©2026 Business Next Media Corp. All Rights Reserved. No.102, Guangfu S. Rd., Da'an Dist., Taipei City 106, Taiwan
Kneron Unveils KL1140 and Full Chip Roadmap, Marking Its Transformation into an AI Infrastructure Company
AI chip company Kneron officially launched its next-generation flagship KL1140 NPU (Neural Processing Unit), alongside a full roadmap of AI chips targeting edge, enterprise, and GenAI applications. The announcement marks Kneron's shift from an edge chipmaker to a full-stack AI infrastructure company.
KL1140 Pushes Large AI Models to the Edge
KL1140 is the first NPU capable of fully running Mamba, a next-generation neural network model, directly on edge devices. According to Kneron Founder and Chief Executive Officer Albert Liu, the chip offers three times the energy efficiency and one-tenth the cost of current cloud-based GPU solutions. Running four KL1140 chips in parallel supports models with up to 120 billion parameters, with power consumption at just one-third that of traditional GPUs.
Albert Liu stated, "The high cost and energy burden of today’s cloud-based AI models show that this approach cannot be sustained long term. KL1140 and our next-generation chips are designed as an answer to that. By running large models at the edge, we bring the full power of LLMs to everyday devices."
University of California, Berkeley testing data confirmed KL1140's significant gains in efficiency over existing edge processors. It supports real-time applications including natural language understanding, smart vision, and robotic decision-making, without needing a constant cloud connection.
A Scalable Lineup for Different AI Use Cases
Alongside KL1140, Albert Liu also unveiled Kneron's full product roadmap for the next three years. The chips span low, medium, and high compute categories:
- KL640 and KL650 are low-power chips designed for always-on camera and smart surveillance use
- KL540 and KL515 are general-purpose chips for IPCam, machine vision, and consumer AI devices
- KL840 and KL1150 are high-performance chips aimed at enterprise edge servers, in-vehicle systems, and GenAI deployment
These models collectively represent a complete, scalable AI product line that Kneron says is now ready for mass production. They are intended to deliver better efficiency and lower cost across a broad range of AI scenarios.
Building a Complete AI Infrastructure Stack
During the press conference, Albert Liu emphasized that cloud compute for AI is already reaching unsustainable levels of power usage. "By 2035, global data centers will consume more than 175 gigawatts of electricity," he said. "That is not a viable path. We need more efficient compute, closer to where data is created."
Kneron is addressing this challenge by scaling edge AI infrastructure. The company also presented its KNEO350 AI server and the KNEO Pi developer board, part of its broader ecosystem that now includes over 28,000 developers worldwide.
The company has already deployed solutions across sectors including healthcare, education, and sovereign AI projects. It is also partnering with Italian firm Spark to build LLM servers based on Kneron chips.
Albert Liu added, "We are not only designing chips. We are building full-stack AI systems. From silicon to applications, we are working to create a sustainable, scalable AI infrastructure that supports security, privacy, and accessibility."
The launch event underscored Kneron’s ambition to lead a new generation of AI development, centered on edge performance and practical deployment. As Albert Liu concluded, "The future lives at the edge. Our goal is to make AI faster, safer, and more available to everyone, everywhere."
©2026 Business Next Media Corp. All Rights Reserved. No.102, Guangfu S. Rd., Da'an Dist., Taipei City 106, Taiwan