The AAUP is wrong about the Anthropic IPO
The truly scary inhuman intelligence is the market
(I’m Henry Snow, and you’re reading Another Way.)
On November 13th, the American Association of University Professors issued an unusual statement attempting to block the IPO of a private firm— AI developer Anthropic. I’m a dues-paying union member of the AAUP, and I think this statement was largely a mistake. I don’t often find myself defending billion-dollar companies, let alone trillion-dollar ones. Tomorrow Anthropic is expected to go public at an initial valuation well over $1 trillion. But I think this well-intentioned letter reflects misplaced priorities and a mistaken approach.
The AAUP letter offers four arguments against Anthropic’s IPO, none of which are without merit. First, the company will “likely begin trading at an extremely high, speculative valuation,” allowing “insiders” to “cash out” while leaving “working people” holding the bag. This happens through index funds— funds that match the composition of indexes, like the S&P 500, which track significant firms. Indexes require certain criteria for a company, and the most important of these is generally “is the company very valuable.” In total then an index fund leaves you with a panoply of investments in large companies that probably won’t all implode at once, rather than one particular sector or firm. For some time now, the general responsible investing strategy has been “throw money in an index fund and don’t touch it”— most people aren’t going to effectively pick winners and losers, some economists would argue it is impossible to do so anyway, and therefore betting on the market as a whole is the best you can do.
Unfortunately, in an era of rising inequality and wealth concentration, index funds are prone to abuse. If a firm can make its way onto an index— in other words, get a big enough speculative value to be considered a major firm and appear in, say, the S&P 500— then it immediately unlocks guaranteed buyers. This gives millions of people a financial stake in that company’s profitability overnight. Maybe you disdain SpaceX, but if Elon gets his way into your retirement fund then the government can’t crush them for fraud or abuse without screwing you too. And if the company runs into trouble, it has to be bailed out.
As you can tell, I’m sympathetic to this argument. But delaying investment in one IPO is not a structural solution. This letter is first and foremost a signal, because I’m not sure anyone is going to be swayed by it. That’s fine— boycotts can be an organizing tool. But it’s organizing and communicating on flawed premises at best. We need reforms to how index funds work, and certainly I and probably many other AAUP members could devise all kinds of investment schemes that in the long term would make the world a little less focused on the market. I don’t think you get there by playing whack-a-mole with unpopular firms or even industries— if you want to turn anti-datacenter sentiment into index fund or pension reform sentiment you have to make that case.
Meanwhile, if anyone does listen to this letter, it could be counterproductive. Anthropic might be undervalued. Plenty of university professors dealing with AI cheating every day don’t want that to be the case, but harmful things can be immensely profitable! The costs of bad advice here are high. If your state pension officer takes their advice and there isn’t a bubble, that causes real financial harm. Do I really trust the AAUP for investment advice? The whole point of index funds is that you can’t trust anyone at all. The market is fickle and unfair, but that doesn’t mean you can easily beat it.
This is a reader-supported newsletter! If we lived in a world of thriving magazine and book publishing, and well-funded public institutions, I wouldn’t need to ask for a cent. Unfortunately we don’t, and I appreciate your support. In addition to the two articles a month free subscribers get by email, premium subscribers receive one bonus post a month. On that note, here’s another reminder that you can purchase my book Control Science from Verso right now. I also have a new personal website, henrys.now, where you can find more of my work.
I worry bubble predictions are wishful thinking by my colleagues who want AI to be a scam that fails. They— and this used to be me, too— have been wrong repeatedly about this. AI critics have been warning of a bubble for years. If you took their investment advice you missed out on major returns. Markets are not an arbiter of truth or justice. They won’t save us. In the long run everything is a bubble.
Maybe this “bubble” will pop soon! Jensen Huang’s recent ouster from Nvidia threatens to collapse large parts of the AI market due to informal compute deals. If I worked at Anthropic I’d be nervous right now. If I were a state pension officer I probably would hold off on making any kind of investment this speculative! But I wouldn’t be sure. I wouldn’t want to bet on it. And “the AAUP is mad about AI” should not be understood as a signal that has any value as investment advice.
The AAUP’s second argument against the IPO is more sound— but still suggests a better approach than the one they are taking. Their letter explains that holding Anthropic stock would not mean having power over company decisions the way shareholders should. This is a typical problem in the tech industry, and one we should fix legislatively— it is ridiculous that tech founders rule like kings. But again, much of your retirement savings are regrettably already invested in firms like this.
Moreover, Anthropic theoretically has a very good reason for limiting shareholder power here. From the beginning, the company was organized to balance an effective altruist conception of ethics with the profit-making necessary to fund it. That means requiring itself to do things that aren’t profit-maximizing. I am immensely skeptical of this effort— EA ethics are deeply flawed, and clever corporate architecture is unlikely to overcome the sheer power of capitalist markets. But why single them out for trying? This makes far more sense as an argument for divesting from Mark Zuckerberg’s kingdom of Meta rather than avoiding this one IPO.
The third and fourth arguments are more generic criticisms of AI itself. Three is more bubble fears. It’s unfortunate that so much of our economy is bound up in AI— but it is. “You have a responsibility to protect our investments from the AI bubble” they tell officials— sure, but pension funds also have an obligation to maximize returns and not to make investment decisions based on, say, angry professors. We might not like the results, but there are good reasons for this.
More importantly, states should not need pension funds invested in the market at all. Right now a state pension works like this. Connecticut takes some of your money now, invests it, and uses the returns to fund future spending when you retire. This is unnecessary and roundabout. Instead, they could simply… promise you money later. If the state wants to invest in the meantime that’s fine, but it shouldn’t have to, and your pension should not depend on that. State governments should be able to commit to taxes now and payments later— there’s no reason that retirement money needs to be booked as an “asset” to be invested now and spent later. A private company does this in case they go bust, cease to exist, and therefore liquidate all obligations, but states can’t do this! Perhaps there’s some value in state governments making investments in order to juice revenues, but there’s no need for them to tie retirement to investments like this.
The only reason we treat pensions like we do is a lack of trust in ourselves. Voters can’t be trusted not to cut spending in the future and target pensions, so workers understandably demand money set aside now. But that’s a paper restriction. In a true emergency governments might raid pension funds— do you really trust voters and courts? They probably won’t and this would be illegal, but we could make it just as illegal not to pay out future pensions. Financial markets’ role in government pensions is fundamentally parasitic: we don’t trust our own public, so we let our money pass into private hands as a caretaker, juicing corporate valuations. Are solutions easy or obvious? No. But they are available, and we should be pursuing those instead of fiddling with individual cases. (Here’s some research on this by Tom Sgouros for Berkeley’s Haas Institute, which has influenced my thinking here.)
Fourth and finally the AAUP letter points to serious harms by AI writ large. Some of these are overblown and some underestimated, but the response here should be regulatory action, not calling for pension officials to punish an IPO. Even if this were effective leverage, a privately-held AI developer is perfectly capable of doing all these same harms— cynically you could say that if we’re going to do nothing else we might as well at least try to socialize any benefits. I don’t think cynically and cynicism is usually bad. But keeping obviously very powerful firms private does not keep them from doing harm, and can exacerbate it. Indeed, the entire point of the complaint about marginalizing shareholders is that it keeps too much power in private hands.
The AAUP has also circulated a petition directed to private financial CEOs, like BlackRock’s Larry Fink, asking them to boycott the IPO, “pause all investments in AI and datacenters,” and “pledge not to bail out AI companies” with public money. Why on Earth would anyone trust a voluntary pledge by JPMorgan Chase CEO Jamie Dimon? A union should not be begging a banking CEO to deliver social change. This is ineffective and repellent. I want investment decisions made in a democratic way— pleading with the wealthy and powerful to be responsive to the public is representative, but not democratic, in the same way threatening the King’s ministers once was. I get the do something instinct, but this is the wrong thing.
And unfortunately bailouts are sometimes necessary. We should be acting to avoid their necessity in the first place, not to stop them once they are necessary. It’s much too late for this. By the letter’s own reasoning, a bubble popping here has the potential to cause a massive financial crisis. Should we just let that happen if it’s going to? Should we wipe out investors and states and retirees who have money in Nvidia? This doesn’t seem fair or effective to me. “Let the market punish the unworthy” is a classic free-market libertarian argument, not a working class one. We could talk about how a bailout might be structured differently, avoided in the future, etc. But if you buy other premises in their very letter, opposing a future bailout is catastrophic.
All of this I fear reflects an unseemly tendency in our financialized world to look to markets as at best the primary battleground of political and economic conflicts, or even worse, as their arbiter. The popular writer Ed Zitron has done quite well for himself on his own newsletter insisting that the bubbble will burst any day now— for years. (Oddly, Zitron is also a PR guy. Including for AI firms. Interesting). (A premium subscription to this newsletter, in contrast, is a financial commitment to getting my best effort at the truth even when you or I might not like it!). His demonstrably false predictions are popular— I see colleagues mention him often— because they promise salvation without actually doing anything, in an age where doing anything feels difficult or impossible. Fret not! The market finds the True Value of the firm eventually— the bubble will pop, the righteous will be delivered, the unworthy doomed.
Markets cannot be trusted to do this. The 2008 financial crisis punished Black homeowners and state pension funds, not bankers— only direct government action mitigated this, and then only partially. Markets do not care about you or about anything. We retreat into the world of markets because elected officials seem to do less and less— a malapportioned Senate and gerrymandered House chokes any regulation, American aversion to taxes strangles state and local budgets even and at times especially in blue states, and the Supreme Court has largely decided laws Democrats pass don’t count. But these are developments to contest and not accept. I don’t want decisions about the public good to be made with one eye glancing down at the Fidelity app.