A billion dollars that doesn’t exist yet
Last month, Coefficient Giving (funded by Dustin Moskovitz and Cari Tuna) and GiveWell announced a $276 million grant for roughly 90 million insecticide-treated mosquito bed-nets in the DRC. GiveWell estimates it could prevent about 67,000 deaths, mostly kids under five. This is genuinely great news.
Where I do a bit of that 😬 is that the Chronicle of Philanthropy reports the timing is tied to an expected windfall if OpenAI and Anthropic go public. Coefficient’s CEO told Semafor he expects AI wealth to add something like $40 billion a year to American philanthropy, if those listings happen.
So now anti-malarial prophylaxis in Kinshasa are, in a small but real way, a bet on tech IPOs. If you’ve never heard of effective altruism (EA), the movement behind this kind of giving, this is a good time to start. If you have, and you have feelings, same. (Full disclosure, I attended 1 out of 3 days of EAx in Amsterdam in 2025. It wasn’t intensely culty).
What is Effective Altruism?
The origin story usually starts with philosopher Peter Singer’s 1972 thought experiment: if you walked past a child drowning in a shallow pond, you’d ruin your shoes to save them. So why does distance let us off the hook for kids dying of preventable disease on the other side of the planet?
That question got an operating system in 2007, when two former hedge fund analysts, Holden Karnofsky and Elie Hassenfeld, founded GiveWell to rank charities by evidence and cost per outcome. Giving What We Can followed in 2009, the name “effective altruism” stuck around 2011, and Open Philanthropy (now Coefficient Giving) grew out of GiveWell with Moskovitz and Tuna’s money behind it.
The core logic is a three-question filter, which Coefficient still uses: How many people does a problem affect? How neglected is it? How tractable is it? Run a cause through that, use evidence multipliers, and you get a ranking. Affectionately, it’s spreadsheet compassion.
I know I shouldn’t be so offended that its moniker implies that other altruism isn’t effective (if I follow the same logic that BLM doesn’t mean other lives don’t matter.)
The movement’s gotten bigger and stranger of late. A wing called “longtermism” argued that future people count as much as present ones, which the movement into preventing AI catastrophe. The ethical/logical inner circle is tight, maybe incestuous? Anthropic’s president is married to Karnofsky, who joined Anthropic last year. (Fun full disclosure: the AI I used to polish this draft is made by Anthropic. Noted, and heap of side-eye to myself acknowledged.)
Credit where it’s due
Bed nets through the Against Malaria Foundation are the flagship. Seasonal malaria medicine for kids through Malaria Consortium, small cash incentives that get Nigerian babies vaccinated through New Incentives, and no-strings cash transfers through GiveDirectly all have randomized trials behind them. Projects via the Lead Exposure Elimination Project are smaller but very EA in spirit: neglected, cheap-to-solve, and eminently measurable.
We might need a reminder here that measurable impact isn’t a revolutionary idea from EA. Plenty of field orgs have been counting carefully for decades (see RAG Time: How Mercy Corps Is Using AI in the Field).
So why the cringe face?
Hala Hanna of MIT Solve tells the Chronicle that EA-style funders search for impact where the light is, meaning where outcomes are easy to count. Infectious disease gets the money; noncommunicable disease (which kills more) gets far less. Philosophers like Amia Srinivasan made a cousin of this argument years ago: a cost-per-life lens can’t see politics, power, or the systems that keep people poor in the first place.
Another point: who decides? A handful of very rich, unelected people choosing priorities for whole regions is a legitimacy problem no matter how good the spreadsheet is. Sabermetrics looks simplistic if you compare it to the impact measurement of human lives in the Global South or The ClimateTM. Critics like Stanford philanthropy ethicist Rob Reich (not Robert, though I’m sure the two would get on like old shoes) have been making this point about “big philanthropy” for a while, so AI money could only compound the issue further.
Not to put another point too fine on it: Maybe the most famously EA person at this point is Sam Bankman-Fried. A lot of people have started wondering whether “effective” has come to mean “whatever smart guys in the Bay Area find sexy” and its claims of “altruism” to quite simply smack of Orwell.
My first exposure to EA
I read Bruce Holsinger’s novel Culpability (2025), maybe the first book I’ve read within a year of Oprah’s stamp of approval. I actually really do recommend it, very twisty, very breezily philosophical. In it, a self-driving minivan crashes with a scholarship jock in the driver’s seat, his dad, the narrator, beside him, and his mom, a celebrated AI ethicist, and sister, a Gen Z piece of work, in the backseat. In the story, the family gets buddy buddy with the car manufacturing tech billionaire, who is an active EA, if I remember correctly. Ultimately, in the story, responsibility is spread so thin across all these people (and machines) that it nearly disappears. That’s what I can’t shake when it comes to AI-fueled philanthropy… a bed-net program is funded by an IPO, chosen by a private board, and justified by a model, and just who exactly is accountable if something goes awry?
As mentioned, I had the opportunity to meet actual EAs at a conference in Amsterdam (basically EA’s x to TED, especially if we want our analogies to actually look more like algebra [I already told my SAT students that analogies are word algebra]!) I met some cool people. I got sick and missed the last two days. The vibes were definitely semi-cultish, but we’ll just call the rush of mid-20s Euro-techbros… somehow adulterated by a moral compass.
Paper money, real kids
Coefficient’s own money is real; Moskovitz’s Facebook and Asana wealth already exists, but the wave it’s preparing for is not. Sources say OpenAI and Anthropic filed confidentially for IPOs this summer, without much news since. We do know OpenAI is still bleeding billions, but Anthropic supposedly just posted its first operating profit. Either way, philanthropy pegged to a stock price, while not unprecedented or even frowned upon, inherits that stock’s vicissitudes. Coefficient has actually dubbed these assets “extremely volatile.”
We’re not in a generation of much long-term financial follow-through, unsurprisingly. The Chronicle of Philanthropy cites an Institute for Policy Studies analysis finding Mark Zuckerberg and Priscilla Chan have given away roughly 6% of their net worth in 15 years, despite pledging most of it. And Silicon Valley Community Foundation’s Nicole Taylor points out that most AI employees are in their twenties and thirties and not really thinking about philanthropy.
Now compare that to what Elon Musk et al just dismantled, even if it’s apples to oranges (they’re both roughly spherical and fruit, OK?). A Lancet study estimated USAID funding prevented about 91 million deaths between 2001 and 2021, on a budget that was 0.3% of federal spending. That’s about 17 cents per American per day. USAID answered to Congress, an inspector general, the GAO, and auditors who did not f*** around (just ask my Dad, a career bookkeeper for international AID subcontractors). Was money ever mismanaged? Of course. But it was a rounding error next to where the rest of our tax dollars go, especially the nonsense they’re going to now.
I’ll be upfront about my bias here: I think USAID was the cheapest, most effective diplomacy the US has ever bought, and I’ve seen how the loss has affected people I know. The agency wasn’t flawless. Its compliance demands were so heavy that big US contractors often got the money while local groups struggled to qualify, which the “Shift the Power” movement has rightly called out (we’ve written about that principle in Nothing About Us Without Us). But that’s a fixable flaw in an accountable system. Billionaire philanthropy has the opposite problem: flexible, fast, and answerable to about no one.
Peering into the void looking for an altruistic billionaire
The Chronicle piece spends a lot of time asking whether nonprofits can absorb a flood of AI money. Stripe’s Nan Ransohoff suggest thousands of new “philanthropic startups.” Meanwhile, USAID’s closure put nearly all of its 16,000 employees and an estimated 280,000 contractors and local staff out of work. The people who know how to move bed nets through customs, keep vaccines cold, and survive a federal audit are right there.
So here’s what I’d love to hear from the next newly minted AI philanthropist concede:
- Public aid was already doing this cheaply, at scale, with oversight I’m not subject to.
- My giving is filling a hole that policy choices created, not discovering an empty vessel.
- The people who ran those programs, especially national staff in the countries themselves, are the talent I should hire first.
- Counting lives saved per dollar wasn’t invented by my friends. Government auditors and M&E teams have been doing it for decades.
- A budget line passed by Congress is more stable than my stock options, so I owe grantees multi-year commitments that include salaries and overhead, not just peachy keen projects.
That last one echoes Hanna, Taylor, and Founders Pledge’s David Goldberg, who all land in roughly the same place: back existing organizations fully, use community foundations and experienced advisers, and build new institutions only when there’s a real gap. If you’re a nonprofit leader wondering what kind of staff you’ll need when the money shows up, we explored that question in Wanted: Nonprofit Staff Who Can Wrangle AI (and Still Care About People).
I’m very glad 90 million bed nets are headed to the DRC. Effective altruism earned its reputation for thinking through valuable contributions to the wider world creatively and carefully, and that matters when people and the environment hang in the balance. But it’s important to see EA’s seeming blind spots (what it can’t measure, who it doesn’t answer to, how much of its future rests on paper wealth, and ultimately, its hubris of studying whom to serve without having practitioners who do the serving) are about to scale with whatever the IPO market decides to do.
The cheapest fix isn’t a new foundation or a clever model, but rather hiring the people who already did this work under real scrutiny, and agreeing to a little scrutiny themselves.
Interested in AI that serves nonprofit missions? Read our monthly update for spotlights on funding, tools, and people building responsibly.
Leave a Reply