allowing the integration of multiple devices like AI accelerator chips, processors, and memory resources.

According to the company, memory constraints in AI processing have posed significant hurdles for data center managers. Panmnesia aims to tackle this challenge through its CXL technology, which boosts efficiency by pooling devices.

As Panmnesia CEO Myoungsoo Jung pointed out to Reuters, 'Without a standardized interface, accommodating the diverse range of AI accelerators in data centers remains quite a challenge.'

Notable companies specializing in chip design software, including Synopsys and Cadence Design Systems, have also shown a keen interest in the innovative technology offered by Panmnesia.

Panmnesia's latest funding follows an ambitious initiative by the South Korean government aimed at strengthening the national AI chip industry, with plans to allocate $800 million over the next five years. The goal is to boost the domestic share of AI chips in data centers from almost negligible levels to an impressive 80% by 2030.

Introducing Panmnesia’s AI Acceleration System

In April, Panmnesia revealed The cutting-edge AI acceleration solution dubbed TrainingCXL leverages CXL technology to meet the increasing demands presented by memory-intensive AI applications like ChatGPT.

TrainingCXL provides an enormous memory capacity, adaptable performance, and dramatically reduced training times, representing a transformative approach for AI applications.

According to Jung, the system can scale memory capacity up to an astonishing 4PB (petabytes), with training execution times slashed by a factor of 5.3 compared to traditional systems that depend on PCIe interconnects.

As industry giants such as Meta and Microsoft strive to expand their AI models for competitive advantage, the need for robust computing systems capable of managing large-scale AI has surged immensely.

Read more:

• AMD Introduces New Chipset Aimed at Generative AI to Compete with Nvidia
• Five Reasons Why Nvidia is Poised to Achieve a $2 Trillion Market Cap by 2030
• NeuRRAM: Researchers Have Developed Cutting-Edge Neuromorphic Chips Tailored for AI Applications
• Co-Founder of Google DeepMind Advocates for the Responsible Use of Nvidia’s AI Chips in the United States.