$370 billion. That is the combined philanthropic asset base Nan Ransohoff calculates from three sources alone: the OpenAI Foundation’s 26% stake in OpenAI, Anthropic’s seven co-founders pledging 80% of their wealth, and Anthropic employee donor-advised funds. Writing in her Substack essay “The Third Wave of American Philanthropy,” Ransohoff argues this wealth will flow out at $37 billion to $100 billion per year, adding 6 to 17 percent to annual US charitable giving. Her arithmetic is defensible. Her conclusion deserves sharper scrutiny.
The thesis holds up better as directional sociology than as financial modeling. None of this capital is liquid yet. OpenAI and Anthropic are private companies. A valuation is not a bank balance; it is an agreement about price that exists until it does not. The comparison Ransohoff herself raises is apt: the FTX collapse evaporated roughly $40 billion in pledged effective altruism capital within months. IPO timelines for both companies have slipped before and could slip again. Even under the most optimistic scenario, converting paper wealth into deployed grant dollars requires liquidity events, transfer mechanisms, and organizational buildout that do not happen on a quarter’s notice.
That said, Ransohoff’s historical framing is the most valuable part of her argument. The first wave of American philanthropy came from Carnegie and Rockefeller’s industrial fortunes and built the institutional infrastructure of modern civic life: the research university, the public library, the perpetual foundation. The second wave, roughly 1990 to 2020, was driven by Gates and Buffett through the Giving Pledge and focused on quantifiable global health outcomes. The effective altruism movement gave that wave its measurement vocabulary.
The third wave, she argues, starts now, and its priorities differ from both predecessors. The likely cause areas are not hard to read from the funders’ worldviews. AI safety and alignment research will absorb significant early capital, because the people writing the checks believe the risk is real and the window is short. Biosecurity and pandemic preparedness are a natural second cluster; both OpenAI and Anthropic leadership have made this concern explicit. Longevity science and basic research infrastructure are plausible third-tier bets, especially for founders who believe the post-scarcity upside of AGI justifies long time horizons.
Compare this to the Gates Foundation’s portfolio and the contrast is clear. Gates, advised by GiveWell’s framework, directed capital toward problems with measurable, near-term outcomes: malaria nets, oral rehydration therapy, childhood vaccination. The AI philanthropy wave is likely to fund problems whose outcomes are not measurable on a five-year timeline. AI safety research does not produce a body count avoided. Alignment progress does not show up in a DALY metric. This creates a genuine evaluation problem that the effective altruism toolbox, built for poverty interventions, is poorly designed to solve.
Ransohoff acknowledges this tension and her recommended response, building a new philanthropic ecosystem modeled on Silicon Valley’s venture capital architecture, is analytically coherent. She calls for something resembling a YC for philanthropic startups, for independent capital allocators analogous to specialist VCs, and for pay structures that let top talent compete with private-sector compensation by distributing Anthropic or OpenAI equity as performance bonuses. The idea is not fanciful; it maps directly onto how Open Philanthropy, which now disburses over $300 million annually, was actually built.
The bottleneck Ransohoff identifies is not the money, and it is not even the organizational capacity. It is the talent pipeline. She calculates that deploying $50 billion per year at average grant sizes comparable to Coefficient Giving would require roughly 50,000 grants annually and 5,000 full-time grant staff. That is, in her phrase, an Alphabet-worth of employees, directed at problems most Alphabet employees have never thought about professionally.
The structural skepticism worth naming here is concentration risk. About half of Ransohoff’s projected philanthropic capital traces to a single entity: the OpenAI Foundation. If OpenAI’s restructuring negotiations, commercial trajectory, or regulatory environment produces a different outcome than today’s $850 billion valuation implies, the entire model compresses sharply. The Gates Foundation comparison is instructive in the other direction: Gates’ philanthropy worked partly because Microsoft’s equity was already liquid before the foundation scaled its operations significantly. This wave is being planned before the liquidity events have occurred.
The reader who should act on this analysis in the next 18 months is not a casual observer. Founders considering the next company, researchers inside biosecurity or AI safety labs seeking institutional backing, and cause-area organizations positioned in AI governance or longevity science should start building relationships with the allocators forming now, because the capital, when it does become liquid, will flow fastest to organizations with existing track records and trusted referral networks already in place.
Based on Nan Ransohoff’s essay “The Third Wave of American Philanthropy,” published on her Substack at nanransohoff.substack.com.
