Government Mainframes Versus DOGE: Showdown At The COBOL Corral

March 17, 2025 Alex Woodie

Elon Musk is a force of nature, slashing the federal payroll through the Department of Government Efficiency (DOGE). But the government’s IBM mainframes are immovable objects in their own rights, lasting the test of time. If Musk and DOGE are serious about reforming how the government does business, the government’s legacy systems will pose a serious challenge.

As the head of DOGE, Musk has ruffled plenty of feathers with his comments and his actions, including firing federal employees en masse and wielding a large “chainsaw for bureaucracy” given to him by Argentine President Javier Milei, an ally of President Donald Trump who has shrunk the size of his government.

By some counts, more than 200,000 federal employees have been fired or accepted early retirements in the eight weeks since President Trump has been in office and set Musk and DOGE loose on the bureaucracy. But that pace of change hasn’t been fast enough for Musk, who is a famously impatient CEO who has found great success with his “move fast and break things” management style at Tesla, SpaceX, and X.

In February, Musk ridiculed the Office of Personnel Management’s (OPM) and its storage of paperwork in an Iron Mountain facility housed in a large, abandoned limestone mine in rural Butler County, Pennsylvania, saying it slowed down the pace of federal firing.

“The speed at which the mine shaft elevator can move determines how fast people can retire from the federal government, and if the elevator breaks down sometimes then nobody can retire,” Musk said at a White House event last month. “We were told the most number of people that could retire possibly in a month is 10,000 because all the retirement paperwork is written down on a piece of paper, then it goes down a mine.”

The OPM does store paperwork used to process federal employee retirements at that facility, a former OPM manager who was the Retirement Operations Center director told CBS News. However, the federal portion of the facility is located on the same level as the entrance, he said. “I’m not aware of an elevator in the mine,” he said.

Elon Musk waves the “chainsaw of democracy” given to him by Argentine President Javier Milei.

If Musk thought the limestone mine represents a “time warp” back to a 1950s office, then he likely has similar thoughts about the various IBM mainframes scattered around Washington DC. These mainframes generally run applications written in COBOL, the language that started as a joint IBM-Defense Department project in the late 1950s.

Fast forward nearly 80 years, and a lot of that COBOL is still running the government’s day-to-day business. But while those old COBOL apps are still running, it would be a stretch to say they’re running well. The size of the government has grown tremendously since the 1950s, when governments of all sizes started installing IBM mainframes to automate business processes. In many cases, those old applications have not kept up with the times.

Cracks in the COBOL wall began to appear in earnest during the COVID-19 pandemic, when IBM mainframes used by various state employment offices were called upon to process millions of unemployment claims. The result was predictable: a nearly impenetrable backlog due to slow processing times on overwhelmed mainframes. That gave us headlines like this one in The Verge: “Unemployment checks are being held up by a coding language almost nobody knows.”

It’s not COBOL’s fault, of course. And lots of people know it (although skills are definitely a growing issue). The fault lies with the folks in charge of hiring COBOL programmers and designing the systems to scale, and partially with the organizations that have failed to adequately train the next generation of COBOL programmers, as well as the RPG programmers who make the IBM i server run.

“Functionality, we have not seen it as a big problem. Scalability is definitely a problem,” says Miten Marfatia, the CEO of Evolveware, a Silicon Valley company that provides COBOL code migration software and services for mainframes and IBM i shops. “It wasn’t the functionality that was the problem. I think the applications could not stand the load, so it was more of an architectural problem.”

Marfatia has worked with dozens of banks, healthcare companies, and government agencies over the years to migrate their COBOL apps. He’s seen the Social Security’s infamous COBOL application, which has been a candidate for modernization for decades, only to keep on chugging along. That particular COBOL app made the news when DOGE discovered that its data store listed tens of millions of Americans as being age 100 or higher, many in excess of 150 years of age, and some nearly 300.

The Federal Government stores records in this Iron Mountain facility in an old limestone mine in Butler County, Pennsylvania.

“They should be on the Guinness Book of World Records. They’re missing out,” Musk joked in his White House appearance. “I think they’re probably dead. That’s my guess. Or they should be very famous. One of the two.”

The government faces a number of challenges when it comes to modernizing legacy systems, Marfatia said. For starters, the vendors that service them tend to be very large and promote solutions that require a large number of people to implement. “Number two is their proposals are quite exorbitantly expensive,” he says. “From that standpoint, I do believe that the new administration might make some substantial changes there.”

Eric Kimberling, the CEO of Third Stage Consulting, has seen his share of bloated technology proposals for ERP migrations and digital transformation projects. Upward of 80 percent of these projects fail, Kimberling says, often because the buyer is too infatuated with new technology and doesn’t pay enough attention with the “soft skills,” such as planning, change management, and setting realistic expectations.

When it comes to running large IT projects, like modernizing mainframes or IBM i-based systems, the government has a few things going against it, Kimberling says.

“I think government in general, as a broad generalization, is more challenged when it comes to change in general,” he tells IT Jungle. “There’s obviously regulatory and political forces that that sort of slow things down in government entities. There’s limited budgets, there’s limited competencies and skill sets and just sort of an overall – I don’t want to say resistance to change, as in, it’s intentional. It’s just that backdrop is not conducive to being on the cutting edge of technology.”

“And then you add that to the fact that so many of these government initiatives are such big failures and big money pits that I think it leaves a lot of government agencies thinking that they could probably better spend that money or better spend those resources somewhere else,” he continues. “I’m not saying it’s right or wrong. It’s just I think that’s the dynamic we see in the government space.”

In the private sector, many companies are attempting to move their old mainframe and midrange applications to newer Web- and cloud-based architectures. They want to break up the functions contained in monolithic applications and re-implement them as a large number of smaller microservices, often running in a containerized (i.e. Kubernetes) environment. Some customers are moving to this architecture, even if they’re planning to keep running on-prem.

While that’s the state of the art for enterprise applications in the private sector, refactoring giant COBOL or RPG apps as containerized microservices running in the cloud may not make a lot of sense for government agencies, Kimberling says.

“I think the problem is a lot of government entities think ‘Well, if I’m going to modernize, I’ve got to modernize. I’ve got to go all in on cloud and multi-tenant, SaaS or whatever,” he says. “And you don’t have to. I mean, the vendors are going to try selling you that, but that’s not necessarily what you need to do.”

Instead, governments would be better served by playing it safe, he says. They should plan for incremental change rather than going for the big bang, and focusing on some of the common gotchas that can sneak up in any digital transformation project, such as ensuring effective communication and managing change effectively.

“If you’re a government entity, you should play it safe. You should be low risk,” he says. “Maybe government will start to think more that way as a result of fear [of DOGE]. But in general, I don’t think government entities need or should or would want to swing for the fences and go for the most modern bleeding edge technology if they’re five years behind.”

Technology has improved remarkably over the past decade, Kimberling says, particularly when it comes to data analytics and AI in the cloud. Private-sector companies rightfully are looking to take advantage of the new capabilities. But moving too fast can be dangerous for an entity like a government agency.

“The technology has definitely gotten better, but the ability to implement, I would argue, has gotten worse, because now the technology has changed so much that the divide between where we are today and where technology is . . . that hill becomes steeper,” he says, “especially if you’re trying to jump too far ahead of yourself, and jump from an old mainframe homegrown system to now suddenly an AI-driven, cloud-based system. That’s a huge, huge jump, and the risk is exponentially higher than if you’re just taking a more incremental approach.”

Whether you’re looking to modernize a large government mainframe system or an RPG system for a midsized company, the lessons of past failures remain the same. Careful planning, aligning the project with business outcomes, and setting reasonable expectations will increase the odds of success.

RELATED STORIES

COBOL Is Not The Problem; The Data Is

The IBM i And Its RPG Decade Of Crisis

Modernization Starts with the Business, and the Tech Follows