# Expedera

> Expedera licenses configurable neural-processing-unit IP and a supporting software stack to semiconductor and device makers. Its packet-based architecture is designed to keep compute blocks busy while cutting memory movement, an expensive and power-hungry part of AI inference. The company says its earlier Origin designs have shipped in tens of millions of devices, while the newer Origin Evolution line expands the approach to large language models, automotive systems and data-center inference.

- **Founded:** 2018
- **Headquarters:** Santa Clara, United States
- **Founders:** Siyad Ma (CEO and co-founder), Da Chuang (Executive Chairman and co-founder), Sharad Chole (Chief Scientist and co-founder)
- **Products:** Origin E1 / LittleNPU, Origin E2, Origin E6, Origin E8, TimbreAI T3
- **Notable:** Origin silicon emerged from stealth in 2021 with third-party measurements cited at up to 18 TOPS/W on 7nm., The first disclosed production deployment shipped in 2022 for on-device 4K low-light video denoising., Expedera said Origin IP had shipped in more than 10 million devices by 2024 and in tens of millions by 2025.

## Products & services

- **Origin E1 / LittleNPU** — A compact NPU IP family for always-on and cost- or area-sensitive applications around the 1-TOPS class.
- **Origin E2** — Configurable 1-to-20-TOPS NPU cores for smartphones, edge nodes and consumer or industrial devices.
- **Origin E6** — Higher-performance configurable NPU IP for phones, computers, edge servers and automotive systems.
- **Origin E8** — The high-end Origin family for demanding inference uses including data centers and autonomous-vehicle systems.
- **TimbreAI T3** — An NPU offering listed within Expedera's current product portfolio for efficient edge inference.
- **Origin Evolution** — A packet-based NPU IP platform for LLM, CNN and RNN inference, configurable to 128 TFLOPS per core and scalable across multiple cores.
- **NPU software stack** — Model import, quantization, compilation, estimation, profiling and multi-job APIs, with support including TVM, Hugging Face and llama.cpp representations.

## Achievements

- Origin silicon emerged from stealth in 2021 with third-party measurements cited at up to 18 TOPS/W on 7nm.
- The first disclosed production deployment shipped in 2022 for on-device 4K low-light video denoising.
- Expedera said Origin IP had shipped in more than 10 million devices by 2024 and in tens of millions by 2025.
- Opened a sixth development center in Hyderabad, India, in June 2024.
- Earned ISO 26262 ASIL-B automotive functional-safety readiness certification in July 2024.
- Earned ISO 9001:2015 quality-management certification in April 2025.
- Origin Evolution won the Edge AI and Vision Alliance's 2026 Best Edge AI Processor IP award.

## Latest updates

- **2026-04** — Origin Evolution named Best Edge AI Processor IP in the Edge AI and Vision Product of the Year Awards.
- **2025-05** — Launched Origin Evolution for LLM, CNN and RNN inference from edge devices through data-center systems.
- **2025-04** — Received ISO 9001:2015 quality-management certification for Santa Clara and Taipei operations.
- **2024-09** — Co-founder Siyad Ma became CEO; founding CEO Da Chuang moved to Executive Chairman.
- **2024-07** — Received ISO 26262 ASIL-B automotive safety-readiness certification.

## Links

- Website: https://www.expedera.com/
- LinkedIn: https://www.linkedin.com/company/expedera/
- Twitter/X: https://twitter.com/ExpederaInc
- YouTube: https://www.youtube.com/@ExpederaInc
- Facebook: https://www.facebook.com/ExpederaInc/

---

Profile page: https://yespress.io/expedera
Published by YesPress — https://yespress.io
Last updated: 2026-09-26