Qualcomm-backed firm launches latest AI chips — posing a challenge to giants like Nvidia

GPT, or generative pre-trained transformer, refers to AI algorithms trained on huge amounts of data that are able to produce text and images. Examples include OpenAI’s ChatGPT.

However, many of the companies wishing to use generative AI functions and apps today rely on giants like Microsoft and Amazon through their cloud businesses. These so-called “hyperscalers” are buying huge amounts of chips from companies like Nvidia to train up these massive AI models in data centers running complex servers.

Kneron is betting that businesses will not always want to rely on these cloud giants for their AI needs.

Instead, KNEO 330 is aimed at businesses wishing to run their own AI applications from their own premises rather than relying on servers located elsewhere and owned by one of the big cloud players.

“We find a strong market need in big high-tech companies, or medical centers or financial institutions, they care about their privacy,” Albert Liu, CEO of Kneron, told CNBC ahead of the product launch. “They don’t want to upload that to OpenAI.”

Liu believes that businesses might not want to give their data over to a third-party to use their AI tools. Instead, with Kneron’s technology, companies can build their own AI applications using the company’s servers located on-premise at a business.

Kneron’s launch comes just days after both Nvidia and AMD launched their latest AI chips, with both appearing to ramp up the pace of launches. The products from Nvidia and AMD are aimed at huge data centers from tech giants that can train up massive AI models.

On Wednesday, Kneron also announced that its latest KL830 processor chip will be in a new PC. This lays down a challenge to companies like Intel and AMD that are making PC chips. The KL830 will allow the PC to run AI applications on device rather than connecting to the internet. The idea is this will improve privacy as data is not connected to the internet.