Solar Energy Storage,Battery System,Battery Structure,Lifepo4 Battery Cells Guangdong Yuqiu Intelligent Technology Co.,Ltd , https://www.cntcetltd.com
Recently, the enigmatic and diverse nature of technological intelligence has finally come into the spotlight. Wu Ren, CEO of Heterogeneous Intelligence, shared his vision: “Heterogeneous intelligence encompasses chips, models, and supercomputers. Our goal is to bring AI models from the cloud to local devices and terminals, making them smart.†Wu Ren emphasized that the company aims to move artificial intelligence closer to end-users, enabling edge devices to perform intelligent tasks independently.
Wu Ren’s calm and composed demeanor made a strong first impression. Since founding Heterogeneous Intelligence in 2015 after leaving Baidu, he has remained relatively quiet, contributing to the company's mysterious image. However, this silence has not diminished the interest in his work.
Rather than viewing the recent media event as a simple exchange, it can be seen as an opportunity for Wu Ren to share insights about the current state and future direction of Heterogeneous Intelligence. He emphasized the importance of bringing AI to the edge, rather than relying solely on cloud-based solutions.
“Cloud IoT is a false proposition; it’s hard to achieve,†Wu Ren stated. He explained that while the Internet of Things seems to be widely connected, it actually relies on extensive infrastructure—submarine cables, data centers, and more. The exponential growth of edge devices has outpaced the capacity of traditional cloud systems.
To address this challenge, Heterogeneous Intelligence is focusing on moving AI capabilities from the cloud to edge devices, allowing them to think and process data locally. This approach, known as Intelligent Internet of Things (I2oT), enables edge devices to make decisions and connect back to the network, creating a more efficient and responsive system.
“Doing AI locally is the future,†Wu Ren said. “Heterogeneous Intelligence will support companies in this direction.†This shift presents two major challenges for AI chips: high performance to handle complex computations and low power consumption to fit into small, edge-based devices.
“It’s a challenge, but also an opportunity,†Wu Ren said with confidence. “We’re ready to take on the challenge and seize the opportunity.â€
In response to these demands, Heterogeneous Intelligence has developed specialized AI chips designed for efficiency and performance. Their NovuTensor V1 chip delivers 15 TOPS of computing power while consuming only 5 watts—making it ideal for edge deployment. Compared to Google’s high-power chips or Huawei’s Cambrian chips, which are either too energy-intensive or lack sufficient performance, NovuTensor strikes a perfect balance.
The 5 watts of power required by the chip is equivalent to what a USB port can provide, making it highly portable and suitable for a wide range of applications. “This is the key to breaking through from the cloud to the edge,†Wu Ren said.
Heterogeneous Intelligence has also developed a full-stack solution, including patented technologies for three-dimensional tensor convolution, data scheduling, merging, and processor collaboration. These innovations ensure optimal performance and efficiency across different industries.
“We want to empower other companies,†Wu Ren repeated often. By providing tools for training smarter models, more complex neural networks, and better data utilization, Heterogeneous Intelligence helps businesses leverage AI without needing to build everything from scratch.
Their flagship product, NovuBrain, targets areas like security, manufacturing, healthcare, smart cities, and autonomous driving. They have already partnered with industry leaders such as HP and KUKA.
Looking ahead, the company is advancing its 28nm ASIC chip into mass production, while simultaneously developing 16nm and 7nm versions. According to Wu Ren, the team is confident in meeting the timeline for the next-generation chips. Mike Li, who previously worked on 16nm chips at Inphi, is leading the development and is fully confident in the process.
“The future world must be led by the intelligent Internet of Things,†Wu Ren concluded. “Heterogeneous Intelligence aims to transition the Internet of Everything into the Internet of All Things, providing the technology needed for this transformation. As the network grows, so will our company, and we aim to become a great global enterprise.â€