Forecasts on US Global Health Funding

In early 2025, the US government made significant changes to foreign aid, creating substantial uncertainty about global health funding.1 In March and April 2025, we collected forecasts on how US global health spending will change over the next few years.2 We think these forecasts may help us make more informed decisions about where to allocate our funding and identify potential gaps that may need to be filled.

Experts, superforecasters, and others from whom we solicited forecasts predict ~50% cuts in US health aid spending over the next few years, though with wide ranges (15% to 90%). Given high levels of uncertainty, we don’t put substantial weight on any specific numerical estimate. Our main takeaway is that large cuts to health aid are plausible, and it's difficult to forecast future changes with any confidence at this point.

In addition to making grants for urgent needs, we plan to continue exploring ways to respond to changes to US foreign assistance in the near and longer term and collecting forecasts as one of many pieces of information to help inform our work.

Published: May 2025

What we did

We collected forecasts from:

  • 11 external experts (former government officials, members of advocacy groups, global health and development experts, and other funders)3
  • 25 professional forecasters (Good Judgment Inc.’s “superforecasters”)4
  • GiveWell research staff who have been speaking with government, nonprofit, and funder contacts
  • individuals contributing via Metaculus (a crowd-sourced forecasting platform)5

We also incorporated information from:

  • large language model analysis of conversations with partners and internal discussion6
  • recent analyses by the Center for Global Development on FY 2025 spending and cuts7

Forecasts were collected in March 2025 (mostly between March 7-21) and reflect information available at that time.8

We collected forecasts on changes to US health aid (overall and for specific categories like HIV, malaria, tuberculosis, family planning, etc.) in FY 2025-2027 relative to FY 2024, as well as projections for Gavi and Global Fund funding in their next cycles.9

We asked for forecasts on the amount that will be spent, rather than on the amount appropriated. These values can differ due to impoundment, program terminations, administrative delays, or other executive actions preventing full disbursement of the appropriated amount. While these values have typically moved in tandem historically, that may be less likely going forward because several of the mechanisms for reducing the amount spent are under the control of the executive branch.

Findings

The forecasts show significant uncertainty about US global health funding, with most forecasts predicting substantial cuts but varying in severity.

On average, the external experts we surveyed predicted that annual US health aid spending would be ~50% lower in FY 2025-2027 compared to FY 2024, with a range across experts (range of ~15%-80% cuts).10 Experts predicted Global Fund replenishment for the 2026-2028 replenishment cycle (from all funders) would be ~40% lower than the requested amount (range of ~15%-70% cuts),11 and Gavi replenishment for the 2026-2030 replenishment cycle (from all funders) would be ~30% lower than the requested amount (range of ~15%-50% cuts).12

Superforecasters estimated ~50% cuts in health aid overall in FY 2026 relative to FY 2024,13 with Global Fund replenishment ~15% lower than requested14 and Gavi replenishment ~10% lower than requested.15

US funding for health overall and in top areas

This table shows forecasts on FY 2025-2027 spending as a percentage of FY 2024 spending (i.e., 90% means a 10% cut) for US health aid overall and for HIV and malaria specifically.

For US health aid overall, key areas of spending include infectious disease control (HIV/AIDS, malaria, tuberculosis), maternal and child health, nutrition, immunization, family planning, and global health security. This also includes annual contributions to Global Fund and Gavi. Kaiser Family Foundation estimates this was ~$12 billion in FY 2024.16

We focus on these areas because they constitute a large share of US health aid funding and were the areas for which we solicited the most forecasts. Forecasts for other areas are included in this spreadsheet.

FY 2025 spending as a % of FY 2024 spending FY 2026 spending as a % of FY 2024 spending/th> FY 2027 spending as a % of FY 2024 spending
Global health (total)
External experts17 55%(range across 9 experts: 20%-83%) 52%(range across 9 experts: 20%-85%) 58%(range across 6 experts: 40%-85%)
Superforecasters18 - 52%(rough 80% confidence interval: 18%-73%)19 -
GiveWell staff20 45%(range across 7 staff: 30%-65%) 36%(range across 7 staff: 10%-80%) 35%(range across 7 staff: 10%-80%)
Metaculus21 78%(4 forecasters, 25th-75th percentile range: 64%-92%) 71%(4 forecasters, 25th-75th percentile range: 63%-91%) 68%(4 forecasters, 25th-75th percentile range: 57%-89%)
CGD analysis22 66% - -
LLM-assisted forecasts23 75% 68% 62%
HIV24
External experts25 68%(range across 8 experts: 50%-90%) 64%(range across 8 experts: 50%-85%) 64%(range across 5 experts: 50%-85%)
Superforecasters26 - 56%(rough 80% confidence interval: 12%-77%)27 -
GiveWell staff28 51%(range across 5 staff: 33%-75%) 49%(range across 5 staff: 15%-75%) 49%(range across 5 staff: 15%-75%)
Metaculus29 74%(5 forecasters, 25th-75th percentile range: 65%-83%) 69%(5 forecasters, 25th-75th percentile range: 59%-81%) 59%(5 forecasters, 25th-75th percentile range: 47%-74%)
CGD analyses30 78% - -
LLM-assisted forecasts31 70% 63% 58%
Malaria32
External experts33 63%(range across 9 experts: 40%-88%) 61%(range across 9 experts: 40%-85%) 63%(range across 6 experts: 40%-85%)
Superforecasters34 - 45%(rough 80% confidence interval: 9%-72%)35 -
GiveWell staff36 54%(range across 7 staff: 20%-80%) 38%(range across 7 staff: 5%-80%) 38%(range across 7 staff: 5%-80%)
Metaculus37 71%(9 forecasters, 25th-75th percentile range: 53%-100%) 78%(9 forecasters, 25th-75th percentile range: 70%-93%) 85%(9 forecasters, 25th-75th percentile range: 69%-108%)
CGD analyses38 66%
LLM-assisted forecasts39 80% 75% 70%




Global Fund and Gavi

We also asked how Global Fund and Gavi replenishment will compare to requested funding. The Global Fund is requesting $18 billion for its eighth replenishment (2026-2028). Gavi’s stated need is $11.9 billion, and it is requesting $9 billion in new pledges for 2026-2030.40 You can see the full results of these forecasts in this spreadsheet.41

Global Fund replenishment (2026-2028) as a % of requested funding
External experts 59% (range across 5 experts: 33%-85%)
Superforecasters42 83% (80% confidence interval: 63%-105%)
GiveWell staff 66% (range across 7 staff: 43%-80%)
Metaculus43 79% (5 forecasters, 25th-75th percentile range: 67%-94%)
LLM-assisted forecasts44 82%
Gavi replenishment (2026-2030) as a % of requested funding
External experts 71% (range across 4 experts: 50%-85%)
Superforecasters45 92% (80% confidence interval: 60%-128%)
GiveWell staff 71% (range across 4 staff: 60%-80%)
Metaculus46 96% (8 forecasters, 25th-75th percentile range: 90%-108%)
LLM-assisted forecasts47 78%

Limitations

These forecasts have several limitations:

  • These forecasts represent input from experts, superforecasters, and others but are not a comprehensive or statistically representative sample.
  • The forecasts reflect information available in March 2025 that may quickly become outdated as new developments occur.
  • Given unprecedented changes to US foreign aid policy, historical patterns may be less reliable predictors than usual, which makes forecasts even more unreliable.
  • We asked about actual spending rather than congressional appropriations. While we think this better reflects resources reaching recipients, this adds additional uncertainty.

Takeaways and next steps

We view these forecasts as indicating a high possibility of substantial cuts to US health aid but with a high level of uncertainty. We plan to continue to investigate urgent funding gaps and explore new funding areas that may emerge as a result of cuts. We describe our strategy and highlight some of the grants we’ve made in response on this page.

To keep up to date on changes to US foreign aid, we plan to continue to speak with partners in government, nonprofit organizations, and other funders. We plan to conduct another round of forecasts in one to two months to incorporate new information. We'll also make use of Good Judgment Inc.’s public dashboard tracking global health aid forecasts for FY 2026 vs. FY 2024 and may also request additional forecasts for other areas.

  • 1

    For more context on the US foreign aid cuts, see KFF’s U.S. Foreign Aid Freeze & Dissolution of USAID: Timeline of Events. We also summarize these changes on this page.

  • 2

    For more on how GiveWell is responding, see this page.

  • 3

    This was a non-representative sample of individuals drawn from our networks.

  • 4

    Good Judgment Inc. (GJI) provides forecasting services using their network of "superforecasters", individuals with demonstrated track records of making accurate predictions about future events. GiveWell and Open Philanthropy paid GJI $144,000 for forecasts on six questions. For the forecasts, see GJI’s page here.

  • 5

    For the Metaculus questions and forecasts, see their page here:

    • Questions on specific funding levels over the next three years (e.g., “Compared to FY 2024, what will the US government's funding for international HIV/AIDS prevention and treatment be in the following years?”) typically had 3-8 forecasters per question.
    • Questions on whether specific programs will exist (e.g., “Will the PEPFAR program cease to exist before January 1, 2026?”) typically had 19-36 forecasters per question.

  • 6

    These forecasts are experimental and were included to test this approach. We roughly followed guidelines from Schoenegger et al. 2024 on LLM-assisted predictions, but consider these results more speculative than other forecasting methods presented here.

  • 7

    We included two analyses from the Center for Global Development:

  • 8

    Metaculus forecasts are the exception, as they were submitted between March 4th, when the questions opened, and April 3rd and 4th, when we pulled them for this page.

  • 9

    A copy of the survey we sent experts is here. We tailored the survey to each expert, asking for their forecasts on overall funding levels and on the areas they had expertise in. Some experts opted not to provide forecasts for all questions.

  • 10

    See these forecasts.

  • 11

    See these forecasts.

  • 12

    See these forecasts.

  • 13

    See this forecast.

  • 14

    See this forecast.

  • 15

    See this forecast.

  • 16

    Information and descriptions for these questions are drawn from KFF’s 10 Things to Know About U.S. Funding for Global Health, February 4th, 2025.

  • 17

    Expert forecasts are in this spreadsheet. Responses have been anonymized. Forecasts were not requested from every expert for all funding areas and years; gray cells indicate where no forecast was requested or an expert chose not to respond to a forecast question.

  • 18

    To see the full reports from Good Judgment Inc.’s superforecasters, see their page on “Superforecasting the US Government’s International Health Funding.”

  • 19

    The rough 80% confidence interval is the 10th percentile (22.3%) and 90th percentile (91.4%) of forecasted appropriated funding, multiplied by the median forecast that 80% will be spent. This understates 80% confidence interval, since there’s uncertainty on the median forecast of amount spent.

  • 20

    GiveWell staff forecasts are in this spreadsheet. Responses have been anonymized. Gray cells indicate where no forecast was requested or staff decided not to provide forecasts.

  • 21

    For the Metaculus forecasts on total US global health funding, see their page here. The numbers above were pulled on April 3, 2025 and additional forecasts may be made after that date.

  • 22

    For the forecast of total US global health funding in FY 2025, we use “Table 1. USAID award cancellations by sector,” “Total” row: 34% cut by value. The USAID Cuts: Little Sign of Mercy for “Life-Saving” Health Programs (March 14, 2025).

  • 23

    LLM forecasts are in this spreadsheet. We produced these forecasts using Claude, uploading notes from conversations and internal messages. Prompts and output are here, with all sensitive information removed. Claude does not use our conversations for training.

  • 24

    This includes funding through PEPFAR (the President's Emergency Plan for AIDS Relief) but not HIV funding through the Global Fund. In future years, if PEPFAR is discontinued, this would include funding for similar HIV programs (testing, prevention, treatment) that were previously part of PEPFAR. KFF estimates the US government’s global funding for HIV was ~$5 billion in FY 2024: “Consistent with this trend, in FY 2024, the latest year of data available, most funding was provided to HIV efforts ($5.4 billion or 44%).” KFF’s 10 Things to Know About U.S. Funding for Global Health (February 4, 2025).

  • 25

    Expert forecasts are in this spreadsheet. Responses have been anonymized. Forecasts were not requested from every expert for all funding areas and years; gray cells indicate where no forecast was requested or an expert chose not to respond to a forecast question.

  • 26

    To see the full reports from Good Judgment Inc.’s superforecasters, see their “Superforecasting the US Government’s International Health Funding” page.

  • 27

    The rough 80% confidence interval is the 10th percentile (15.0%) and 90th percentile (96.7%) of forecasted appropriated funding, multiplied by the median forecast that 80% will be spent. This understates 80% confidence interval, since there’s uncertainty on the median forecast of amount spent.

  • 28

    GiveWell staff forecasts are in this spreadsheet. Responses have been anonymized. Gray cells indicate where no forecast was requested or staff decided not to provide forecasts.

  • 29

    For the Metaculus forecasts on US international HIV funding, see this page. The numbers above were pulled on April 3, 2025 and additional forecasts may be made after that date.

  • 30

    For the forecast of US international HIV funding in FY 2025, we average the values in these two reports:

  • 31

    LLM forecasts are in this spreadsheet. We produced these forecasts using Claude, uploading notes from conversations and internal messages. Prompts and output are here, with all sensitive information removed. Claude does not use our conversations for training.

  • 32

    This includes funding through PMI (the President’s Malaria Initiative) but not malaria funding through the Global Fund. In future years, if PMI is discontinued, this would include funding for similar malaria programs (diagnosis, treatment, prevention) that were previously part of PMI. KFF estimates the US government’s global funding for malaria was ~$1 billion in FY 2024. KFF, U.S. Global Health Budget Figures, Figure 6a, U.S. Global Health Funding (in millions), By Program Area, FY 2024.

  • 33

    Expert forecasts are in this spreadsheet. Responses have been anonymized. Forecasts were not requested from every expert for all funding areas and years; gray cells indicate where no forecast was requested or an expert chose not to respond to a forecast question.

  • 34

    To see the full reports from Good Judgment Inc.’s superforecasters, see their page on “Superforecasting the US Government’s International Health Funding.”

  • 35

    The rough 80% confidence interval is the 10% percentile (11.2%) and 90% percentile (90.0%) of forecasted appropriated funding, multiplied by the median forecast that 80% will be spent. This understates 80% confidence interval, since there’s uncertainty on the median forecast of amount spent.

  • 36

    GiveWell staff forecasts are in this spreadsheet. Responses have been anonymized. Gray cells indicate where no forecast was requested or staff decided not to provide forecasts.

  • 37

    For the Metaculus forecasts on US international malaria funding, see their page here. The numbers above were pulled on April 3, 2025 and additional forecasts may be made after that date.

  • 38

    For the forecast of US international malaria funding in FY 2025, we average the values in these two reports:

  • 39

    LLM forecasts are in this spreadsheet. We produced these forecasts using Claude, uploading notes from conversations and internal messages. Prompts and output are here, with sensitive information removed. Claude does not use our conversations for training.

  • 40

  • 41

    Gray cells indicate where no forecast was requested or respondents decided not to provide forecasts.

  • 42

    The median of the superforecasters’ estimates for funding contributed to the Global Fund was $15 billion: $15b / $18b = 83%.
    The 80% confidence interval was calculated by dividing their 10th percentile forecast ($11.3 billion) and their 90th percentile forecast ($18.9 billion) by the total requested for the cycle ($18 billion).
    To see the full reports from Good Judgment Inc.’s superforecasters, see their “Superforecasting the US Government’s International Health Funding” page.

  • 43

    The median of the Metaculus forecasts for funding contributed to the Global Fund was $14.28 billion: $14.28b / $18b = 79%.
    The 25th-75th percentile range was calculated by dividing their 25th percentile forecast ($12.13 billion) and their 75th percentile forecast ($16.91 billion) by the total requested for the cycle ($18 billion).
    For the forecasts, the Metaculus page here. These numbers were pulled on April 4, 2025 and additional forecasts may have been made since then.

  • 44

    LLM forecasts are in this spreadsheet. We produced these forecasts using Claude, uploading notes from conversations and internal messages. Prompts and output are here, with sensitive information removed. Claude does not use our conversations for training.

  • 45

    The median of the superforecasters’ estimates for funding contributed to Gavi was $11 billion. To calculate the forecast of the percentage of the funding fulfilled, we divided that by the total Gavi target ($11.9 billion): $11b / $11.9b = 92%.
    The 80% confidence interval was calculated by dividing their 10th percentile forecast ($7.1 billion) and their 90th percentile forecast ($15.2 billion) by the total requested for the cycle ($11.9 billion).
    To see the full reports from Good Judgment Inc.’s superforecasters, see their “Superforecasting the US Government’s International Health Funding” page.

  • 46

    The median of the Metaculus forecasts for funding contributed to Gavi was $11.4 billion: $11.4b / $11.9b = 96%.
    The 25th-75th percentile range was calculated by dividing their 25th percentile forecast ($10.7 billion) and their 75th percentile forecast ($12.85 billion) by the total requested for the cycle ($11.9 billion).
    For the forecasts, the Metaculus page here. These numbers were pulled on April 4, 2025 and additional forecasts may have been made since then.

  • 47

    LLM forecasts are in this spreadsheet. We produced these forecasts using Claude, uploading notes from conversations and internal messages. Prompts and output are here, with sensitive information removed. Claude does not use our conversations for training.