Bringing AI To The Core: IBM’s Bets On In-Platform Intelligence
IBM’s 2025 infrastructure refresh — z17, LinuxONE 5, and Power11 — isn’t just a silicon upgrade. It’s a performance, TCO, and sovereignty play. In-platform AI acceleration, post-quantum cryptography (PQC), and hybrid cloud integration make a bold case for keeping your data — and your AI — close to home and hopefully on IBM hardware. While the vision is strong, execution will be key. IBM’s challenge isn’t just technical; Big Blue will need to convince buyers that mainframe-class infrastructure is relevant in a cloud-native world.
- Spyre accelerator: AI where it belongs. IBM announced the Spyre accelerator, an AI accelerator card with 32 AI-optimized NPU cores, with the z17 launch this spring. In July, IBM revealed that Spyre will be available for Power11, as well. The new accelerator will deliver a cost-effective AI inferencing alternative to cloud-hosted NVIDIA H100 or Intel Gaudi 3 AI solutions by eliminating data movement (and egress fees), reducing latency for inference, leveraging a more power-efficient architecture, and simplifying compliance by keeping data local.
- Post-quantum cryptography: IBM leads while others lag. IBM has led the market in hardware-accelerated PQC (Kyber and Dilithium) since its 2022 z16 and corresponding LinuxONE iterations. With z17 and LinuxONE 5, it has expanded PQC support and extended capabilities into the OS layer for a full-stack solution. Competitors such as HPE, Dell, and Lenovo lag with either partial portfolio coverage or software-only implementations, forcing a trade-off between leaving critical workloads vulnerable or slowing performance to apply complex encryption algorithms in software.
- AI-assisted operations: watsonx Assistant for Z and beyond. The latest watsonx Code Assistant for Z supports Linux on Z and integrates with Z Operations Unite, enabling natural language queries for real-time system insights and automation. IBM also released COBOL Upgrade Advisor for Z (CUAZ) to accelerate the transition to the newly released COBOL version 6.5. CUAZ, Automatic Binary Optimizer (ABO), and IBM watsonx Code Assistant for Z are all part of IBM’s app modernization lifecycle. Similar capabilities are being extended to Power11 and LinuxONE via watsonx Orchestrate and Code Assistant.
- HashiCorp: an integration that matters. IBM’s $6.4 billion acquisition of HashiCorp brings Terraform, Vault, Consul, and Boundary into its ecosystem, embedding secrets management, service discovery, and identity-aware access across z/OS, PowerVS, and LinuxONE. This streamlines the integration of IBM Z and Power-based workloads with enterprise cloud infrastructure by strengthening hybrid cloud security and automation.
- Value: “Big Blue” still means big green, at least up front. While IBM platforms have high up-front costs, they offer unmatched uptime, energy-efficient AI inferencing, clear privacy and compliance advantages, and long-term TCO savings, especially when you factor in the Telum II + Spyre combo — though IBM must still simplify pricing and clearly communicate the benefits of running AI on-prem versus cloud alternatives.
What It Means For Enterprise Buyers
My take? Here’s what I am advising clients to do:
- Run AI where your data lives. With Spyre and Telum II, IBM enables low-latency, secure AI inferencing directly on-platform — eliminating data movement, reducing power use, and avoiding cloud egress fees for better TCO.
- Secure the future with PQC. IBM leads in post-quantum cryptography with full-stack support, making it a strong choice for industries such as finance, healthcare, and government that manage long-lived, sensitive data.
- Modernize operations with automation and control. The integration of HashiCorp’s control plane tools and watsonx Assistants brings scalable automation, secrets management, and AI-driven ops to IBM’s legacy platforms.
- Bridge the adoption gap. IBM’s vertically integrated stack and AI-assisted tools help reduce complexity, but broader adoption still hinges on improving developer experience, simplifying pricing, and attracting cloud-native teams.
IBM is working to make its Z mainframe and Power midrange platforms relevant again with compelling AI-driven updates. But success will depend on how well it can translate technical differentiation into operational simplicity, developer accessibility, and economic clarity. If you are looking to understand how IBM’s recent announcements affect your mainframe strategy, please reach out for a guidance session to help figure out your next steps.