IBM has announced that its new AI hardware, the IBM Spyre Accelerator, will soon be available for commercial use. The company stated that the accelerator is designed to enable low-latency inferencing to support generative and agentic AI applications, while also focusing on security and resilience for core business workloads. According to IBM, the Spyre Accelerator will be available starting October 28 for IBM z17 and LinuxONE 5 systems, with Power11 server availability expected in early December.

The move comes as businesses increasingly shift from traditional IT workflows to those powered by AI agents, which require fast inference times and real-time responsiveness. IBM says its new hardware addresses the need for mainframes and servers to run complex AI models alongside demanding enterprise tasks without sacrificing performance or security. The company emphasized that the Spyre Accelerator is built to allow clients to keep critical data on-premises, reducing risk and improving operational efficiency.

IBM developed the Spyre Accelerator through its research-to-product pipeline, incorporating innovation from the IBM Research AI Hardware Center and enterprise-grade development from IBM Infrastructure. The chip was initially introduced as a prototype and refined through testing at IBM’s Yorktown Heights campus and in collaboration with partners such as the University at Albany's Center for Emerging Artificial Intelligence Systems.

Now a commercial system-on-a-chip, the Spyre Accelerator features 32 individual cores and 25.6 billion transistors, produced using 5nm node technology. Each unit is mounted on a 75-watt PCIe card, allowing up to 48 cards in an IBM Z or LinuxONE system or up to 16 cards in an IBM Power system for scalable AI deployment.

"One of our key priorities has been advancing infrastructure to meet the demands of new and emerging AI workloads," said Barry Baker, COO, IBM Infrastructure & GM, IBM Systems. "With the Spyre Accelerator, we're extending the capabilities of our systems to support multi-model AI – including generative and agentic AI. This innovation positions clients to scale their AI-enabled mission-critical workloads with uncompromising security, resilience, and efficiency, while unlocking the value of their enterprise data."

"We launched the IBM Research AI Hardware Center in 2019 with a mission to meet the rising computational demands of AI, even before the surge in LLMs and AI models we've recently seen," said Mukesh Khare, GM of IBM Semiconductors and VP of Hybrid Cloud, IBM. "Now, amid increasing demand for advanced AI capabilities, we're proud to see the first chip from the Center enter commercialization, designed to deliver improved performance and productivity to IBM's mainframe and server clients."

For businesses using IBM’s mainframe systems together with processors like Telum II for z17 and LinuxONE 5 models, Spyre aims to enhance security while delivering low latency and high transaction rates. Companies can use this hardware stack with software tools from IBM to scale multiple AI models for applications such as fraud detection or retail automation.

On Power-based servers, customers will have access to a catalog of installable AI services supporting end-to-end enterprise workflows. The combination of Spyre with an on-chip accelerator (MMA) allows faster data conversion for generative AI tasks. The system supports large prompt sizes—enabling ingestion of over eight million documents per hour—to integrate knowledge bases into business operations efficiently.

More information about the product is available at http://www.ibm.com/solutions/ai-accelerator.