Always be closing

The odds are shifting in the AI casino. We see—public opposition to data centres; business resistance to token-gouging; scientists’ doubts about AGI; and economists’ disagreements about productivity and employment effects. More subtle shifts emerge as talk of bailouts from the promoters, as leaks from lenders about how the vig is no longer enough, and even as speculation about laying off some of the action to the feds¹.

Our AI punters, having seen these muddy stretches of turf even before the gun, have accordingly mounted a PR campaign with messages to different audiences. With the confident mantra of “Message: We care” from smiling billionaires to anxious county executives, with the soothing encomia to the wisdom of markets in the FT and the Times, another message, one more of joy than of apprehension, is to recount the benefits we may all anticipate from the efficient generosity of the new philanthropists.

A bearer of joy is Nan Ransohoff, from whom we learn in “The Third Wave of American Philanthropy” (May 19, 2026), that : “AI is about to generate hundreds of billions in new philanthropic funding; and “We have a huge amount of work to do to make the most of it”. The first proclamation depends on projections —how much money will AI make and how much will its barons hoard ?—which are so easily manipulated as to be worthless. ² The following question—how to use these hypothetical windfalls ?— is the greater one, which Ransohoff fails because she, like the Abundance grifters³, has been indoctrinated to fail.

My critique of Ransohoff’s happy warrior sketch of a philanthropic AI is this: (1) wrangling bigly numbers to create a premise of urgency ⁴ is a dishonest starting point; (2) relying on the fickle generosity of the tech caste to fund basic entitlements — pension security, public education, health, disaster prevention and relief, mitigation of and adaptation to global heating, support to artists — is a bad way to produce public goods; (3) glorifying the personality of the “good billionaire” ⁵ worsens inequality by further concentrating power in the production of public goods; and (4) anointing “elite capital allocators”, as Ransohoff worshipfully calls them, to manage potential AI windfalls will require cadres of militantly ambitious factotums ⁶ working for the allocators and being accountable only to them.

What is the philanthropic potential of the AI valuations ?

Philanthropic potential depends on how much the AI companies make and on how much they hide in their pet foundations. Ransohoff’s model is the equation below.

Liquid Capital = Valuation * Endowment share

“Liquid capital” is the potential beginning endowment of the new generosity, to be allocated by “philanthropic startups”. For valuation, Ransohoff cites May, 2026 guesses of US$850 billion for the Open AI Foundation and of US$900 billion for Anthropic. That these books are unmarked to market and cannot be verified is the lesser of two defects⁷. The greater problem with gray market valuations is that they will be manipulated to entice new rounds of suckers. Bernie Madoff had nominal control of US$65 billion as he celebrated Thanksgiving in 2008. The business of Elizabeth Holmes was once valued at US$9 billion. The issuer of USDC got a federal bailout of more than US$3 billion after SVB collapsed. The enterprise value of federal prisoner (as of June 22, 2026, anyway) Sam Bankman-Fried was theoretically US$ 15 billion. We Work had a gray market number of US$47 billion, went public at about US$9 billion, and stood under US$50 million when it filed for Chapter 11. I urge all to read Hubert Horan in Naked Capitalism on the profitability of Uber and Ed Zitron on the economics of AI in general.

Ransohoff whisks some flocculent numbers about what will be donated (the “endowment shares”), as percentages of valuations, by the Open AI Foundation, by the founders of Anthropic, and by Anthropic employees. Given that “Liquid capital” cannot reliably be calculated if “Valuations” and “Endowment shares” are unknown, I won’t badger Ransohoff about endowment shares except to say that no one really knows how much will flow from the new wells. She guesses $US50 billion as the potential annual flow of liquid capital and compares it to the pittance distributed each year by Gates Foundation (US$9 billion) to intimidate readers with the size of the task that “we” face.

Though the volumes of liquid capital and the shares to endow are not bankable because they are imaginary, they serve their masters in other ways. Scale creates urgency to act — “customer doesn’t come on the lot unless he wants to buy”. Scale burnishes the icons of the new philanthropists. The personalization of generosity is a strategy — it takes “exceptional individuals” to give away such wealth and the public must be warned not to disturb their tranquility, like children being told not to play loudly when mommy is coding an app. Ransohoff describes the techno-Randian heroes required for the new philanthropy: “Traditional philanthropic orgs and people won’t cut it. These problems demand tech-caliber talent and execution”; in her words, “we need elite capital allocators”. There is nothing to respect in Ransohoff’s game, but, damn, telling Gates that his foundation sucks does merit some faint admiration.

What will the “philanthropic startups” do with their 50 billion ?

Ransohoff is less definitive on this question, a vagueness deriving from the fact that it is easier to talk about raising money than it is to talk about using it. A practiced cluelessness limits Ransohoff’s ability⁸ to suggest any projects more substantive than the vaporware one sees on LinkedIn. She urges us to “make the problems we need to solve more legible”. (One imagines that heat, poverty, violence, homelessness, illiteracy, lack of clean food and water, are legible to those personally acquainted with them). She boldly informs readers that “Defining the problems we want solved is no small task, and vision is in short supply”. (I had been like totally worried about that). Beyond such banalities, Ransohoff has no idea about good investments or how to manage them unless she has been guided by the practice of Geoffrey Firmin, formerly of Quauhnahuac, Mexico, who mused late in life that his book of secret knowledge could never appear, precisely because of its subject.

What are the risks of philanthropic start-ups ?

Despite the faults in Ransohoff’s message, it is urgent to prepare for the potential windfalls from AI by identifying operational risks and in stating how they can be managed for higher public returns. The philanthropical start-up model has no advantages in the operational cycle—not in project selection, not in project administration, and least of all, in results evaluation⁹—and, moreover, its cult of personality aspects aggravate the weaknesses of current foundation programs.

The model devalues honest analysis. Those making grant proposals will dare not disagree with the allocators or risk their funding. Those who have attached themselves closely to the allocators will see to that.

Tax evasion will grow. A virtue of Tim Schwab’s work on Gates is to remind us that foundation money is not private—it is public because it is tax-sheltered. Tax-sheltered foundation cash is a way of buying information and influence while having the preterite pay.

There will be continued theft of intellectual property (IP) as we see in the ongoing construction of the LLMs. There will be wider capture of potential new IP as the startups will surely insist on owning the products of their grants.

The philanthropic model corrupts grant-making because it is accountable only to allocators and their inner circles. Philanthropy controls money, careers, institutions, and, by those levers, guides entire fields of analysis. This corruption operates through three mechanisms:

Researchers learn which proposals attract funding. The foundations, by signaling what they will buy, narrow research methods and agendas to “what pleases the allocators and their minions” not to “what answers questions”.¹⁰

Foundations fake transparency by publishing lists of recipients and reports of results, but their decision-making is opaque: How were proposals reviewed ? This opacity is not accidental — it protects the foundation from scrutiny while allowing it to present itself as a neutral, technocratic arbiter of the public good, as we saw with the effective altruism racket.

The product of the philanthropic model is a tax-exempt and self-perpetuating patronage network. The “good billionaire” is not a benefactor but a politician without voters, spending tax-sheltered public money to widen private dominion over the public good while starving what should be countervailing powers.

What can be done with the AI valuations ?

The first thing is to tax them. A tax-funded mechanism would then allocate large block grants to purposes with successful histories, low transaction costs, and a requirement for open-access project appraisal.

I have previously suggested such large grants as examples of what the Gates Foundation might do in its final years (said to be 2025-45) and I reformulate those suggestions here. ¹¹ The Gates Foundation is projected to commit US$10 billion annually over 20 years. Getting impact from the US$200 billion demands focus. I proposed four domains: (1) public services in states (California, Washington) where the tech fortunes have been made; (2) debt relief to poor countries; (3) response to pandemics and other public health emergencies; and (4) development of next generation vaccines. After the Project 2025 assault on US scientific capacity, a fifth domain is to invest in restoring that capacity. Grants would be made over 10-year periods.

The advantages of a portfolio of a few large grants are:

A concentrated portfolio can work in proven domains that are easy to explain; people want greater pension security, they want protection against diseases, they are willing to support debt relief, they want good science, they don’t want their mothers to be crippled by polio;

A concentrated portfolio uses a few existing institutions; CALPERS, WHO, GAVI, IDA, and a few others, saving administration and grant-preparation costs and avoiding the juvenile conflicts among multiple donors to the same effort;

The stated domains have widely distributed benefits, a feature that serves to mute regional opposition within the US;

Methods of estimating benefits in those domains are well-understood; and

Funding the global health organizations has the effect of restoring some of the USA’s lost credibility without further brightening the halos now shining around the US plutocrats.

1 Instructive writers on the AI polyverse are Brad DeLong, Gary Marcus, and Ed Zitrin.

2 Credit Ransohoff for the deft “we” there. She is telling the preterite “We’re all in this together”. This is definitely a step up from Leona Helmsley.

3 I reviewed Abundance in “Midcult Redux” (The Apricot Tree of Tangiers, May 2, 2025) arguing that Klein and Thompson refused to understand that many good ideas (public transport, de-carbonization, public health, denser housing) face rich, organized, and violent opposition from reactionary elites and that even supposedly progressive rich people want a government that works mainly for them and therefore reject systemic change. Ransohoff suffers from the same strategic blindnesses as Klein and Thompson. It is regrettable that Jennifer Pahlka, who writes well on governance and service delivery and somehow emerged dry from the wreck of Abundance, not only fails to spot Ransohoff’s game, but encourages the preterite to buy into it.

4 It is not a coincidence that Ransohoff published this piece as California is considering a billionaire tax.

5 Insightfully criticized by Tim Schwab in Bill Gates: The Myth of the Good Billionaire.

6 A reviewer has dissuaded me from writing “camp followers” in place of “factotums”.

7 The fiercely empirical Ed Zitrin has caught one of the (potential) new philanthropists hesitating to disclose full financial information to a lender. If that (potentially) generous public benefactor won’t speak honestly to someone who is trying to lend him money, then why would he answer outsiders about how his endowments are used ?

8 Read Ransohoff’s papers on climate (“Red teaming the next ~ 5 years of carbon removal”) or on policies (“There should be general managers for some of the world’s more important problems”) if you think I am being unfair.

9 Readers unfamiliar with the governance of foundations and development banks should consider this argument. (A) The governors (boards of directors, etc) rarely think in terms of the expression productivity = project benefits/project costs; (B) the same governors are only vaguely aware of how project benefits are measured (hint: one has to go out and look); (C) following (B), governors accept weak and incomplete measures of project benefits in the form of hasty project evaluations, which can be fixed and where inconvenient results are easily ignored; (D) governors worrying about productivity and not understanding how project benefits are generated and evaluated, have one budget instrument—cut costs; (E) project evaluation costs are an easy target in the spirit of “we must move quickly, kids are starving”; and (F) project quality enters a death spiral as less is spent on deriving lessons from successes and failures. Parminder Brar at mdbreform.com gives strong empirical support to this line.

10 “Second prize is a set of steak knives”.

11 “How Bill Gates can spend US$200 billion in 2025” (The Apricot Tree of Tangiers, May 15, 2025).