Lisa Su, CEO of AMD (photo), unveiled the new graphics processing unit (GPU) “Instinct MI455X” in a keynote speech held on the 5th (local time), ahead of the opening of CES 2026. At around 1 p.m. on the same day, Jensen Huang, CEO of NVIDIA, unveiled the next-generation AI semiconductor “Vera Rubin,” prompting a head‑to‑head showdown with AMD’s new product.

In the CES 2026 keynote speech held on the 5th at the Venetian Hotel in Las Vegas, USA, Su stated, “We are preparing for the era of ‘yotta computing,’” adding, “AMD’s products will be game changers in preparing for the yotta era.” “Yotta” is the 10 to the 24th power unit, signifying astronomical computational performance capable of operating and training AI models estimated to be comparable to or exceed the human brain. Su also said, “Over the next few years, the world’s computing capacity will need to increase by 100 times.”

The MI455X unveiled by Su is an AI accelerator composed of two graphic computing dies (GCDs), two memory controller dies (MCDs), and 16 HBM4 units. Su said, “With the MI455X GPU, we will deliver up to 10 times higher performance compared with the previous generation (MI355X).” This product uses the latest 2‑nanometer process technology and has more memory capacity than NVIDIA’s Rubin GPU, which is known to incorporate eight units of next-generation high-bandwidth memory, HBM4. It will be used in data center servers of AI companies, including OpenAI.

Su said, “Within five years, an era will arrive in which more than 5 billion users worldwide use AI every day,” adding, “We will provide optimized computing resources across all domains, from the cloud to PCs and edge computing.”

Popular News