Skip to main content

The digital world has transitioned from the consumption of content delivered to mobile devices to massive data creation across a spectrum of edge devices. This data is collected and analyzed to produce valuable insights enhancing device performance, user experiences, and automation. As industries continue to deploy smart connected devices, the volume of digital data generated annually is growing exponentially and is expected to eclipse 180 zeta bytes globally by 2025. That is 180 x 1021, or the equivalent of 180 trillion 1 Gigabyte USB sticks! This is “Big Data”.

Over the past decade, the cloud has been the platform of choice for storing, processing, and monetizing big data. However, the cloud has its limitations and cannot service every connected device or application.
These limitations include persistent connectivity, real-time response and decision, privacy, and the escalating cost of cloud services as a dominant factor on the balance sheet.

As a result, 75% of digital data will be processed closest to the point of origin at the edge. Edge computing will become the dominant technology of the current decade, representing a US$274 billion market opportunity in 2025. The fundamental innovation underpinning the next wave of edge computing is artificial intelligence and machine learning, with a predicted semiconductor market size approaching US$30 billion by 2026.

With model complexity, size and performance ever increasing, commodity compute offerings have proven incapable of delivering the power, cost, and multi-sensor processing requirements of edge use cases.

To address this accelerating demand, AiM Future today announced a family of NeuroMosAIc (NMP) processor intellectual property (IP) and software. Available in three versions, the NMP-300, 500, and 700, are tiered by performance and feature set for users seeking fully verified, pre-configured accelerators to cut design time and fast-track device production. These new AI accelerators inherit their fully
configurable predecessor’s performance and energy efficiency advantages to enable new device classes from always-on, battery-operated sensors to edge gateways and servers.

“Adding artificial intelligence to everyday devices is improving the way we communicate, learn, and create”, said ChangSoo Kim, Chief Executive Officer of AiM Future. “Our team has been developing highly efficient machine learning technology addressing how humans safely and securely conduct their daily lives for over 6 years. We are grateful for our customers’ success and their guidance has led us to deliver a family of pre-configured edge AI accelerator IP and software to further ease the development of intelligent devices and broaden the addressable market to become truly ambient AI.”

“The motivation to deliver ever smarter vehicles and robotics is fueled by driver, pedestrian, and worker safety, and therefore a greater human experience”, says Alex Kim, CEO of PnP Networks. “The market for intelligent devices is rapidly expanding, driven by continually changing requirements and expectations. To keep pace requires innovative new architectures, such as the NeuroMosAIc Processor, to lower the complexity and energy consumption of adding machine learning to edge devices.”

NeuroMosAIc Processor Family Highlights

The NMP-300 is an ideal match for machine learning applications in ultra-low power, far-edge devices such as wearables and environmental and image sensors. It offers 0.5 TOPS while consuming microwatts on common at-sensor applications.

The NMP-500 targets applications requiring performance efficiency, such as smartphones, drones, AR/VR, and high-end home appliances. It delivers up to 4 TOPS or may be scaled down to 2 TOPS to match the power and area constraints of cost-sensitive applications.

The NMP-700 is the highest-performance member of the family targeting edge gateways, servers, robotics, and UAVs. It delivers 8 or 16 TOPS where performance is the primary design objective.

Software Enables End-User Applications

The co-designed NeuroMosAIc Studio provides developers with a comprehensive set of tools to run pre-trained machine-learning models on the NMP hardware. This includes a sophisticated hardware-aware model converter, mapper, and profiler to produce maximum efficiency. Support for industry-standard frameworks, such as ONNX, PyTorch, and TensorFlow Lite, offer ML engineers flexibility while also
leveraging GLOW and TensorFlow XLA compilers to achieve highly optimized models. This equates to a production-ready hardware and software solution generating the highest performance and lowest power computation of edge AI applications.

Availability and Additional Reference

• The NeuroMosAIc Processor IP and Studio are available to lead customers today.
• For more information, please contact us at

About AiM Future

Based in Seoul, South Korea, the company was founded in late 2020 with LG Electronics as a strategic investor. Subsequently, it attracted a seed round investment from a group of leading venture capital firms We Ventures, KB Investment, and D.Camp. The company is focused on the development of neural network hardware accelerator IP and compiler software. Its flagship NeuroMosAIc architecture achieved commercialization in 2019 and the company has executed several license agreements with partners around the world.

See more at