The launch of the IBM z17 this year represents a huge leap forward for the practical deployment of artificial intelligence within the enterprise. Built on decades of mainframe reliability, the z17 integrates new processors, accelerators and AI capabilities that help enterprises make real-time decisions, cut costs, automate operations and improve security.

Here is a reminder of five reasons why the latest mainframe deserves our attention.

1. Bringing AI to where the data is

Instead of having to transfer sensitive information - such as customer or financial data - out to the cloud for AI inferencing, the IBM z17 lets you run AI on the platform, alongside your enterprise data. Companies can benefit from the security of the mainframe environment while avoiding the latency of moving data between systems.

The result is faster decision-making, lower risk of breaches and stronger compliance.

This is thanks to the IBM z17's Telum II processor and the complementary Spyre Accelerator chip, which help to improve performance and provide more efficient handling of data-intensive workloads. They combine to deliver high-speed AI inferencing with low latency to support generative AI and large language models (LLMs) at enterprise speed and scale.

With the vast amounts of mission-critical data held on mainframe systems, there a number of key reasons why IBM customers will prefer running AI inferencing on-platform over outsourcing to a separate system:

• Data locality: by avoiding data transfer to external systems such as cloud services, you cut network bottlenecks and enhance performance.
• Data privacy: processing AI tasks on the highly secure Z environment helps maintain compliance with data protection regulations.
• Better resource utilization and cost efficiency: by using existing mainframe infrastructure for AI tasks, you avoid the additional cost of setting up separate AI processing systems.

Another important reason is the need for real-time processing. By supporting AI directly on the mainframe, you enable real-time data analysis and decision-making even for the massive data and transaction volumes that typify mainframe workloads (and you can do this using multiple AI models as explained next).

2. Enabling real-time AI decisions with multiple AI models

The IBM z17 enables multiple AI models to run in parallel to make real-time decisions with results delivered in milliseconds. Also known as Ensemble AI, this is another breakthrough made possible by the system's advanced on-chip processor and the Spyre Accelerator.

Ensemble AI provides the power to unlock more value from transactional AI by using multiple models to improve accuracy and reduce false positives - ideal for applications such as enhanced fraud detection, anti-money laundering, anomaly detection and sensitive document summarization.

For instance, in insurance claims fraud detection, Ensemble AI techniques can combine the strengths of multiple models to improve accuracy and performance. Traditional machine learning models provide an initial risk assessment for a claim; LLMs then enhance the analysis by processing unstructured data, such as claim descriptions or supporting documents. In this way, the complementary capabilities of different AI models work together to detect fraudulent claims more effectively.

3. Delivering energy, space and cost savings

The z17 delivers superior performance while dramatically reducing energy use and cost. Not only does it enable real-time AI inferencing, it does this while minimizing energy impact, cutting power consumption by up to 83% compared with using standard x86 servers for AI-enhanced online transaction processing workloads.

In addition to using 17% less energy than its predecessor, the IBM z16, the z17 can consolidate the workloads of hundreds of x86 servers while consuming only a quarter of the power. It can reduce energy use and data center space while achieving up to 44% lower Total Cost of Ownership (TCO) over 5 years with cloud-native, containerized workloads.

In locations where square footage is expensive and tightly regulated, IBM clients plan to consolidate x86 server racks into z17 systems to free up space and cooling capacity.

4. Bringing AI-powered self-management to Z

The IBM z17 uses its own AI capabilities to monitor logs, detect anomalies and even recommend or implement fixes. This is increasingly important as the older generation of mainframe engineers retires. Over time, it can potentially mean less downtime, faster recovery and lower maintenance costs.

For example, the z17 uses IBM Concert for Z an AI-powered IT operations management tool that uses machine learning to spot problems early by identifying deviations from normal system behaviour. It provides precise remediation recommendations with generative AI and built-in IBM Z expertise, ensuring faster resolution of complex issues.

The IBM Spyre Accelerator will enable IBM watsonx Assistant for Z to run natively on the IBM z17. This is an AI assistant that provides users with accurate, up-to-date self-service answers to their IBM Z queries - helping to simplify the execution of both complex and repeated tasks. Importantly, it can 'codify' the knowledge of IBM Z experts into automations, giving authorized users the option to trigger automations to complete certain tasks.

5. Supporting advanced security capabilities

The latest mainframe is packed with advanced AI technology to secure and protect enterprise data. For example, IBM Threat Detection for z/OS will use AI to detect and identify potentially malicious anomalies that could be the result of cyber-attacks. It will issue alerts about suspicious incidents and include a quarantine facility and dashboard to help analyze and diagnose them.

There is also a new AI solution to discover and classify sensitive data on the platform. Using natural language processing and other newly created AI techniques, it can identify and protect confidential information before using it in the AI data pipeline. By using automated tagging and classification with AI, enterprises will be able to avoid labour-intensive error-prone manual data classification.

Looking ahead, the z17 also addresses the emerging risk of quantum computing. Encryption that would take today's computers centuries to break could be cracked in a matter of hours by quantum systems. To prepare, the z17 includes quantum-safe cryptography certified by the National Institute of Standards and Technology (NIST).

All of this underscores by the IBM z17 is the ideal solution for enterprises that process large volumes of mission-critical data to integrate AI into the core of their operations.

This blog was originally published on the IBM Community.