1. Home >
  2. Computing

Samsung to Battle Nvidia in 2025 With Its New Mach-1 AI Chip

The company is pouring billions into R&D with plans to disrupt the AI market in the near future.
By Josh Norem
Samsung 3nm
Credit: Samsung

Samsung is looking to enter the AI market with a new product it calls Mach-1. The company has announced it intends to go head-to-head with established names like Nvidia with some ambitious plans. As always, we'll need to wait a while to see if any of them come to fruition. Samsung expects Mach-1 to arrive in 2025, so it might end up competing with Nvidia's just-announced Blackwell platform.

According to Wccftech, Samsung is not just looking to dip a toe into the AI market but rather to become a leading provider of AI hardware. It is in the unique position of being a company that manufactures both chips and memory, and it's hoping to capitalize on that advantage to crush its biggest Korean rival, SK Hynix. The company has reportedly already stated it plans on opening an artificial general intelligence (AGI) laboratory, which will help it launch its first AI chip, Mach-1.

Samsung HB3e
Though Samsung is trying to reduce the usage of HBM with AI chips, it's also trying to compete with rival SK Hynix in the nascent HBM3e market as well. Credit: Samsung

Details on this chip are vague at this point, but one of Samsung's goals is to reduce bottlenecks between the GPU and the nearby stacks of memory. Samsung says these GPU and memory bottlenecks degrade performance and consume too much energy. Its solution, which involves several new algorithms, will allegedly improve power efficiency by up to 8x. Samsung says this massive boost in efficiency will allow AI chips to run low-power memory instead of power-hungry high-bandwidth memory (HBM) like they currently require.

Samsung expects to have prototypes of its new Mach-1 system ready by the end of 2024, which should lead to its deployment in early 2025. The company has already procured an order for Mach-1 chips from a Korean firm trying to wean itself off of Nvidia hardware, which is currently both expensive and hard to acquire.

Tagged In

Samsung Semiconductors Artificial Intelligence Memory

More from Computing

Subscribe Today to get the latest ExtremeTech news delivered right to your inbox.
This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of use(Opens in a new window) and Privacy Policy. You may unsubscribe from the newsletter at any time.
Thanks for Signing Up