Flat caps, rising subscriptions: how the debt deal squeezed federal IT from the inside

The week

Over the Memorial Day weekend in 2023, President Biden and House Speaker Kevin McCarthy reached an agreement to raise the federal debt limit, a legal ceiling on how much the federal government can borrow to pay for spending Congress has already approved. For weeks Washington had been counting down to a date nobody wanted to reach: the Treasury had warned that, without action from Congress, the government would soon run out of cash to pay its bills.

The standoff had been building since January 2023, when federal debt reached the statutory limit of about $31.4 trillion. From then on the Treasury relied on what it calls extraordinary measures, accounting steps that let it keep paying bills for a while without new borrowing. In April the House passed its own bill pairing a debt-limit increase with deeper spending cuts. On May 26 Treasury Secretary Janet Yellen told Congress the department could run out of room as early as June 5. The negotiated deal split the difference between the House bill and the White House position.

The House passed the Fiscal Responsibility Act on May 31, 2023, the Senate on June 1, and the president signed it on June 3, days before the date the Treasury had said it could run out of cash. The votes were bipartisan in both chambers, 314 to 117 in the House and 63 to 36 in the Senate, with opposition from both parties’ wings.The crisis was averted. The price of the agreement was a set of limits on what the government could spend next.

The law suspended the debt limit until January 2025. It set caps on discretionary spending for fiscal years 2024 and 2025. Discretionary spending is the portion of the budget Congress decides each year through appropriations bills, as opposed to programs like Social Security that run on autopilot under existing law. Agency operating budgets, including almost all federal IT, come from this discretionary pot.

The statute put numbers on it. For fiscal 2024 it set the nondefense limit, which the law calls the “nonsecurity” category, at about $704 billion. The defense, or “security,” limit was about $886 billion. For fiscal 2025 the figures rose to about $711 billion and $895 billion, growth of roughly one percent in each. The law also listed figures for 2026 through 2029, but only for congressional budget procedures, without the automatic cuts that backed the first two years.

The caps held nondefense spending roughly flat in the first year with a small increase in the second. The law included a mechanism to discourage long continuing resolutions by lowering the caps if full-year appropriations were not enacted. A continuing resolution is a stopgap that keeps agencies running at roughly last year’s levels when Congress has not passed a new budget on time. The mechanism was designed to make lengthy stopgaps unattractive.

The detail mattered. If any part of the government was still on a continuing resolution on January 1 of either fiscal year, both caps would drop to one percent below the fiscal 2023 level. The cut would be enforced across the board, through a sequester calculated by April 30. Enacting full-year appropriations for every account before that report would cancel it. Because the penalty hit defense as well as nondefense spending, it gave both parties a reason to finish on time.

It also rescinded some unspent COVID-19 relief funds and part of the additional IRS funding provided in 2022. To rescind is to cancel money Congress had previously provided but agencies had not yet spent.

The IRS rescission written into the statute was about $1.4 billion. A separate, informal understanding between the negotiators, widely reported at the time, contemplated repurposing roughly $20 billion more of the IRS money over two years. The law also ended the pandemic pause on federal student loan payments and amended the National Environmental Policy Act to speed permitting reviews. It added an “administrative pay-as-you-go” rule requiring agencies to offset executive actions that raise mandatory spending.

The headlines were about default and politics. For the people who run government technology, the more lasting news drew less attention: two years of flat budgets were coming, and the way IT costs behave meant that flat would not feel flat.

For agency technology leaders, the deal mattered less for the headline than for what it implied about the next two budgets. Nondefense spending would be roughly flat in nominal terms while inflation was still well above the Federal Reserve’s target, which meant a real decline. “Nominal” means the dollar figure on the page; “real” means what those dollars can actually buy. A budget that stays the same while prices rise buys less each year.

The gap was not small. The Bureau of Labor Statistics reported consumer prices up 4.9 percent in the year to April 2023. Technology contracts are not indexed to the consumer price index, but vendor list prices, cloud rates for some services and contractor labor rates were all moving up that year. A program office whose appropriation stayed level was, in effect, being asked to absorb those increases inside its own line.

Under a flat budget, technology spending does not shrink evenly. It sorts itself by how each line behaves. Some costs grow on their own. Some investments can be postponed, so they are. Some capacity erodes without anyone deciding it should. Each behavior calls for a different control, and together they account for the whole IT budget. The pattern is not peculiar to the federal government: any organization facing a flat or falling technology budget will recognize them.

Where federal IT money comes from

Congress provides money to agencies through appropriations, and that money generally comes with conditions on what it can be used for and how long it is available. Agency chief information officers, or CIOs, oversee technology, but the money flows through many programs and offices. Chief financial officers, or CFOs, manage the budget as a whole.

Inside the IT budget there are two broad kinds of spending. Run costs keep existing systems operating: hosting, licenses, support contracts, help desks, security tools. Modernization spending replaces or rebuilds aging systems. In household terms, run costs are the utility bills and the car payments; modernization is the kitchen renovation. When money is tight, the bills get paid and the renovation waits.

That is the heart of the problem the 2023 deal sharpened. The bills were no longer stable, and the renovation was already overdue.

Rules on when money can be spent

Federal money also comes with time limits. Most operating funds are one-year appropriations: they must be obligated, meaning legally committed through a contract or order, in the fiscal year for which Congress provided them. Some accounts carry multi-year or no-year money, which gives a program more room. The “bona fide needs” rule requires that a year’s money pay for that year’s needs, so an agency cannot simply prepay several years of a subscription out of one appropriation.

The Antideficiency Act sits behind all of this. It bars agencies from obligating money before Congress appropriates it, or beyond what is available. Violations must be reported to the president and Congress, so finance offices treat the limits seriously. OMB then apportions each appropriation, releasing it in portions over the year. A CIO can have an approved budget and still be unable to sign a contract until the apportionment arrives.

GAO has reported for years that most federal IT spending, around 80 percent by its estimates, goes to operating and maintaining existing systems. That ratio is the starting point for everything below. When four of every five dollars are already spoken for, a flat total leaves little room to absorb growth anywhere.

Who the caps did and did not bind

The caps did not fall evenly across government. Defense IT sat under the security cap, which rose a little more than nondefense. Agencies that had received large multi-year sums from the 2021 infrastructure law or the 2022 Inflation Reduction Act held money that did not depend on each year’s appropriations bills. For them, the binding constraint was often staff to spend the money well, not the money itself.

Some agencies sit outside appropriations entirely. The Federal Reserve, the FDIC, the OCC and the NCUA fund themselves from assessments, fees or earnings, so the caps did not limit their technology budgets directly. A bank examiner’s IT shop and a civilian benefits agency’s IT shop therefore faced very different years. Comparisons across agencies need to account for where the money comes from before drawing conclusions about how well each managed.

State agencies felt the federal caps through grants. Medicaid is the clearest case: federal matching rates for state Medicaid IT are higher for building systems than for running them, commonly 90 percent for development and 75 percent for operations. Once a system goes live, more of its cost lands on the state. That structure creates its own version of the run-cost squeeze, independent of whatever Congress does to the discretionary caps.

Costs that grow on their own

A decade earlier, much federal IT spending bought hardware and perpetual licenses, followed by fairly stable maintenance. With a perpetual license, an agency pays once and owns the right to use the software, with a modest yearly fee for updates. By 2023 a larger share went to cloud services and software subscriptions, which are billed by the month and by use. Cloud computing means renting computing power and storage from a provider’s data centers instead of owning the machines.

Those bills recur every year and rise with use. Storage accumulates. Compute expands with new workloads. Per-user licenses follow headcount and new features. Renewals arrive with price increases. Nobody has to decide to spend more; it happens because the organization keeps working.

Security spending added its own growth. Logging requirements under OMB’s 2021 memorandum M-21-31 raised the volume of data agencies had to collect and retain, and storage and analysis tools were often priced by volume. A good security rule came with a meter attached.

M-21-31 set out tiers of logging maturity, labeled EL0 through EL3, and required agencies to keep certain logs for twelve months in active storage and a further eighteen months in cold storage. Each tier added log types and analysis capability. The memo did not come with dedicated funding, so agencies largely paid for the extra ingestion and storage from existing budgets.

Security teams had choices about how much of that meter to run. Not every log source carries equal value for detection or investigation. Identity provider logs, endpoint telemetry and logs from internet-facing systems usually justify full retention in a searchable tier. Verbose application debug logs often do not. Agencies that mapped each source to the tier the memo actually required, and filtered or summarized the rest before ingestion, met the requirement for less money. Those that sent everything to a premium analytics platform paid for volume that rarely answered a security question. The decision belongs jointly to the CISO and whoever owns the cloud bill, because each can see only half of the trade-off.

Why consumption billing fits poorly into annual appropriations

Cloud billing and appropriations law were designed around different assumptions. A cloud provider bills for what was used last month. An agency must obligate funds before the service is delivered. To bridge the gap, contracting officers typically obligate an estimated amount on a task order and track usage against it.

When usage runs ahead of the estimate, the agency has a problem. It must add funds through a contract modification, shift money from elsewhere, or throttle the service. Continuing to run past the funded ceiling risks an Antideficiency Act violation. Some agencies responded by over-obligating as a cushion, which ties up money that could have funded something else. Others set tight ceilings and accepted the risk of a mid-year scramble. Neither choice is free.

Federal law does let agencies buy severable services, such as hosting or support, for a period of up to twelve months that crosses fiscal years. That flexibility helps align a cloud period of performance with a renewal date. It does not solve the deeper mismatch, which is that consumption is variable and appropriations are fixed.

Putting a meter on the meter

Under a flat budget, every dollar of automatic growth comes out of something else. The discipline usually called FinOps, short for financial operations, addresses this. It brings finance, technology and program staff together to manage cloud spending by establishing who is using what and stopping payment for what nobody uses.

The practices are concrete. Tag cloud resources to the programs that use them and show each program its costs. Set budgets and alerts. Rightsize instances, meaning shrink virtual servers that were set up larger than they need to be. Move old data to cheaper storage tiers. Use committed-use discounts where demand is predictable, accepting less flexibility in return. Reconcile software licenses against actual use before each renewal.

Each practice has a cost. Tagging needs an enforced standard and someone to chase untagged resources. Moving logs to cold storage cuts storage spend but slows an investigation that needs old records quickly. Rightsizing saves money until a seasonal workload, such as tax season or open enrollment, needs the capacity back. A FinOps program that ignores these trade-offs tends to produce savings on paper and outages in practice.

Evidence an auditor would ask for

Inspectors general and GAO auditors tend to ask for things that can be counted. Useful measures include the share of cloud spend tagged to a program and the number of idle or orphaned resources found and removed. Others are license utilization before and after each renewal, and unit costs such as cost per case processed or per claim handled. Unit costs matter because a rising bill can be healthy when volume rises with it. A rising bill with flat volume is the warning sign.

Suppose a mid-sized federal program office’s cloud bill has risen every quarter for two years. Nobody approved the increases; they came from new test environments that were never switched off, years of stored records, and licenses for staff who had moved on. Once resources are tagged, the program manager can see that a chunk of the bill belongs to projects that ended. Turning those off returns money to the mission rather than cutting it.

The appropriations constraint

Federal appropriations law adds a constraint. Agencies generally cannot commit funds beyond what has been appropriated. Multi-year discounts therefore have to be structured as options or within available funding, and that limits some savings commercial buyers take for granted. A private company can sign a three-year deal for a better price; an agency often cannot promise money Congress has not yet provided.

Contract terms can still help. Caps on renewal price increases, notice of consumption spikes and the right to reduce seat counts at each anniversary are all negotiable. These are unglamorous clauses, and they are where a surprising amount of money is won or lost.

Options deserve care. Under the Federal Acquisition Regulation, an agency can award a base year with priced option years and exercise each option only if funds are available. Pricing the options at award locks in rates without committing future money. Agencies that let the vendor reprice at each option lose most of the benefit. Multi-year contracts with cancellation ceilings exist as well, though they are used less often for software.

A single metric shows the pressure. Compare the growth rate of run costs with the growth rate of the IT budget. The gap is the amount modernization loses each year if nothing changes. If run costs grow faster than the budget, the renovation fund shrinks automatically, every year, without anyone ever voting to shrink it.

Large departments and small agencies feel this differently. A cabinet department with a central cloud contract can pool commitments and negotiate volume discounts across bureaus. A small independent agency often buys through a governmentwide vehicle at list-adjacent prices and lacks a dedicated cost analyst. The small agency’s percentage exposure to price increases is usually higher, and its capacity to resist them is lower.

Investments that get deferred

Modernization projects run for years. Replacing a legacy financial system, case management platform or benefits system can take five years or more. “Legacy” systems are older systems still in use, often running critical work, often expensive and hard to maintain, and often understood by a shrinking number of people.

Under flat caps they become the residual: what is left after run costs and mission priorities are paid. Continuing resolutions make it worse. They typically fund agencies at prior-year rates and restrict new starts, which can delay awards and stall work already planned. A project scheduled to begin in October may sit idle for months while the agency waits for a full-year budget.

How a continuing resolution stalls a project

The new-start restriction is narrower than it sounds and more damaging. A “new start” is an activity that was not funded the year before. A modernization phase that was planned and approved, but never funded, counts as one. OMB often apportions CR funding in fractions tied to the CR’s length, so a program may receive only a few weeks of money at a time. Agencies can ask Congress for “anomalies,” exceptions written into the CR, but those are reserved for urgent cases and rarely cover a routine modernization phase.

When the full-year budget finally arrives in late winter or spring, the agency has roughly half a year to obligate a full year of money. Contract actions pile up in the final quarter. Contracting officers become the bottleneck. Rushed awards are more likely to have weak requirements, and protests can push them further. The work that slipped in October often slips again because the procurement could not be run properly in the time left.

Modernization also carries a cost that plans often understate. During migration, old and new systems run side by side, so run costs rise before they fall. The agency pays for both until the old one is switched off. If the old system is never fully decommissioned, the savings never arrive. Treating decommissioning as a funded deliverable with a date, rather than an assumption, is one of the more reliable ways to protect a modernization budget.

Deferral also has a defensible logic, which is why it persists. A legacy system that works is a known quantity. A replacement carries delivery risk, and many federal modernization efforts have overrun. A budget officer choosing between a guaranteed run cost and an uncertain project can reasonably pick the run cost for one more year.

The case against repeated deferral rests on accumulating risk. GAO’s 2019 review of federal legacy systems identified ten critical systems most in need of modernization, several running on outdated languages or unsupported hardware. Each year of delay makes the staff who understand those systems older and scarcer, and makes vendor support harder to obtain. The cost of deferral is real; it simply does not appear in any one year’s budget.

Tools that soften the blow

Agencies had some tools. The Technology Modernization Fund provided funding for projects, usually with repayment expectations. It received $1 billion from the American Rescue Plan in 2021, and its board later relaxed repayment terms for some awards. IT working capital funds authorized by the 2017 Modernizing Government Technology Act let some agencies carry money across years, which matters because many appropriations expire if not used within a set period.

Modular contracts that deliver usable capability in each phase reduce the damage when funding pauses. Instead of one giant contract that only pays off at the end, the work is split into pieces, each of which leaves the agency better off on its own. If the money stops after phase two, phases one and two still deliver something.

Modular contracting has costs too. More, smaller contracts mean more procurement actions, more integration work between vendors and a heavier load on the government team that coordinates them. An agency without architects to hold the pieces together can end up with several functioning modules that do not work well as a system. The approach pays off when the agency keeps enough technical staff to own the overall design.

A modernization plan that assumed steady annual increases was fragile. A sturdier plan named the phases that could proceed on flat funding and defined a usable stopping point if the money ran short.

Capacity that erodes by attrition

The third behavior concerns people. Flat budgets slow hiring, and federal technology talent was already hard to recruit against private-sector pay. Staff numbers then fall through attrition, without a decision. People retire or leave and are simply not replaced. The gap is filled by contractors, or by nobody.

Attrition also falls unevenly. One vacancy in a two-person cost management team halves its capacity, while the same vacancy in a large service desk barely registers. Leaders who look only at total headcount will miss this, because the total barely moves while a critical function hollows out.

Why the hiring system amplified the effect

Federal hiring is slow even in good years. Positions must be classified, announced and filled through processes that often take months. General Schedule pay is set governmentwide and lags private-sector technology pay, especially for cloud engineers and security specialists. Tools such as direct-hire authority and term appointments help at the margin, but they do not change the budget arithmetic. When a hiring freeze or a flat budget pauses backfills, the time to replace a departing specialist can stretch past a year.

A small agency is most exposed. It may have one person who understands the cloud bill, one who manages a legacy database and one who runs the contract for the help desk. Each is a single point of failure. A large department can usually move someone from another bureau; a small agency usually cannot.

The roles most at risk are the ones that control the other two categories. Cloud cost management needs analysts who understand both billing and architecture. Holding contractors accountable needs technical staff who can judge their work. Losing those people saves salary and raises run costs and project risk later. The bills and invoices those people used to check then go unchecked.

A more durable approach protects a small core of government staff in architecture, security, cost management and contract oversight, and uses contractors for delivery capacity. The government keeps the people who decide and check; contractors supply the people who build and run.

The strongest counter-argument is that contractors are more flexible and often more skilled in specific technologies, so agencies should buy capacity rather than build it. For delivery work, that is often right. The limit is set by law and policy. The Federal Acquisition Regulation reserves inherently governmental functions to federal employees, and OMB’s Policy Letter 11-01 asks agencies to keep enough in-house capacity to control critical functions. Deciding what to buy, accepting deliverables and judging whether a contractor is performing fall on the government side of that line.

Bringing the three together

These three behaviors interact, which is why they belong in one review. Rising run costs squeeze modernization. Deferred modernization keeps expensive legacy systems running longer, which raises run costs. Losing cost and oversight staff makes both worse. Treating them separately, in different meetings with different owners, hides the trade-offs.

Who owns which lever

Each behavior has a natural owner. Run-cost growth belongs to the CIO’s office, working with program managers who consume the services. Deferred modernization belongs to the CFO and the investment review board, which decide what gets funded. Staffing belongs to the chief human capital officer, who controls hiring priorities. When each owner optimizes its own lever, the agency as a whole loses, because the levers trade against each other.

The Federal Information Technology Acquisition Reform Act already gave CIOs a role in agency IT budgets. A joint CIO and CFO review of growing costs, deferred investments and staffing, held at least annually, makes the trade-offs explicit. What matters is a single conversation, rather than another report, in which someone can say, “If we do not fund these two cost analysts, we should expect the cloud bill to grow faster, and this modernization phase to slip.”

State and local agencies faced a parallel version. Many technology projects funded with one-time federal pandemic relief money created recurring subscription costs. Those costs would fall on general funds once the relief money had to be obligated, by the end of 2024, and spent. A county that bought a new permitting platform with relief funds, for example, would find the annual subscription landing in its regular budget once the one-time money was gone.

The relief program allowed recipients to keep spending obligated funds through the end of 2026, which softened the cliff for some. It did not change the underlying arithmetic. A subscription bought in 2022 still needed a permanent funding source by its renewal date in the years after. Counties and cities with balanced-budget requirements had to find that money by cutting elsewhere or raising revenue. Those that had budgeted for the operating tail from the start had fewer hard choices later.

Where the argument is weakest

A fair objection is that flat budgets impose discipline, and that much federal IT spending deserved scrutiny. Some agencies did use the pressure to retire duplicate tools and cancel unused licenses. The argument here does not deny that. It holds that the discipline arrives unevenly: it falls hardest on the things nobody defends at budget time, such as decommissioning, cost analysis and contract oversight. Discipline that targets those first produces higher costs later, which is the opposite of what the caps intended.

Two flat years at one bureau

To see how the three behaviors combine, suppose a federal bureau spends two years under flat caps. It runs a benefits system built decades ago, a newer case management tool bought as a cloud subscription, and a modest team of in-house technologists supported by contractors.

Year one, autumn. The fiscal year opens on a continuing resolution. The bureau is funded at last year’s rate and cannot start new work. The first phase of the benefits system replacement, planned to begin in October, waits. Meanwhile the cloud bill keeps rising, because the case management tool has added users and the security team, following the federal logging rules, is storing more data than ever. Nobody approved those increases; they simply arrived.

Year one, spring. A full-year budget finally passes, at roughly the same level as before. After paying the higher run costs, there is less left for modernization than planned. The bureau’s cost analyst, one of two, retires. The position is not backfilled. The remaining analyst now has twice the ground to cover and stops reviewing license usage before renewals.

Year two. The case management subscription renews with a price increase that nobody challenged, for a seat count that includes staff who have left. The benefits replacement starts late and runs alongside the old system, so the bureau now pays for both. Because no one scheduled the old system’s shutdown as a funded milestone, it keeps running “for now.” The contractor building the new system is capable, but with fewer government technical staff, fewer people can judge whether its work is on track.

Year two, autumn again. The new fiscal year opens on another continuing resolution. The second modernization phase, planned to follow the first, is now a new start and cannot begin. The contractor’s team for phase one finishes its work and rolls off, taking with it knowledge the agency will need to pay for again when phase two is finally funded. The legacy system, still running, needs an emergency hardware refresh that nobody budgeted for, paid for out of the same flat line. By spring the bureau is two years into a five-year plan with less than one year of progress, and its run costs are higher than when it started.

Few of these were bad decisions; most were not decisions at all. Under a flat budget, the outcome is shaped less by what leaders choose than by what they fail to choose. The same bureau, with tagged cloud costs, a renewal clause allowing seat reductions, a protected cost analyst role and a dated decommissioning plan, would have entered year two with more money for its renovation and a clearer picture of where it stood.

What it means for risk, customers and oversight

It is easy to treat this as an accounting story. It is also a service story. Legacy systems that stay in place longer are harder to secure and harder to change when policy changes. Delayed modernization means the public waits longer for faster, simpler services. And thinner oversight of contractors increases the chance that a project overruns or underdelivers, which in government tends to attract the attention of inspectors general, auditors and congressional committees. The budget mechanics described here shape, very directly, how well an agency serves the people who depend on it.

Security deserves its own note. Older systems often depend on software that vendors no longer patch, and compensating controls such as network isolation and extra monitoring add run cost. Agencies report their security posture each year under the Federal Information Security Modernization Act, and inspectors general audit those reports. A legacy system kept alive by deferral tends to show up in those audits as a recurring finding.

Oversight bodies had been watching this for a long time. GAO added the management of federal IT acquisitions and operations to its High-Risk List in 2015. Flat budgets do not create that risk, but they make it harder to retire.

One page that shows the pressure

A well-run agency can show, on one page, how fast its run costs are growing compared with its IT budget; which cloud resources belong to which programs; which subscriptions renew in the next year and on what terms; which modernization phases can proceed on flat funding; and which small teams are one departure away from failure.

Each item on that page needs a named owner and a data source. Run-cost growth comes from the cloud billing data and the license inventory, refreshed monthly. The renewal calendar comes from the contract writing system, with the notice date for each renewal flagged at least ninety days ahead. Modernization phases come from the investment plan, each tagged as fundable at flat levels or not. The staffing view comes from human resources, organized by function rather than by office. Federal agencies already report IT spending to OMB in a standard taxonomy, so much of the raw data exists; the work is joining it.

The failures are familiar in any organization under budget pressure:

• Cutting evenly. Trimming every line by the same percentage, which ignores that some lines grow on their own.
• Letting renewals auto-renew. Accepting price increases and seat counts without checking them against actual use.
• Counting migration savings early. Booking the savings from a new system before the old one has actually been switched off.
• Planning modernization on hope. Assuming budgets will rise steadily, with no defined stopping point if they do not.
• Treating headcount as a single number. Missing that a few small, specialist teams carry disproportionate weight.
• Outsourcing oversight itself. Relying on contractors to judge contractors, because the government staff who could have done it were not replaced.

Flat budgets reward early answers

The questions below are worth answering before the next budget cycle, not during it. Each one maps to one of the three behaviors, and each has an owner who can act on the answer without waiting for Congress. The fifth applies mainly to state and local agencies, and to federal programs that pass money through to them.

1. How fast are our cloud and subscription run costs growing compared with our IT budget, and what does the gap cost modernization each year, and can each program see its own share of the cloud bill?
2. Which major software contracts renew in the next twelve months, and do they cap price increases and let us reduce seats?
3. Is decommissioning the old system a funded, dated deliverable or an assumption, and does each modernization program have a usable stopping point if funding runs short?
4. Which small teams, such as cost management, architecture or contract oversight, would be crippled by one more departure, and when did the CIO and CFO last review staffing alongside run costs and deferred investments?
5. Which recurring costs did we take on with one-time money, and where will they land when that money is gone?

With hindsight

The caps held in broad terms, though final fiscal 2024 appropriations were not enacted until March 2024, after several continuing resolutions, and fiscal 2025 ran largely on a full-year continuing resolution. In practice that meant agencies spent long stretches unable to start new work and unsure of their final numbers, which is exactly the condition under which modernization slips by default.

The fiscal 2024 sequence showed how the one-percent trigger worked. Agencies were still on a continuing resolution on January 1, 2024, which tripped the provision. Congress then passed the remaining full-year bills in two packages in March 2024, before the April 30 report, so the automatic cut never took effect. The deadline shaped the timing; it did not stop the stopgaps. Agencies still spent nearly half the fiscal year without final numbers, and that delay was the part that hurt modernization schedules.

The debt limit returned when the suspension ended on January 1, 2025. Congress raised it again in July 2025 through the budget reconciliation law, by $5 trillion. That increase came without new discretionary caps, so the 2023 deal remained the most recent statutory limit on annual appropriations at the time it expired.

In 2025 the new administration canceled or cut many federal contracts, including technology consulting work, and reduced the federal workforce substantially. The Office of Personnel Management offered a deferred resignation program early that year, and GSA closed 18F, its in-house digital services team, in March 2025. The government shutdown that autumn was the longest on record. It began on October 1, 2025, lasted 43 days and ended with another continuing resolution. GSA also began negotiating governmentwide software pricing under a OneGov approach, which addressed renewal costs directly. That last step was a recognition that subscription renewals are a budget problem worth tackling at scale, not one office at a time.

OneGov came with its own timing risk. GSA’s published list of agreements shows many of the deepest discounts, including several introductory offers priced at a dollar or less, carrying expiration dates between late 2026 and 2028. Agencies that built new services on those prices will face renewals at whatever terms follow. The renewal calendar described above applies to governmentwide deals as much as to an agency’s own contracts.

Together these made budget uncertainty, even more than flat budgets, the defining feature of federal IT finance. A flat budget can be planned around. An uncertain one demands plans that can stop, pause and restart without losing what has been built, which is why modular, phase-by-phase modernization matters even more.

The attrition category proved the most acute. Agencies that lost technical staff found it harder to manage cloud costs and to oversee the contractors they still relied on. That is the interaction described above playing out in real life: fewer cost and oversight staff, higher run costs, more risk in the projects that remain.

What did not hold up was the framing of the problem as two flat years followed by a return to normal. After 2025 the constraint on federal IT combined uncertain funding, deliberate workforce reductions and governmentwide procurement changes. The three behaviors still describe how a constrained IT budget sorts itself. Attrition, though, was overtaken by decisions taken at the center, and the agencies best placed were those that already knew which of their small teams carried the most weight.

Run-cost management remains the part of the budget agencies control most directly. Leaders cannot set appropriations or hiring policy, but they can insist on tagged cloud resources, scrutinized renewals and a protected core of people who understand both the bills and the systems. Those are the levers that keep a flat or uncertain budget from consuming the future by default.