The Intermediaries

How foundations and donor-advised funds shape the flow of charitable capital

The capital flow

Most charitable dollars pass through an intermediary first

Charitable dollars rarely travel directly from donor to operating organization. Most pass through at least one intermediary vehicle — a foundation, a donor-advised fund, a fiscal sponsor, or a community fund. Each intermediary adds friction, delay, and its own institutional incentives.

Donor
Individual or
corporation
Intermediary
Foundation, DAF,
fiscal sponsor
Operating Org
Charity, research lab,
advocacy group
Beneficiary
People, communities,
ecosystems

The tax deduction is claimed at the first arrow — when the donor transfers money to the intermediary. But the charitable impact doesn't happen until the last arrow — when the operating organization deploys the money. The growing gap between these two moments is the structural problem this analysis examines.

5%
Required annual payout for private foundations
0%
Required annual payout for donor-advised funds
$2.1T+
Combined intermediary assets — ~$1.8T foundations (2025) + ~$330B DAFs (2024)
Foundations

The 5% floor anchors payout — but it's a mix, not a monolith

The Tax Reform Act of 1969 established a minimum 5% payout requirement for private foundations. The idea was simple: foundations shouldn't be allowed to sit on tax-advantaged assets indefinitely without deploying them for charitable purposes.

In practice the sector is a genuine mix. Aggregate grantmaking runs ~$117B a year on ~$1.8T of assets — about 6.5%. Behind that average: a few foundations spend far above the floor (Bloomberg ~12%, Gates ~11% counting its direct programs), a large cluster sits within a point of 5%, and a tail — including some of the very largest — sits below it. What's nearly universal isn't the rate; it's that investment returns (8-12% in good years) outrun payout, so endowments grow even after all grantmaking.

Foundation payout rate, year by year
Qualifying distributions as % of fair market value of assets, with the 5% required minimum marked
CG calculation from IRS SOI data. Qualifying distributions as % of fair market value of assets.

Foundation by foundation, the spread is wide — from Novo Nordisk at ~1% to Bloomberg above 12%. And grants aren't the whole picture: operating-heavy foundations like Rockefeller run large in-house programs that count toward the IRS test but never appear as grants. The snapshot below stacks both layers — grants (solid) plus operations, direct programs and PRIs (shaded):

Payout rate by foundation, 2024
Grants paid (solid) + operations, direct programs & PRIs (shaded), ÷ assets — the shaded layer also counts toward the IRS test. Dashed line = the ~5% legal floor, drawn only across the US private foundations it binds; gray bars (right) aren't subject to it
990-PF figures ÷ prior-year assets. Solid = grants paid; shaded layer = disbursements for charitable purposes beyond grants (operations, direct charitable programs, PRIs — FY2024 990-PFs via ProPublica). The layer ranges from ~0.1 points at lean grantmakers (Lilly, Bloomberg) to ~3 points at operating-heavy ones (Rockefeller, Gates). Open Society shows ~0 because its network's staff costs sit in OSF entities outside these two 990s. Gray bars = the 5% rule doesn't apply (Novo and Wellcome are foreign — Wellcome's figure already includes direct spend; HHMI is an operating foundation, shown as program spend). Even on total spend, sub-5% bars aren't necessarily violations — the callout below explains the mechanics.

How each foundation got there, year by year:

Payout rates, foundation by foundation
Grants paid ÷ total assets for 14 large foundations, with the 5% required minimum marked. Dashed lines = the US 5% rule doesn't apply
CG calculation from IRS 990-PF filings (grants paid ÷ total assets). Dashed lines = not subject to the US 5% minimum: Novo Nordisk and Wellcome are foreign foundations; HHMI is an operating foundation (program spend, not grants).
Can a foundation legally sit below 5%? Often, yes — these lines measure grants ÷ assets, which is stricter than the IRS test. Qualifying distributions also count charitable admin costs, direct charitable programs, and program-related investments — worth anywhere from ~0.1 points at lean grantmakers (Lilly, Bloomberg) to ~3 points at operating-heavy ones (Rockefeller's FY2024 990-PF shows ~$380M of charitable disbursements but only ~$190M of grants; on total spend it's at ~6%, comfortably above the floor). The denominator is also the year's average investment assets, not year-end totals inflated by a stock run-up; excess payout carries forward five years; and each year's required amount can legally be paid as late as the end of the following year. Foreign foundations (Novo, Wellcome) face no US minimum at all, and operating foundations (HHMI) satisfy a different test. Lilly Endowment is the case where even the lagged test broke: its Eli-Lilly-stock surge outran its grantmaking, and its 2024 audited statements disclose a ~$3.6B distribution shortfall it must now make up.
Every additional percentage point of payout on the sector's ~$1.8 trillion is roughly $18 billion a year of additional grants — two points would free ~$36B/yr, more than half of what every DAF in America combined grants in a year ($65B in 2024). Model what a higher payout does to a foundation's lifespan →

The argument for low payout rates is perpetuity: foundations are designed to exist forever, so they need to preserve capital in real terms. But perpetuity is a choice, not a necessity. Some of the most impactful foundations in history — the Rosenwald Fund, Atlantic Philanthropies, the Aaron Diamond Foundation — spent down their endowments entirely and arguably achieved more per dollar by concentrating their giving in time. And the choice is being made again at the very top of the sector: the Gates Foundation will spend ~$200B and close by 2045, Good Ventures has long planned to spend down in its founders' lifetimes, and MacKenzie Scott is giving her fortune away as fast as she can move it. The Overview's “new playbooks” are, in structural terms, a revolt against the 5% drip.

The DAF question

A DAF has no legal requirement to ever distribute its funds

Donor-advised funds are the fastest-growing charitable vehicle in the United States, and the most controversial. They combine the immediate tax benefits of a charitable gift with the ongoing control of a private foundation — but without the foundation's 5% payout requirement, excise taxes, or public disclosure obligations.

DAF aggregate payout rate, 2013–2023
Grants as a share of beginning-of-year assets, all US donor-advised funds
CG calculation from NPT DAF Report 2024. Payout rate = grants / beginning-of-year assets.

The aggregate payout rate for DAFs has hovered around 20-25%, which looks healthy. But aggregate figures mask enormous variation. Some donors use DAFs as short-term holding accounts, granting everything within a year or two. Others use them as de facto endowments, letting assets accumulate for decades.

Unlike foundations, DAFs have no legal requirement to ever distribute their funds. A donor can claim a full tax deduction upon contribution, then leave the money invested indefinitely. The sponsoring organization (Fidelity Charitable, Schwab Charitable, etc.) earns management fees on the assets regardless of whether they're ever granted out.

The parking lot

Roughly $100–200 billion is parked beyond what grantmaking needs

How much charitable capital is effectively "parked" in intermediaries? We can estimate this by looking at the gap between what intermediaries hold and what they distribute annually.

Foundations hold roughly $1.8 trillion (2025) and distribute ~$117 billion a year (~6.5%) — while their endowments earn returns that typically outrun that rate, so the corpus keeps growing. Every additional point of payout would move ~$18 billion more per year.

DAFs hold roughly $330 billion (2024) and granted about $65 billion that year. The aggregate payout rate looks reasonable (~25%), but the asset base keeps growing — 2024 contributions ($90B) outran grants by $25 billion. If contributions were capped at the prior year's grants — preventing net accumulation — roughly $50-80 billion in additional capital would have flowed to operating organizations over the past decade.

How much is “parked” depends on what you count. The full $2.1T stock is roughly eleven years of the sector's entire grantmaking held in reserve. Even the most conservative counterfactual — foundations paying two points more, DAF contributions capped at prior-year grants — would have moved $100-200 billion more to operating organizations over the past decade. Either way: capital that has already generated a tax deduction but hasn't yet produced charitable impact. (The Effectiveness framework's Tier 4 measures the annual flow into this parking lot — this page measures the accumulated stock.)

The counterargument is that this capital isn't idle — it's being invested, generating returns, and will eventually be distributed. But "eventually" is doing a lot of work in that sentence. Discounted to present value, a dollar granted ten years from now is worth substantially less than a dollar granted today — especially if the charitable problem it's meant to address (disease, poverty, climate change) is growing or compounding in the interim.

The incentives

Every actor's incentives align toward accumulation, not deployment

The intermediary accumulation problem isn't caused by bad actors. It's the predictable result of structural incentives:

For donors: The tax deduction is immediate upon contribution, regardless of when (or whether) the money is deployed. There's no tax cost to delay, and significant social and psychological benefits to maintaining a large philanthropic portfolio.

For foundation boards: Perpetuity is the default expectation. Spending down requires an affirmative decision to end the institution. Staff jobs depend on the foundation continuing to exist. Investment performance is measured; grantmaking effectiveness usually isn't.

For DAF sponsors: Revenue comes from management fees on assets under management. More assets = more revenue. There's no business incentive to encourage faster grantmaking, and the competitive dynamics push sponsors toward more accommodating terms, not stricter payout expectations.

None of these actors is behaving irrationally. They're each responding to incentives that happen to align toward accumulation rather than deployment. Changing the outcome requires changing the incentive structure — through regulation, disclosure requirements, or new norms.

The stress test

The AI wave routes straight through these vehicles

All of this machinery is about to be stress-tested at a new scale. The $280B+ of AI-wealth philanthropy now forming — the OpenAI Foundation, Anthropic founders' pledges, employee DAFs — will flow almost entirely through private foundations and donor-advised funds: the exact structures described on this page. The early returns look familiar: the OpenAI Foundation controls ~$130B+ of equity and has granted only tens of millions so far.

Whether the new money behaves like MacKenzie Scott — fast, unrestricted, spending down — or settles into the 5% drip is one of the biggest open questions in the Forecast. At these asset levels, the difference between those two playbooks is tens of billions of dollars a year reaching (or not reaching) operating organizations.

So the money is parked — does the rest do good?

The Effectiveness framework maps where deployed dollars land. Or model how fast a foundation could spend down instead.