Research Finder
Find by Keyword
Why Mainframes (and COBOL) Are Secretly Saving Your Tax Dollars
Mainframes and COBOL: Debunking Myths of Government Inefficiency
Recent media discussions have highlighted the critical role and continuous evolution of mainframes and COBOL in modern U.S. government operations, particularly within agencies like the IRS and Social Security. Far from being outdated relics, as some uninformed commentators have suggested, mainframes remain essential and efficient, a fact often overlooked in portrayals that focus solely on their age rather than their ongoing advancements and importance.
So why does COBOL and the Mainframe get treated differently?
The term "1960s COBOL mainframe" is a misnomer that does not reflect the reality of today's mainframe technology. Mainframe systems are updated annually, with a roadmap of three systems being worked on at any one time. The latest mainframe, the z16, has been enhanced with capabilities like sophisticated AI-driven fraud detection and quantum-safe encryption. The z16 Telum processor is based on 5nm technology and was designed in Poughkeepsie, NY and the fab that churns out these processors is also based in the US. The upcoming generation, Telum II, is set to integrate Generative AI algorithms directly into the hardware using the Spyre DPU, both showcasing a trajectory of innovation rather than stagnation. Check out my detailed breakdown of the Telum II processor announced last year here.
COBOL's Misunderstood Role
COBOL, often criticized for its age, is indeed a mature language but far from obsolete. With its strong procedures for data handling, especially post-Y2K, COBOL has shown resilience and adaptability. The language's use in mission-critical systems is due to its robustness, readability, and efficiency in handling large datasets, which are crucial for government applications like tax processing, social security, and banking systems.
Moreover, COBOL's ecosystem is vibrant with an active open-source community (Open Mainframe Project), open-source codebases such as ZOWE, and educational initiatives in over 200 universities worldwide. This community drives the language and mainframe platform forward, ensuring it remains relevant and on trend. Tools like IBM's watsonx Code Assistant and, BMC AMI DevX tools, powered by
GenAI-driven BMC AMI Assistant leverage AI to analyze, explain, and modernize COBOL codebases, making them more accessible and maintainable for new generations of developers. Vendors like Broadcom and Rocket Software also provide skills programs, one of which includes learning COBOL in a day, and software tooling to ensure the mainframe remains vibrant and current to trends like micro-services, DevOps and AI.
Efficiency and The Mainframe
Contrary to narratives suggesting inefficiency, mainframes offer some of the most efficient operational platforms in the industry. In a 2024 report commissioned by Broadcom the findings were clear. Moving mainframe workloads to the cloud could cost in excess of 5x more in total cost of ownership (TCO).
Furthermore, Mainframes currently account for 74% of the world’s transaction workloads yet only represent 8% of total IT spend. This cost efficiency stands out when contrasted with on-premise servers, which handle 13% of the world’s transaction workloads yet eat up a staggering 65% of total spend. Public cloud sits in between these two asset classes; it continues to prove its cost efficiencies, having grown to 5% of the world’s workloads at 14% of overall spend.
In the annual BMC Mainframe Survey, the company saw a platform on the rise. This year’s results show that 94% of respondents have a positive perception of the mainframe, with 62% citing it as a platform to grow and attract new workloads. This tracks the 62% who report adopting DevOps practices on the platform. It’s no surprise, then, that two of the top five priorities cited by respondents are efforts to enhance automation and modernize applications. It is true, some workloads will benefit by moving off the mainframe into the cloud or other distributed platforms, but the data and processing power will need to stay on the mainframe as it is by far the most efficient and cost-effective platform.
Another datapoint I found while researching this article was that by leveraging DevOps, organizations can enhance existing applications with minimal code changes. This approach reduces costs by 85% compared to application rewrites while maintaining scalability, performance, and reliability.
The Reality of Government IT Infrastructure
The U.S. government's IT infrastructure has simply not operated on systems untouched since the 1960s. There are no ‘Hidden Figures’ era, room sized, mainframes left in production. The mainframes of today are all in industry standard 19” racks. Instead, there's a continuous cycle of updates and maintenance. The average age of server estates in government departments like the IRS is likely under a decade, aligning with industry norms for UNIX or large x86 deployments. If we want genuine transparency then vendors or the government should publish the server that these systems are running on, they won’t, but if any US government is running on a system older than z13 (the current is z16) I would be shocked.
AI, Blockchain and Modernization Efforts
AI's integration into mainframe environments is not just about updating old code; it's about transforming how these systems operate. AI tools can now quickly decipher complex, legacy monolithic codebases, predict maintenance needs, automate routine tasks, and even assist in real-time decision-making processes. This not only addresses inefficiencies but also propels these systems into realms of predictive and adaptive computing, which are essential for modern governance.
In addition to AI, other cutting-edge technologies like blockchain could further enhance government system capabilities. Blockchain could be used for securely logging and verifying transactions in government databases, ensuring data integrity and transparency. Heck you could even run the blockchain on a Linux partition on the mainframe and take advantage of the highest level of secure key management available. Mainframes are certified to FIPS 140-2 Level key management, where cloud and x86 systems top out at level 3 typically. There are numerous examples of Blockchain innovators running confidential computing environments on its platforms which further demonstrates the flexibility and ability for this chip architecture to support cutting-edge applications and workloads.
Mainframe and the Cloud
Mainframes connect to the cloud. Moreover, the integration of cloud services with mainframes, a hybrid cloud approach, allows for the flexibility of cloud computing while maintaining the robust security and data handling of mainframes. Various options exist to create RESTful API’s to mainframe systems with one example being z/OS Connect. These API approaches enable applications on other platforms to seamlessly access the mainframe and have these robust data servers be a fully paid up member of a hybrid cloud architecture.
Mainframe Skills
Sourcing cutting edge IT skills is hard. Getting AI skilled data scientists is hard, Sourcing skilled cyber security professionals is hard. These roles command a premium, with salaries often in the $500,000 range for the most skilled engineers. According to a 2024 report by the Futurum Group, the mainframe sector has seen a noticeable 65% rise in skilled workers over the past 5 years. There is strong intent among employers to bolster their teams, with 91% planning to hire new mainframe talent within the next 2 years, reflecting the ongoing critical need for mainframe expertise in modern IT infrastructures. The mainframe skills landscape is challenging for sure, but schemes exist such as Broadcom’s ‘Beyond Code’ program and the IBM Mainframe Skills Council to address the short term challenges, and the long term trends look positive with the worst skills shortages in the rear view mirror.
Furthermore, the development of skills among government IT staff is crucial, regardless of the underlying technology platform.. Training programs focusing on both COBOL and contemporary languages, along with understanding how to bridge these technologies, can ensure that government IT remains at the forefront of efficiency and security. This holistic approach to IT management underscores the need for a narrative that appreciates the complexity and adaptability of government systems, rather than simplifying them into outdated stereotypes.
Energy Efficiency and the Mainframe
As the US looks at rampant AI-fueled growth and the impact on the energy grid it is worth noting that Mainframes offer improved sustainability. When you consider energy efficiency and consolidating workloads, mainframes achieve high utilization rates (over 90%), and therefore consume 30-50% less energy per workload compared to distributed server systems, though their total power draw and cooling needs remain significant.
Modern mainframes, like the z16, are upgradable and compact, fitting standard racks while using advanced power management and liquid cooling to optimize energy use. Software optimizations, including virtualization, containerization, and AI-driven resource allocation, further enhance efficiency, though legacy applications may offset gains if not updated. Mainframes integrate with hybrid cloud systems, supporting a circular economy by minimizing infrastructure sprawl. Overall, while not without environmental costs, mainframes provide a sustainable computing option due to their durability, efficiency, and liquid cooling capabilities.
Looking Ahead
While the last few paragraphs may be seen as a love letter to the mainframe, that is not the case. Certain workloads should not be on the mainframe, they should be moved to other platforms. If a workload is more cost effective on the cloud, put it there. The analogy I always use is transporting goods. If a flat bed truck is cheaper and more suited to the goods being transported, then use a flat bed truck. However there will also be times when air freight is a better option, or putting the container on a boat is the best option. Put another way - just because an F250 exists doesn’t mean that a train is obsolete.
The premise that mainframes and COBOL lead to inefficiencies in government operations is largely a myth fueled by a lack of understanding of modern IT infrastructure. Mainframes, with their continuous updates, robust security features, and integration with cutting-edge AI, are not just avoiding obsolescence but are at the forefront of technological innovation in government IT.
Also crucially these systems are designed, fabricated and manufactured in the US of A, by a US based company. The software that runs and operates these systems is made by US based companies such as Broadcom, BMC and Rocket Software amongst many others.
The challenge isn't the technology itself, the age of the programming language, but ensuring that these systems are managed, updated, and utilized with foresight and expertise. As we move forward, the focus of those looking to drive efficiency should be on leveraging these platforms' strengths while continuously modernizing the applications that run on them, ensuring they serve the public with the efficiency and security they demand.
Steven Dickens | CEO HyperFRAME Research
Regarded as a luminary at the intersection of technology and business transformation, Steven Dickens is the CEO and Principal Analyst at HyperFRAME Research.
Ranked consistently among the Top 10 Analysts by AR Insights and a contributor to Forbes, Steven's expert perspectives are sought after by tier one media outlets such as The Wall Street Journal and CNBC, and he is a regular on TV networks including the Schwab Network and Bloomberg.