DOGE Watch · Federal IT · Legacy Systems · 10 Sources
$100B+
Annual federal IT spend
~80%
Goes to legacy systems
50+ yrs
Age of some systems
§ DOGE Watch / Federal IT Legacy Systems

The Federal Government Spends $100 Billion a Year on IT. Most of It Keeps 1960s Systems Running.

§ 01 / The Numbers

The federal government spends approximately $100 billion per year on information technology, according to OMB’s IT Dashboard. GAO has consistently found that roughly 80 percent of that spending goes to operating and maintaining existing systems — including systems that are decades old — leaving only about 20 percent for modernization and new development. That ratio has not changed meaningfully in fifteen years.

Some of those legacy systems are genuinely old. GAO identified in 2016 — and reconfirmed in subsequent reports — that several critical federal IT systems were running on hardware and software that was 40 to 50 years old. The Social Security Administration relies heavily on COBOL, a programming language whose primary use peaked in the 1980s. The IRS runs COBOL-based mainframe systems that date to the Kennedy administration. The Department of Defense operates 8-inch floppy disk-dependent nuclear coordination systems — a fact that briefly went viral in 2016 when it was disclosed publicly.

What This Means
$100 billion per year, 80 percent of it spent keeping old systems alive, and a GAO High Risk designation that has been continuous for over a decade. This is not a DOGE discovery — GAO has been reporting on it since at least 2002. What DOGE added is visibility. Whether the solution is aggressive modernization, consolidation, or some combination depends on agency by agency analysis. What is not in dispute: the current allocation — 80 cents of every IT dollar keeping 1960s code running — is indefensible as a long-term strategy.
§ 02 / Why It Persists

Federal IT modernization is hard for reasons that are structural, not merely political. Procurement rules require competitive bidding processes that favor large established vendors, adding 18–36 months to the front end of any project before a line of new code is written. Program managers who initiate modernization projects rarely survive long enough to see them completed — the average tenure of a federal CIO is under two years. And the systems being replaced are load-bearing: the IRS mainframe processes trillions of dollars in tax transactions annually. A failed migration is not just embarrassing; it can mean millions of Americans don’t receive refunds.

The result is risk aversion at scale. Agencies maintain legacy systems because replacing them is genuinely dangerous and because the incentives for program managers skew heavily toward not being the person who broke the IRS during tax season. DOGE’s contribution — accurate or not — was to make the maintenance cost visible. The $100 billion figure is real. Whether confronting it requires DOGE-style shock therapy or a long-term funded modernization plan is the real policy question.