At the Consumer Electronics Show in Las Vegas, AMD Chair and CEO Lisa Su introduced the company’s latest AI accelerators, including the new MI300X series designed for high-density data centers and the upcoming Helios system for on-premise deployments. The MI300X delivers up to 1.2 exaflops of mixed-precision performance per rack, while Helios pairs these accelerators with AMD’s 5nm EPYC CPU to offer a turnkey solution supporting over 100 petaflops of training throughput. Su noted that these products will begin sampling to strategic partners in Q2, with general availability slated for Q4 2026, positioning AMD to capture enterprise workloads that demand both compute density and power efficiency.

During her keynote, Su emphasized that AMD’s headcount has grown by 18% year-over-year, reaching over 30,000 employees globally. She stated that the company is hiring ‘AI-forward’ engineers at twice the rate of general technical roles, focusing on candidates with expertise in large language models, hardware-software co-design, and system integration. This approach follows AMD’s investment of $2 billion in R&D over the past twelve months, reflecting Su’s commitment to scaling the workforce in lockstep with the firm’s expanding AI roadmap.

In a separate announcement, AMD detailed a long-term roadmap aimed at narrowing the gap with the market leader. The company unveiled the Instinct 7000X series for midsize corporate data centers, delivering 600 teraflops of FP16 performance per card and a 30% boost in memory bandwidth versus the previous generation. Analysts at Bernstein Projects estimate that this lineup could grow AMD’s share of the global AI accelerator market from 10% to 15% by the end of 2027, driven by demand from financial services, healthcare, and telecom operators seeking on-premise inference and training capabilities.