Advanced Micro Devices Inc (AMD) revealed more information about its upcoming artificial intelligence (AI) chip that could pose a threat to market leader Nvidia Corp.
However, they did not provide any details about the chip’s cost or customers. Despite Santa Clara, CA- based AMD’s showcase of its MI300X chip that wrote a poem about San Francisco, shares closed 3.6% lower on the AI strategy news.
In contrast, Nvidia shares rose 3.9% to close at US$410.22, making it the first chipmaker with a market cap above US$1 trillion.
Although AMD has claimed that their MI300 and MI300A chips will have 192 gigabytes of memory, allowing tech companies to save on costs for services similar to ChatGPT, analysts are unclear on whether AMD’s software solution is competitive with Nvidia’s.
After the presentation, Kevin Krewell, principal analyst at TIRIAS Research, commented that the lack of a mention of a major customer for AMD’s chips might have disappointed investors.
Already, Nvidia dominates the AI computing market with an 80-95% share, making them hard to replace.
Intel, Cerebras Systems, and SambaNova Systems are among the few competitors that have similar products. Nevertheless, AMD has started shipping their “Bergamo” central processor chips to companies such as Meta Platforms.
Alexis Black Bjorlin, who works with the computing infrastructure at Facebook’s Meta, mentioned that they have adopted the Bergamo chip. Moreover, AMD updated their Rocm software to rival Nvidia’s Cuda software platform.
Soumith Chintala, a Meta VP, stated that the goal is to make it simpler for AI developers to switch to AMD’s platform from the dominant vendor.
Despite the optimism, analysts have suggested that companies may not be convinced of using AMD’s chips if they are unable to match Nvidia’s software.