The GAO's 18th annual review of NASA's major projects finds a portfolio of 36 projects, worth at least 70 billion dollars in life-cycle costs, that is broadly under control: only 4 of 18 projects in development reported new cost or schedule overruns this year. That aggregate stability, however, conceals two concentrated sources of risk that dominate the individual case findings below - a wholesale restructuring of the Artemis lunar-return architecture triggered by a December 2025 executive order, which is forcing simultaneous re-design across half a dozen interdependent projects; and a 22 percent cut to NASA's civil-servant workforce in 2025, whose effects on specific projects are only now surfacing. A handful of individual project assessments - Commercial Crew's Starliner anomaly, Dragonfly's power-source shortfall, Mars Sample Return's cancellation - further illustrate internal-control weaknesses that cut across categories. $70 billion is the estimated total life-cycle cost of NASA's portfolio of 36 major projects (each over $250 million). 18 projects are in development, 17 in formulation, and 1 (Commercial Crew Program) has no set baseline. NASA's civil-servant workforce fell nearly 22 percent in 2025, affecting 25 of the 36 projects. A NASA procedural requirement defines the project life cycle, key decision points, and cost-category thresholds. A federal statute triggers a mandatory replan and congressional notification once development cost growth reaches 15 percent or a milestone slips 6 months. A congressional appropriations mandate established GAO's standing annual major-projects reporting requirement. Two 2025 executive orders separately drove this year's workforce reductions and the Artemis mission restructuring. Portfolio cumulative cost growth rose to about $4.7 billion, up $311 million from last year. Two Artemis projects - Orion and SLS Block 1B - drove nearly all of the year's overruns. Orion's 2026 exit will remove about $3.5 billion, 73 percent of the portfolio's cumulative cost overruns. Portfolio cumulative schedule delay grew 11 months, to 14 years. Only 2 of 18 projects in development slipped this year, by 1 month each. A replan is triggered at 15 percent cost growth or a 6-month slip - the threshold Orion and SLS Block 1B crossed. Testing worked as intended where it ran its course: Roman caught two design flaws pre-orbit, and GRACE-C stayed on schedule. The SLS program at Marshall lost almost 20 percent of its civilian workforce in 2025. GAO's recommendation to stop approving risk-compounding decisions remains only partly addressed. NASA's own root-cause report on Boeing's 2024 crewed Starliner flight test found 'a complex interplay of hardware failures, qualification gaps, leadership missteps, and cultural breakdowns that collectively posed risks to crew safety' (p.61), rooted in NASA's own incomplete systems knowledge of the spacecraft and Boeing's inadequate systems engineering and subcontractor oversight. This is the richest single case in the report: a Testing gap (insufficient thruster testing) compounded by a Responsibility failure (subcontractor oversight) and a Communication breakdown between agency and contractor, with direct exposure to Community trust in human-spaceflight safety. A December 2025 executive order and NASA's February-March 2026 decisions paused Gateway, canceled SLS Block 1B, and halted Mobile Launcher 2, forcing Orion, HLS and EGS to redesign interfaces simultaneously. GAO states plainly: 'managing the changes described in the table presents a significant acquisition management challenge to NASA... changes to one project may have cascading effects on another' (p.24). This is a Regulatory system shock (a government strategy set by executive order) that exposes a Coordination and Design weakness across a tightly interdependent project family. The 2025 deferred-resignation program cost the DAVINCI project at Goddard Space Flight Center its program-management and engineering leads with no backfill: 'DAVINCI lost critical team members with leadership roles in program management and engineering to the deferred resignation program and was unable to backfill the positions in fiscal year 2025' (p.32), forcing it to rescope planned risk-reduction work. A direct hit to Human capital (skills and knowledge retention), triggered by a government strategy (the workforce-reduction executive order) and answered, in part, by a new Guidance directive to rebuild in-house technical roles. Thermal vacuum testing on the Nancy Grace Roman Space Telescope caught two design flaws before shipment: 'project officials said, during the SCIPA thermal vacuum testing, they identified cryogenic frost (microscopic ice) buildup on the WFI detectors' (p.80), and unexpectedly high sun-shield-harness temperatures. Both were mitigated pre-launch. A positive counter-example to the Starliner case: the same Testing function working exactly as intended, followed by Process optimization (new decontamination-cycle procedures) in response. The Department of Energy, which supplies Dragonfly's radioisotope power source, told NASA after the design was set that 'the MMRTG will produce less thermal power than the project requires under the current concept of operations' (p.66), forcing a rework of the lander's surface-recharge planning. An Asset risk NASA does not itself control, surfacing a Domain knowledge gap at the boundary with the body that supplies the component rather than within NASA's own engineering. A government shutdown delayed launch-vehicle selection for ROSA, which in turn delayed the start of the nuclear launch-authorization process for the rover's radioisotope heaters: 'ROSA's top risk is the delay in receiving nuclear launch authorization... The selection was delayed by 3 months because of the government shutdown' (p.81). A macro-level Regulatory system event (an appropriations lapse) propagating into a specific technical Coordination risk with the Department of Energy's licensing process. Mars Sample Return was cancelled as unaffordable and staff drawn down before NASA finished comparing cheaper NASA-led and commercial lander architectures that could have cut cost and schedule substantially: 'NASA began closing out the program before it completed its assessment of these options' (p.73). A Handover and Results review weakness - closeout initiated ahead of the analysis meant to inform it - compounding a Human capital loss (engineering expertise dispersed mid-decision) and a Finance-driven (affordability) cancellation. Orion's overruns account for 52 percent of the portfolio's annual cost overruns (p.16), and its departure from the portfolio after its April 2026 launch will remove 'approximately 3.5 billion dollars in development cost overruns - or 73 percent of the current portfolio's total development cost overruns' (p.19). A single project this dominant distorts aggregate Reporting - portfolio-level figures can mask, rather than reflect, the other 17 projects - and underlines why Finance baselines need to be read project-by-project, not only in the aggregate. Climate and environmental-monitoring missions GRACE-C and SBG-TIR face both cancellation proposals and, for GRACE-C, an unusual cost risk: 'the project is also tracking a risk to its cost reserves due to the exchange rate between the U.S. dollar and the euro, which is increasing the cost of work completed by European contractors' (p.69). These are jointly delivered missions - DLR on GRACE-C, ASI on SBG-TIR - and the currency exposure arises precisely because a substantial share of the work sits on the partner side of the arrangement. A case that ties directly to Nature (the missions monitor water, ice mass and land-surface change), with Finance exposure running through the partnership rather than through domestic budget alone.