The nuclear cleanup bill is $539 billion. In 11 of 21 big projects, DOE picked the answer before asking the question.

The finding, in one paragraph: the federal government's second-largest acknowledged bill after entitlements is the cleanup of the nuclear weapons complex, and the office holding the pen fails at both ends of the ledger.▸ At the back end, the Department of Energy could not give its independent auditor enough evidence to support the liability estimate at four of its biggest sites, drawing a qualified opinion and a material weakness on the FY2025 financial statements. At the front end, the Government Accountability Office found that when the Office of Environmental Management launches the giant projects that determine what the bill will be, it writes the answer into the question — naming a specific facility or technology in the "mission need" document that is, by DOE's own rules, supposed to be solution-neutral — and the options never studied include the cheapest ones on the table.

Five documents, read directly.▸ The new one is GAO's Nuclear Waste Cleanup: Changes Needed to Ensure DOE Is Not Prematurely Excluding Less Expensive Options for Large Projects (GAO-26-108193, released July 2, 2026), built with a twelve-member expert panel. Its siblings: GAO-26-107820 (March 2026), on the accuracy of the cost and schedule numbers EM reports, and GAO-26-107957 (May 2026), on the office's aging infrastructure. The ledger itself is the DOE Agency Financial Report for FY2025 (DOE/CF-0221) — Note 14's liability rollforward and KPMG's audit opinion — read against DOE's own project-management rulebook, Order 413.3B, as quoted in the GAO report.

DOE's audited books put the environmental cleanup and disposal liability at $538.6 billion — three-quarters of the department's entire $714 billion liability balance, and almost none of it covered by existing budgetary resources.▸ The rollforward shows the shape of the problem: even in a year of full-scale effort, the payment barely outruns the growth.

Cleanup work applied $7.4 billion against the bill; estimate revisions put $2.9 billion back.▸ Net progress for the year: $5.9 billion, about 1.1 percent — and that understates the drift, because GAO's March report found EM's operations cost estimates grew by roughly $75 billion since 2022, with its eight most expensive capital projects up another $2 billion. Meanwhile the machinery doing the work is itself a liability: EM operates about 4,300 facilities, many "50–70 years old and well beyond their designed life," with a repair backlog over $1.5 billion and maintenance spending up 80 percent since FY2020.

The July 2 report is about how bills this size get written.▸ DOE Order 413.3B requires that a large project begin with a statement of mission need — the capability gap, not the answer — precisely so the alternatives analysis that follows is real. GAO read all 21 approved mission-need statements for EM's large capital projects. Eleven named the solution.

The pattern in three sites.▸ At Idaho, the analysis of alternatives for calcine waste priced vitrification at "at least $3 billion" against "about $500 million" for direct disposal — but the optimal solution had been "deemed infeasible early on," and the technology actually selected was "later deemed to be technologically immature." At the same site, the Integrated Waste Treatment Unit's steam-reforming choice carried known technical uncertainties into construction and emerged with a 12-year delay and almost $1 billion in overruns before finally operating in 2023. At Hanford, the mission need stated flatly that "a Low-Activity Waste Pretreatment System must be built"; a peer review of that project concluded it would blow through its cost and schedule estimates "by at least 50 percent." And the biggest number on the table — grouting low-activity tank waste instead of vitrifying it, which GAO's March report says could save up to $210 billion — entered the plan only when a renegotiated 2025 agreement finally permitted grouting "portions" of the waste, the cheaper technology Savannah River had been using all along.

Set DOE's rulebook against DOE's practice, and DOE's books against DOE's auditor.▸ Order 413.3B and GAO's own best practices say the same thing — in GAO's words, "allowing mission need to be defined in solution-specific terms creates a potential bias which could prevent the inclusion of viable alternatives and invalidate the analysis." That is the department's standard. Eleven of its 21 large-project charters violate it. The report also names the pressures GAO wants walled off from future reviews: "legal and regulatory constraints, influence from contractors, or existing projects."

On the books: KPMG found DOE's financial information reliable "with the exception of" the environmental liability components at the Portsmouth Paducah Project Office, Savannah River, the Waste Isolation Pilot Plant, and Hanford — the department "was unable to provide sufficient and appropriate audit evidence in FY 2025 to fully support" them — and flagged one material weakness, in EM's financial reporting of the liability estimates.▸ The same office that chooses the expensive option without a full analysis also cannot fully evidence what the total will be. March's report closes the loop on the middle: some of the cost and schedule numbers EM reported were "inaccurate or included increases that had already been accounted for," and EM struggled to explain its own overruns because key documents weren't in its project database.

DOE concurred with both July recommendations — solution-neutral mission needs, and independent outside experts at the mission-need stage — with completion dates promised, and has already implemented the first of March's two fixes (peer-review reports now upload to the central project system).▸ The grouting of Hanford low-activity waste proceeds under the 2025 agreement, in portions. The material weakness on the liability estimate is slated for corrective action in FY2026. The 33 active capital projects, estimated at $44 billion as of March 2026, continue under charters written before any of it.

All figures are from the five documents cited in-line, each read directly. The $210 billion grouting figure is GAO's upper bound ("up to"); savings estimates are potential, not booked.