At CES 2026, Advanced Micro Devices introduced the MI455X, a next-generation AI accelerator built on a 2nm process node and equipped with 432GB of onboard HBM4 memory. By enabling large language models to reside entirely in GPU memory, the MI455X reduces the number of devices required for inference clusters by up to 40%, while its integrated high-speed interconnect supports sustained data throughput of 8TB/s per device. AMD expects to ship production units in late 2027, targeting data center customers seeking to lower total cost of ownership by eliminating external memory tiers and minimizing rack-scale networking overhead.

During its five-year financial outlook presentation, AMD reaffirmed that its data center segment is forecast to achieve a 60% compound annual growth rate, driven by accelerating demand for GPU compute in cloud and enterprise AI workloads. The company’s CEO highlighted that this division accounted for 25% of total revenue in fiscal 2025 and is on track to surpass 40% by 2028. At the current market valuation of $330 billion, management believes sustained double-digit operating margin expansion—supported by economies of scale and design-to-silicon optimizations—could propel AMD toward a $1 trillion market capitalization within four years.

AMD reported a ten-fold year-over-year increase in downloads of its ROCm open-source software platform, underscoring growing developer engagement with AMD GPU architectures. To further drive enterprise adoption, AMD announced a strategic collaboration with Tata Consultancy Services, under which TCS will integrate AMD EPYC CPUs, Ryzen processors and Instinct GPUs into customized AI solutions for life sciences, manufacturing and financial services. The partnership aims to deploy over 1,000 hybrid cloud and edge reference architectures by mid-2026, positioning AMD as a comprehensive compute and software provider for large-scale AI transformations.