IBM has finally pulled the wraps off its next-generation Power11 servers, marking its biggest leap forward in five years for the Power processor line. And it’s clear whom the company has in its sights: businesses eager to embed AI deeply into their day-to-day operations.
At the heart of the new lineup is the Power11 chip, built not just for raw performance but to handle the demanding world of AI inference—the phase where trained AI models actually run and produce results. Unlike Nvidia, which focuses heavily on hardware for training huge AI models, IBM is carving out a different lane, one that aims to help enterprises make AI practical and efficient.
And the numbers are impressive. Power11 promises up to 55% higher core performance over the earlier Power9 generation and as much as 45% more capacity for entry-level and mid-tier servers compared to Power10. Energy efficiency also gets a boost, with IBM claiming these machines sip less power than rival x86 systems—and they even include an energy-saving mode to cut usage further.
But raw speed is just part of the story. IBM is touting “six nines” reliability—an eyebrow-raising 99.9999% uptime—which translates into less than 30 seconds of unplanned downtime per year. Even software updates won’t demand planned downtime, a major selling point for industries like banking, healthcare, retail, and government that can’t afford to blink.
Security, too, gets a new shield. A feature called Power Cyber Vault can reportedly detect ransomware attacks in under a minute, adding another layer of protection for enterprises swimming in sensitive data.
The Power11 family spans high-end, mid-range, and entry-level systems, plus the IBM Power Virtual Server running in IBM Cloud. That gives customers flexibility whether they want to keep workloads on-premises or shift them into the cloud.
Businesses are already seeing the payoff. Jasmine Kaczmarek, VP of technology at distributor MR Williams, shared how she solved a problem 18 times faster using watsonx Code Assistant for i on Power systems. “With just 20 minutes and watsonx, I could investigate a report, trace the logic, understand calculations, and document the issue,” she said. “The day before, a senior developer needed six hours.”
More is coming down the pipeline. IBM plans to integrate Power11 with its Spyre AI accelerator chip later this year, promising even more AI horsepower. And by the end of 2025, the company’s open data lakehouse, watsonx.data, is expected to run on Power11 systems, aiming to simplify enterprise data architectures.
For enterprises looking to navigate the growing tide of AI, IBM’s Power11 servers might arrive just in July 25.
Key Takeaways
- Major upgrade: IBM’s first big Power line refresh since 2020
- AI focus: Designed for AI inference, not just raw compute
- Blazing performance: Up to 55% faster cores vs. Power9
- Sky-high reliability: Less than 30 seconds unplanned downtime per year
- Security boost: New ransomware detection in under a minute
- Cloud-ready: Available on-prem or via IBM Cloud
- Developer speed: Tools like watsonx Code Assistant accelerate coding tasks