IBM launches new chips and servers to boost AI deployment and data centre efficiency

IBM has officially launched its latest Power11 chips and servers, designed to significantly enhance power efficiency and elevate AI capabilities for business sectors such as finance, healthcare, and manufacturing. This marks the first substantial update to IBM’s Power chip series since 2020, showcasing the company's commitment to advancing enterprise technology. With the introduction of Power11, IBM aims to assist organizations in streamlining their artificial intelligence deployments while ensuring improved system reliability and security. Set to be available starting July 25, these Power11-based systems are engineered to eliminate planned downtime during software updates, allowing for uninterrupted operations. The new systems are built with resilience in mind, boasting an average unplanned downtime of just over 30 seconds per year. Moreover, they are equipped with capabilities to detect and respond to ransomware attacks within a minute, enhancing overall cybersecurity measures. Unlike Nvidia’s GPU-centric infrastructure primarily used for training extensive AI models, IBM's Power11 focuses on inference—the application of existing AI models to optimize everyday business processes. IBM is positioning Power11 as a reliable alternative for enterprises eager to handle AI workloads without necessitating a complete infrastructure overhaul. Furthermore, the company plans to integrate Power11 with its Spyre AI chip, first revealed last year, in the fourth quarter of 2025, which is anticipated to further boost the AI inference capabilities of its server systems. Tom McPherson, the general manager of Power systems at IBM, emphasized the importance of reliability and security in the development of Power11. He noted, "We can seamlessly integrate AI capabilities for inference acceleration, assisting businesses in improving their processes." While acknowledging that Power11 may not possess the same power for training large models, he assured that its inferencing capabilities will be straightforward to integrate and highly effective.