Ultimately, Samsung plans to repeatedly launch new variations of AGI Computing Lab chip designs in “an iterative mannequin that may present stronger efficiency and assist for more and more bigger fashions at a fraction of the facility and price,” he stated.
“Via the creation of the AGI Computing Lab, I’m assured that we’ll be higher positioned to resolve the advanced system-level challenges inherent in AGI, whereas additionally contributing reasonably priced and sustainable strategies for the long run technology of superior AI/ML fashions,” Kyung wrote.
Untapped market potential
Samsung’s transfer partly seems to be an effort to search out new income streams in an as-yet untapped market, as its core enterprise, which is reminiscence, has grow to be a commodity, famous Gaurav Gupta, VP analyst, rising developments and applied sciences, at Gartner.
“They’re in search of one other alternative to develop,” he stated. “That is the place chips for inference are available in.”
Certainly, most corporations that at present construct elements for laptop processing and reminiscence try to maintain tempo with the speedy evolution of AI in numerous particular person methods to supply cost-effective computing assets.
At the moment, the generative AI chip marketplace for coaching fashions is dominated largely by Nvidia, with AMD having some share within the house, Gupta stated. However these are for fashions working on GPUs, which might be scarce and expensive and thus aren’t a long-term answer for working AI fashions.
