Catalog of SIP Cores
System on Chip design resources
System on Chip design resources
October 15, 2026 -
New scalable AI accelerator IP delivers energy-efficient, always-on intelligence with native support for neuromorphic computing and TinyML applications.
BARCELONA, Spain — INTERA, a developer of advanced semiconductor intellectual property (IP) for Edge AI acceleration, today announced the launch of EAGLE Cognit, a new near-memory AI accelerator designed to enable high-performance, ultra-low-power artificial intelligence directly within next-generation System-on-Chip (SoC) architectures.
As AI intelligence increasingly moves from centralized cloud infrastructure to connected devices, semiconductor manufacturers face growing pressure to deliver greater computational capabilities while minimizing energy consumption, silicon area and processing latency. EAGLE Cognit addresses these challenges through an innovative near-memory computing architecture designed to reduce unnecessary data movement and maximize AI processing efficiency.
Redefining Energy Efficiency at the Edge
Unlike conventional processing architectures that frequently transfer data between memory and computational units, EAGLE Cognit brings AI computation closer to where data is stored. This approach reduces data movement overhead and enables more efficient execution of neural network workloads.
The accelerator features configurable matrix-vector processing, supports numerical precision from 2 to 32 bits and offers scalable configurations with 4 or 16 processing elements (PE), with architectural testing extending to 32 banks/PE.
Designed for always-on Edge AI applications, EAGLE Cognit combines low-latency inference, compact silicon implementation and an attractive cost-to-performance ratio.
Enabling the Next Generation of Neuromorphic Intelligence
A key differentiator of EAGLE Cognit is its native support for sparse computation and Spiking Neural Networks (SNNs), making it particularly suitable for event-driven and neuromorphic AI applications.
By processing information efficiently when relevant activity occurs, neuromorphic computing offers an attractive approach to implementing intelligent sensing systems with demanding energy constraints.
Besides SNN, EAGLE Cognit also supports conventional neural network architectures including Convolutional Neural Networks (CNNs and FCNN) or Multilayer Perceptron (MLP) providing developers with flexibility across classical and emerging AI workloads.
Designed for Real-World Edge AI Applications
EAGLE Cognit targets a broad range of applications where continuous intelligence, low latency and efficient local processing are essential, including:
By enabling AI inference directly on embedded devices, EAGLE Cognit also supports enhanced data privacy, reduced dependence on cloud connectivity and greater technological sovereignty.
Accelerating Semiconductor Innovation
As part of INTERA's Cognit family of AI accelerator IPs, EAGLE complements the company's portfolio of configurable Edge AI processing architectures, offering semiconductor developers a differentiated solution for energy-efficient, near-memory intelligence.
To facilitate adoption, INTERA provides a development ecosystem supporting established AI frameworks with TensorFlow and Keras model support, Tensor Flow Lite, Quantization and optimization alongside with tools for neuromorphic development supporting NIR compatible models and a toolchain based on MLIR and NIR.
“EAGLE Cognit represents an important step in our mission to make advanced AI acceleration more accessible, scalable and energy-efficient,” said NARCIS AVELLANA (CEO) at INTERA. “By combining near-memory computing with native support for neuromorphic workloads, we are providing semiconductor companies with a new approach to integrating always-on intelligence directly into their products.”
Advancing the Future of Edge Intelligence
With the introduction of EAGLE Cognit, INTERA continues expanding its portfolio of semiconductor IP solutions designed to address the growing computational demands of intelligent connected devices.
The new accelerator reinforces INTERA's commitment to enabling seamless AI integration, maximum performance per watt and greater flexibility in next-generation semiconductor design.
About INTERA
INTERA is an Edge Computing Systems Company focused on advanced semiconductor IP for high-performance, energy-efficient AI processing at the edge. The company develops and licenses scalable AI accelerator architectures and processing technologies that enable semiconductor manufacturers to integrate intelligent inference capabilities directly into their SoCs and embedded devices.
Through its Cognit family of AI accelerator IPs, INTERA supports the advancement of TinyML and neuromorphic Edge AI computing across industrial, healthcare, energy, security and connected-device applications.
Website: www.intera-group.com