The AI gold rush has a familiar rhythm: breathless demos, headline‑grabbing valuations, and a parade of tools promising to solve every problem with a prompt. Global AI spending is projected to exceed $2.5 trillion this year. But when the hype recedes and leaders are accountable for ROI, uptime, security, cost, and compliance, the question won’t be “Which tool did you buy?” It will be “Did you build something enduring?”

Tools are just that: instruments, not a comprehensive solution. At IBM, we’re not just selling tools – we're in the building business. And nowhere is that more evident than in how we approach mainframe modernization. There’s a reason we’ve seen sustained platform growth – more than 4x – over the last decade. Building enduring value requires not just the right tools, but also the right platform and the right approach.

Modernization Starts with the Right Platform

Modernization is not a COBOL translation problem, nor is it a one-time event. Translation is tactical, and modernization is an ongoing journey. It means continually taking advantage of new tools while protecting past investments and securing the environment. It also means simplifying an increasingly complex world of applications that continue to grow in scale, interdependencies, and workloads. Modernization must address the intricate ecosystem these applications run within – including tools like watsonx Code Assistant for Z and watsonx Assistant for Z, which help organizations navigate complexity, preserve institutional knowledge, and create intuitive, streamlined interfaces for both developers and users.

More than anything, modernization is about the entire stack – not just one piece of software. From the silicon and chip architecture, through the operating system, middleware, security frameworks, automation, and AI services, value comes from the way each layer is engineered to work together. For transaction processing, IBM Z remains the platform purpose-built to deliver unmatched efficiency, scale, performance and resiliency. Any other choice will result in tradeoffs.

We modernize at every layer in the stack to deliver superior outcomes. That stack allows our IBM Z systems to scale to tens of billions of encrypted transactions per day on a single system, deliver millisecond‑level AI inference at extraordinary volume, achieve industry‑leading availability, offer quantum‑safe cryptography, and run at sustained high utilization without surrendering SLAs. No amount of code conversion moves those characteristics with you. That’s platform design, not language semantics.

This power is why IBM created “mainframe modernization” in the first place: simplifying access to mainframe applications to achieve innovation and business agility, mitigate business risks and optimize their costs. Thanks to the work IBM’s been doing for years in the mainframe application modernization space, clients are extracting more value, not by migrating away from what already does the job best. In fact, the first generative AI‑powered code assistant we introduced was focused on IBM Z: watsonx Code Assistant for Z, back in 2023.

Since then, we’ve seen hundreds of clients select this technology and begin this type of modernization journey. Clients are also leveraging AI and analytics to extract more value from their data without the cost and risk of moving data off the mainframe, and it’s performing like clockwork. In one instance, Swiss Mobilar used SQL Data Insights software on IBM Z mainframes to achieve near-real time AI inferencing with 94% prediction accuracy, which boosted insurance quote closings by 7% in just six months.

We’re now in our fourth year of leveraging AI accelerators on IBM Z for transaction workloads, and our third year of leveraging AI for application development on the mainframe. The point isn’t to abandon your institutional logic; it’s to understand it, enhance it, refactor it, and connect it to the AI capabilities that will power the next decade of growth.

The Right Approach: Mainframe Modernization Goes Beyond Vibe Coding

Many enterprise IT teams follow the 80/20 rule: about 80% of application code works fine. It’s the remaining 20% of code that needs to be updated that unlocks business value. Targeting that 20% of coding on‑prem is faster, cheaper, and less risky than rewriting 100% of the code for a destination that wasn’t engineered for high‑volume, high‑integrity transaction processing. After speaking with many clients, we’ve learned they don’t want language migration. Instead, they want better access to core capabilities, like open APIs, integrations across hybrid cloud, and safe data sharing for model development. And, they want it without sacrificing performance, QoS, or resiliency.

On the mainframe, AI is a tailwind, not a tradeoff. It narrows the skills gap by helping developers – regardless of whether they grew up on COBOL, Java, Python, or Go – understand, explain, and improve mission‑critical code. It accelerates the adoption of modern, git‑based DevOps on IBM Z, enabling agentic CI/CD workflows and true continuous delivery. And it speeds up the hard parts of modernization: deep code comprehension, dependency mapping, performance tuning, and safe refactoring.

Success requires more than a frontier model guessing from source files. Enterprise applications often span tens of millions of lines and tens of thousands of artifacts. IBM’s approach fuses LLMs with the context that actually matters: deep static analysis across the entire application landscape, and dynamic runtime intelligence from z/OS that reveals real execution paths and performance hot spots. It’s this deep system insight that is uniquely used to enrich our AI software resulting in more accurate outcomes. Almost half of z17 clients have invested in new AI capabilities; with generative AI software on IBM Z up 200% year over year. This is what it takes to modernize without sacrificing resiliency or security.

Clients are already proving the pattern. Royal Bank of Canada is using AI‑assisted understanding to blueprint modernization across complex, interdependent systems. The National Organization for Social Insurance (NOSI) dramatically compressed the time to analyze superfluous COBOL routines, freeing teams to focus on change that matters. ANZ Bank is using modern DevOps on IBM Z to accelerate delivery while reducing manual operations. These aren’t “lift‑and‑wish” migrations; they’re measurable advances in speed, safety, and value – on platform.

The Future Belongs to Builders

There’s another misconception worth addressing, that a SaaS‑only solution can “modernize the mainframe” from afar. Given on‑premises dependencies, sovereignty requirements, and the sensitivity of core transactions, that proposition doesn’t hold up to real‑world governance or risk management. The pathway to sustainable growth is hybrid by design. Keep transactions on the platform built for them; integrate with cloud‑native capabilities where they create value; and use AI to bring both worlds closer, faster, and safer.

There’s a reason that the vast majority of the world’s credit card transactions run through IBM Z. While other industries have seen the rise of GPUs and other platforms, IBM Z systems remain the industry’s best solution for transaction processing – from silicon to operating system to middleware. You can try to replicate it elsewhere. You’ll pay for it in cost, performance, resilience, and security.

When the dust settles from the AI gold rush, those with true wealth will be the ones who didn’t confuse tools with outcomes. They’ll have partners who helped them modernize the right 20% of code, kept guarantees intact, and brought AI to the infrastructure that’s best for their business.

That’s IBM’s role: not just a seller of tools, but a builder with the right platform and approach for enterprise IT that endures. In the AI era, the world’s strongest transactional platform is only getting stronger.