International Business Machines (IBM) has unveiled its latest line of data center chips and servers, the Power11, with claims of improved power efficiency and enhanced support for artificial intelligence (AI) deployment in enterprise operations. This marks IBM’s first major upgrade to its “Power” chip family since 2020. The company says the new systems are tailored for sectors such as healthcare, finance, and manufacturing, and are set to rival offerings from Intel and AMD in terms of performance and efficiency.
The Power11 chips are engineered to operate as an integrated system of hardware and software—similar to Nvidia’s AI servers—with a strong emphasis on reliability and security. According to Tom McPherson, IBM’s general manager of Power systems, the chips require no planned downtime for software updates and experience just about 30 seconds of unplanned downtime annually. Notably, the systems can detect and respond to ransomware attacks in under a minute, enhancing their appeal in today’s cybersecurity-conscious climate.
IBM also disclosed plans to integrate the Power11 with its AI accelerator chip, Spyre, later in the fourth quarter of 2025. Rather than compete directly with Nvidia in training AI models, IBM is positioning the Power11 to excel at AI inference—helping businesses automate and speed up tasks through already-trained AI models. This approach aims to make AI integration smoother and more accessible across industries.
The Power11 systems will be available starting July 25, and IBM has already begun working with select early customers to demonstrate how the chips can enhance business workflows. McPherson noted that while the systems won’t offer the computational heft needed for full-scale AI training, they provide a robust, secure platform for companies to deploy AI features quickly and efficiently.