Neural Processing Unit chipmaker, Kneron, has secured another $49M in funding. The money brings its total Series B funding to $97M. The Series B funding round was led by Horizons Ventures and included Liteon Technology Corp, Adata, Palpilot, and Foxconn.
Albert Liu, Founder and CEO of Kneron, said, “Powerful GPT models are mostly still running out of cloud data centers. This results in a slew of issues, including high latency, high data transfer costs, and inadequacies in user privacy and security protection.
“Kneron’s solutions resolve these industry bottlenecks by creating hyper-efficient AI chips. We’re thrilled to announce the completion of our Series B funding announcement to continue our work in making AI technology more secure, accessible, and energy-efficient.”
Who is Kneron?
Kneron styles itself as “the creator of the world’s first edge AI chip to support transformer neural networks.” It launched its first chip, which underpins all GPT models in 2021. The company goes on to say that it is “On a mission to expand access to energy-efficient and secure AI computing.”
What is interesting is that a number of its investors are in the automotive business, where such chips are seen as essential for autonomous vehicles. They are required to create very high-speed, accurate image processing to allow autonomous vehicles to operate safely on the roads. By providing an end-to-end integrated hardware and software solution, Kneron is making it easy for automotive vendors to adopt and use its chips.
Another of the challenges at the moment for AI is raw compute power. The GPU chips that are so sought after for AI are also key to other markets, including gaming. Kneron has recently launched its latest NPU (neural processing unit), the KL730. It describes this as “An auto-grade NPU chip that supports the most advanced lightweight GPT LLMs, like nanoGPT, bringing further options to a high-demand market.”
What will it do with the money?
Kneron claims that the money is to be spent on R&D. It has announced no plans to expand its sales division or open offices in new locations. Given the demand in the market, that makes sense, although most vendors open offices to attract new staff in different countries.
The R&D spending will help the company continue its push into the autonomous driving market. It is already claiming that compared to the latest ultra-lightweight AI chips, its NPUs increase the accuracy of image-based applications by at least 30%. For autonomous driving, where image recognition is critical, this is a significant improvement.
It will also expand its partnership with Foxconn. It wants to create a nano GPT for both automotive and other applications. The goal is an ultra-lightweight AI chip that can operate GPT models off the cloud.
The latter is exceptionally exciting. If Kneron can do that, then the whole GPT market will take a huge leap forward. In addition to automotive, imagine ships using their own onboard AI to better navigate dangerous waters. It would open up shipping lanes that are currently seen as too dangerous for today’s technology.
It would also allow aircraft manufacturers to add AI chips to aircraft to improve existing autopilot functions. Taken further, it could be the step required to create personal flying machines that replace the current generation of cars.
Enterprise Times: What does this mean?
The current explosion of AI has exacerbated the shortage of GPUs. It has created real challenges for those who want to take advantage of the processing capability of GPUs. It is not just the shortages and demand-driven costs, but GPUs are now highly attractive items for criminals.
Setting that aside and looking at Kneron’s technology opens up a number of interesting possibilities. There is real concern about the ability of large numbers of autonomous vehicles to operate adjacent to people and animals. The unpredictability of people and animals puts a lot of pressure on the ability of onboard image-processing units. If Kneron can solve that, then widespread adoption of autonomous vehicles is a lot closer.
The possibility of nano GPT-capable chips that do not need to sit in cloud data centres is just as exciting. It suggests that we could be seeing the first steps to the commoditisation of AI. It also brings the era of the robot in the house a significant step closer.
Where will Kneron go next with this technology? How fast will it adopted by auto manufacturers? If it does, as claimed, deliver the 30% advantage over existing AI chips, those who are already backers, such as Toyota, could take a serious lead in the autonomous vehicle market.