Expedera develops customizable neural processing unit (NPU) semiconductor IP and a TVM-based software stack for edge AI inference. The company plans to use its latest funding to advance NPU IP products and services and to expand its global engineering footprint. Led by CEO Da Chuang, Expedera focuses on improving performance, power, and latency while reducing cost and complexity for edge applications. Its NPU solutions are scalable and deployed in over 10 million consumer devices, serving use cases from edge nodes and smartphones to automotive. The platform supports importing trained networks, multiple quantization options, automatic completion, compilation, estimator and profiling tools, and multi-job APIs. Expedera operates engineering development centers and customer support offices in the UK, China, Japan, Taiwan, and Singapore. Expedera develops scalable neural engine semiconductor IP branded as the Origin deep learning accelerator and accompanying software platform, which the company says offers industry-leading performance per watt. Its IP scales to 128 TOPS on a single core and to PetaOps with multi-core configurations, and supports frontends including TensorFlow, ONNX, Keras, MxNet, Darknet, CoreML and Caffe2 via Apache TVM. The company licenses its IP to chip designers so they can add AI inference into devices such as smartphones, smart speakers, security cameras, PCs/tablets, wearables, automotive systems and edge servers. Expedera describes its solutions as third-party silicon-validated and applicable from edge nodes and smartphones to automotive and data centers. The company announced an $18M Series A led by Sehat Sutardja and Weili Dai, bringing total funding to over $27M; proceeds will be used to speed product development and expand sales and marketing. Expedera is headquartered in Santa Clara and is led by CEO Da Chuang.

Total raised

$38M

Funding rounds

2

Latest round

Series B

Latest activity

May 2024