I have always had a dumb simple model of OpenAI’s nonprofit/capped-profit/for-profit structure:

1. There is a nonprofit organization that currently controls OpenAI, whose mission is to build “safe and beneficial artificial general intelligence for the benefit of humanity.”
2. There is a for-profit company that employs the employees and raises the money and builds the artificial intelligence models and sells them to consumers and businesses.
3. The ownership structure of the for-profit company is complicated. Different investors (employees, venture capitalists, Microsoft Corp., etc.) are entitled to different cash flows from the company’s future profits. Until recently, every investor’s cash flows were capped at some level: They were entitled to some share of OpenAI’s profits up to some threshold, but after they hit the threshold they got nothing and all of the remaining profits would belong to the nonprofit organization. Also the nonprofit controls the for-profit: The nonprofit’s board makes decisions for the for-profit (like hiring and firing its chief executive officer, etc.), and in making those decisions it exclusively considers its nonprofit mission, not the interests of for-profit investors.
4. But you can ignore much of the complication for most purposes. It is approximately true that OpenAI is a for-profit company whose stock is X% owned by the nonprofit organization and (1 - X)% owned by regular shareholders (employees, VCs, Microsoft), for some X. You can calculate X, with some reasonable degree of precision, using the various waterfalls of cash flows that the for-profit has promised to its investors and some standard option-pricing or discounted-cash-flow models. Microsoft, say, is entitled to ___% of profits up to $___, and then ___% of profits up to $___, and then ___% of profits up to $___, and then 0% of profits above that, but you can abstract away from the details and say “on reasonable assumptions, Microsoft has ___% economic ownership of OpenAI” and be approximately correct. [1]
5. There is room for disagreement on the model to use and the assumptions to plug into it, so actually converting the various capped-profit interests into shares would be a matter of negotiation between OpenAI and its investors. And in fact it has never been quite clear to me what X is, or even whether it is more or less than 50%, that is, whether the nonprofit or investors own a majority of OpenAI’s expected economic value. (Though my impression is that in fact the investors are the majority economic owners.)
6. In any case, whatever X is — whether the nonprofit’s economic interest in OpenAI’s business is 30% of that business, or 50%, or 70% — the nonprofit has super-voting stock, in the sense that it controls the board and will keep controlling the board even as it keeps raising more money from more investors. 

“A regular company with a big block of super-voting stock held by a nonprofit foundation” is unusual, but not that weird. That roughly describes Hershey Co., for instance.

And then the recent news has been that OpenAI was going to “convert” into a more normal public-company structure, which always struck me as less important than it sounded. That conversion consists of:

• Doing the math in my Step 4 above, and reducing all of the various weird profit waterfall interests into common stock; and
• Making sure that X is less than 50%, and eliminating the nonprofit’s super-voting stock, so that OpenAI will be governed by a more normal for-profit board that is answerable to all of its shareholders, including both the nonprofit and the investors.

So there are two elements to the “conversion,” but they are conceptually distinct. You could do either one without the other. You could very easily say “all this capped-profit stuff is weird, everyone’s just going to get stock, but the nonprofit foundation will get 34% of the stock and 99% of the votes.” Or you could say “the capped-profit stuff is fine, but we’re going to give the for-profit investors board representation consistent with their economic ownership, which by the way is more than 50%.” 

Anyway yesterday OpenAI abandoned the conversion, sort of, or rather it abandoned the second part (giving investors control of the board) but kept the first part (converting the weird stuff into common stock). Here’s how OpenAI Chief Executive Officer Sam Altman puts it: 

We made the decision for the nonprofit to stay in control after hearing from civic leaders and having discussions with the offices of the Attorneys General of California and Delaware. We look forward to advancing the details of this plan in continued conversation with them, Microsoft, and our newly appointed nonprofit commissioners.

OpenAI was founded as a nonprofit, is today a nonprofit that oversees and controls the for-profit, and going forward will remain a nonprofit that oversees and controls the for-profit. That will not change.

The for-profit LLC under the nonprofit will transition to a Public Benefit Corporation (PBC) with the same mission. PBCs have become the standard for-profit structure for other AGI labs like Anthropic and X.ai, as well as many purpose driven companies like Patagonia. We think it makes sense for us, too.

Instead of our current complex capped-profit structure—which made sense when it looked like there might be one dominant AGI effort but doesn’t in a world of many great AGI companies—we are moving to a normal capital structure where everyone has stock. This is not a sale, but a change of structure to something simpler.

The nonprofit will continue to control the PBC, and will become a big shareholder in the PBC, in an amount supported by independent financial advisors, giving the nonprofit resources to support programs so AI can benefit many different communities, consistent with the mission. 

Fine. Note that “the nonprofit will continue to control the PBC, and will become a big shareholder” suggests that the nonprofit will be a minority shareholder: The nonprofit might own (say) 30% of the economic value of the for-profit company, but it will have super-voting stock giving it control of the board. 

The point of this, I suppose, is to be enough of a nonprofit to satisfy nonprofit law, but also enough of a normal company to satisfy normal investors. For investors, a weird capped-profit interest is significantly more troublesome than corporate stock, but corporate stock in a company controlled by a nonprofit board is conceptually fine. People own stock in all sorts of companies with weird controlling shareholders; Elon Musk has no trouble raising money. “You will own 20% of our company, but you won’t be able to elect the board, and the board will have limited fiduciary duties to you” is pretty normal.

On the other hand, “you will own 20% of our company up to $10 billion in earnings, and then 40% up to $40 billion, and then 0% above that” is super weird and annoying. Fixing that is easy, though it requires some calculations “supported by independent financial advisors.” And in fact OpenAI’s more recent funding was contingent on getting rid of the profit caps, and this half-restructuring seems to satisfy that requirement: “Altman said the changes proposed Monday would still allow it to access a $30 billion chunk of investment from SoftBank, which had been dependent upon the successful restructuring.”

For nonprofit law, I mean, the weird thing was not the recent “conversion” plan. The weird thing is that OpenAI spun up a capped-profit subsidiary in the first place, which it did in 2019. That made sense: Building safe and beneficial artificial general intelligence turned out to be way more capital-intensive than OpenAI had originally expected, and to raise billions of dollars OpenAI needed to offer investors a return, not just ask for donations. So making OpenAI’s business a for-profit subsidiary of a nonprofit entity was probably necessary, but it’s certainly awkward: If you’re raising money from investors who want a return, there is some unavoidable conflict between your mission of serving humanity and your desire to please investors (and employees!) so you can raise more money. Humanity might be happier paying $10 a month for a ChatGPT subscription, but the investors might be happier at $20.

But that happened, and it seems hard to reverse now. [2] OpenAI’s previous conversion plan would have ended the nonprofit’s control of the for-profit business, which seems to have annoyed the state attorneys general (who supervise nonprofits) and also Elon Musk, who sued to stop it. The annoyance is roughly that, if the nonprofit board had control of the company and then gave it up, it sort of looks like giving away nonprofit assets (which are supposed to be used to benefit humanity) to for-profit investors (who will use them for themselves). I don’t think that this is quite correct. I think that if one day the nonprofit owned 51% of OpenAI’s business, and the next day it sold more stock to bring it down to 49%, that would be fine (if it got fair value for the stock). But it is understandable that people might think that OpenAI was not getting fair value for giving up control to for-profit investors, and in fact one bit of trolling that Musk did was that he offered to pay more for the nonprofit’s stake in OpenAI than the value that the conversion put on it.

Now I suppose that problem is fixed: The nonprofit is not giving up control, so nobody needs to worry that it is undervaluing that control. Of course the nonprofit will now have an economic ownership of the for-profit “in an amount supported by independent financial advisors,” and you could quibble with whatever that amount is, and Elon Musk will:

“This changes nothing,” Marc Toberoff, Musk’s lead counsel in his suit against OpenAI, calling the move a “cosmetic restructuring” that converts charitable assets into private billions. “The founding mission remains betrayed.”

I mean, yes, if you are selling stock in a nonprofit company, something weird has happened. 

But this does make OpenAI … a pretty normal Silicon Valley company, doesn’t it? It’s a giant tech company with a charismatic visionary founder who does not want to deal with the short-term pressures of being answerable to shareholders. OpenAI “was not originally created to be a company. It was built to accomplish a social mission,” and “it’s important that everyone who invests … understands what this mission means to” the company, which has “always cared primarily about [its] social mission,” not about making money. Those quotes are from Mark Zuckerberg’s 2012 letter to Facebook investors; this is totally normal stuff! Sam Altman can prioritize his duty to humanity over making money for shareholders, just like Zuckerberg does, and Elon Musk, and everyone else in his social circle. If there’s one thing that Silicon Valley tech companies know, it’s that a social mission is good for profits.

Of course technically Altman doesn’t control the company; the nonprofit board does. But, uh. You know. “The nonprofit board is now much more tightly under his control” than it was when it fired him last year, as Axios’s Ina Fried notes. If he does what he thinks is best for humanity, the board seems unlikely to stop him, and the shareholders can’t.

If you want to buy a diversified portfolio of stocks of publicly traded technology companies, that’s pretty easy. There are various actively managed tech-focused mutual funds that will do it for a fee, and there are tech-sector index funds and exchange-traded funds that will do it for annual fees of under 0.1%. Or you could just buy the stocks yourself: They all trade on the stock exchange, anyone can buy them, and I bet you could think of a list of 10 big public tech companies off the top of your head.

If you want to buy a diversified portfolio of stocks of private technology companies, that’s much harder. You could probably think of a bunch of big private tech companies off the top of your head — hint, OpenAI is one of them — but you can’t necessarily buy them; their stocks are not traded on any public exchange. Perhaps you are an “accredited investor” and you can buy some of them through your broker or wealth manager, or through a secondary trading platform. But you might not get exactly the companies you want, and diversification might be hard, and certainly getting index-like exposure to the broad private tech sector is difficult.

Obviously a very appealing product would be a fund of private technology companies. Some investment manager who does have access to lots of cool big private tech companies could bundle their shares into a fund and sell them to investors who want diversified exposure. A lot of people seem to want that, but it is not that easy to find. There are some options. ARK Invest has a smallish venture capital interval fund that holds shares in SpaceX, OpenAI and XAI. We talk sometimes about the Destiny Tech100 fund, which is very approximately a diversified fund of cool private companies, and which trades at an enormous premium to its net asset value. That suggests that the demand is high and the supply is low.

You know what seems to be a little easier to find? A diversified portfolio of stocks of public and private technology companies. There are various normal public mutual funds that also include a bit of racy private tech; Baron Partners Fund, for instance, has a big SpaceX holding, and the reason there was quarterly reporting about the value of X (formerly Twitter Inc.) as a private company is that Fidelity mutual funds owned some of it. And now the Wall Street Journal reports:

The often opaque world of Silicon Valley venture capital is opening up a bit wider to individual investors.

Hedge-fund giant Coatue Management is launching a fund that invests in high-growth public and private tech companies, and requires a minimum investment of $50,000. The family offices of Amazon Jeff Bezos and PC company founder Michael Dell have committed a total of $1 billion to seed the new fund, according to Coatue.

“It’s such an obvious idea that if I don’t do it, someone else will,” Philippe Laffont, Coatue’s founder, said in an interview. He said he personally pitched Bezos and Dell, including one over dessert at lunch. (Laffont didn’t divulge which of the two.)

The new fund will invest around 20% to 50% of its assets in private investments, and the rest in public stocks, according to a pitch deck seen by The Wall Street Journal. …

Coatue’s new fund could be a lucrative new revenue stream. It will charge an annual management fee of 1.25% plus 12.5% of the fund’s annual profits, providing the profits exceed 5%.

Redemptions won’t be as easy as with publicly listed funds. Investors will be able to sell their shares once a quarter, with Coatue capping total redemptions at 5% of the fund. Those who sell in the first year will incur a 2% penalty.

Why is this an obvious idea? Why is “half-private half-public stock fund” so appealing? He’s not alone in this; we talked last week about a joint venture between Capital Group and KKR & Co. that is “working on public-private equity solutions.” There is a huge push to sell private investments to individual investors, but that push often takes the form of “a fund that is half public and half private,” or maybe one-quarter private and three-quarters public. Why is that the correct model, rather than “here’s a fund of private companies”?

I don’t know, but here are some ideas. First: It seems to me that there is an important distinction between “public tech company” and “private tech company,” that those are in some sense different asset classes with different characteristics that appeal to different investors and serve different functions in a portfolio, but maybe that is silly and narrow-minded of me. Plausibly to an investor or a professional investment manager, “big Silicon Valley tech business” is the relevant category, and distinctions between private and public companies are minor and arbitrary. An investment manager can focus on finding the best big technology companies, whether they are publicly or privately held; limiting her to buying only private or only public companies would cut off half of her ability to add value, for no real reason. “Private markets are the new public markets,” I often say around here, because companies can stay private while being huge household names and raising money and giving investors liquidity and doing all the things that only public companies used to be able to do; if you take that seriously then there’s no reason for a fund manager to limit herself to only public or only private companies.

That is the conceptual answer, that there is no important distinction between public and private companies and therefore a fund manager should be agnostic about form of organization. There are other, more practical answers.

For instance: If you run a fund, particularly one with lots of smaller-dollar individual investors, you will be inclined to offer redemptions. It is helpful, in asking individuals to pitch in $50,000, to tell them that they can get their money back. “You can get your money back once a quarter, capped at 5% redemptions” is more appealing to individual investors than the institutional venture pitch of “your money is locked up for 10 years,” [3]  and “you can get your money back whenever you want” is even better. [4]

If you run a fund that is 100% invested in private tech companies, and an investor asks for her money back, that’s a problem: You will have to sell some private shares, and those shares are still less liquid than public stocks are. If a lot of investors ask for their money back, you will need to do a fire sale of illiquid private assets. But if you run a fund that is 50% public and 50% private, and some investors ask for their money back, you can quickly raise cash by selling some liquid public stocks without moving the market much. (And then, at your leisure, you can rebalance by selling private stocks and buying back public ones.) If a lot of investors ask for their money back at once, that won’t be great — you’ll sell too much public stock and be overindexed to private companies — but it’s still better than the alternative. If all of the investors ask for their money back at once, of course, you’ll still be stuck dumping the private stocks, but why worry about that now?

Arguably this is not a great first-principles reason to combine public and private funds — in a sense you are offering more liquidity than you can actually provide — but it seems practically useful in raising money.

Here’s one more potential practical advantage of public-private funds. The way an actively managed mutual fund traditionally works is that an investment manager manages a pool of investor capital and buys a bunch of stocks that she thinks are good. For this work, she charges the investors a fee; traditionally the fee was more than 1% of assets, but in recent years those fees have been squeezed by the rise of indexing. It turns out that, in many cases, you can get most of the performance of the actively managed mutual fund — perhaps 80% of the performance, or 100%, or frequently 120% — by just buying an index fund, and the index fund will probably charge you less than 0.1%. The mutual fund manager was charging you a fee for her skill in picking stock, but only a small portion of your returns came from that skill. Most of the returns came from market exposure, but you paid the skill fee on the entire amount of your investment.

This is even more obnoxious when you invest in a hedge fund: If you are paying 2% of assets and 20% of gains, and the hedge fund is mostly just giving you exposure to the stock market, you are wildly overpaying. And so most big modern hedge funds are careful not to do that; they try to avoid correlation to the broad market and rigorously demonstrate that you are paying only for skill, not for market exposure. “Paying alpha fees for beta is bad,” as Cliff Asness puts it. A plausible investment plan might be “I’ll put most of my money in index funds and pay zero-ish fees, and I’ll put some of my money in alternatives — hedge funds, venture, private equity, etc. — that demonstrably provide alpha and charge 2-and-20ish fees.”

If you are an investment manager, this is a bit unpleasant: “Paying alpha fees for beta is bad,” but of course charging alpha fees for beta is good. You want to be paid for giving investors broad market exposure. But it’s hard, because everyone knows this; charging high fees for actively managed mutual funds is an increasingly tough business.

How do you fix that? Plausibly the answer is:

• If you give people public exposure, in any form, the fees you can charge them are approximately zero.
• If you give people private exposure, the fees you can charge them are approximately 2% plus a cut of the profits.
• If you give people blended public-private exposure, the fees you can charge them are not simply the linear combination: If you offer a 50/50 public/private fund, the fees the market will accept might not be “50% of zero plus 50% of 2% is 1%,” but rather much closer to the full private fees.

That is, investors might have only two thought processes, “ooh cool private exposure, that’s worth paying for” and “ugh public stocks, that’s not worth paying for,” and anything that is sufficiently “ooh cool private exposure” might get to charge private fees. [5]  And so if you mix public stocks with private stocks, you might get to charge private fees for the whole lot of them. 

If you are a big US tech company that runs an app store, and if your app store offers TikTok, are you breaking the law? I mean, technically, there is a validly enacted US statute, upheld by the US Supreme Court, saying that “it shall be unlawful for an entity to distribute, maintain, or update (or enable the distribution, maintenance, or updating of)” TikTok, including by hosting it or distributing it in an app store. So, I guess you’re breaking the law? (Not legal advice!) But President Donald Trump has issued an executive order saying that the US Department of Justice “shall take no action to enforce” the law until at least this June. [6]  

And so Apple Inc. and Alphabet Inc. still distribute TikTok through their app stores, and Oracle Corp. still seems to be keeping it online. They … are breaking the law but won’t get in trouble for it? Maybe? Or, like, the law says they can’t host TikTok, but Donald Trump has strongly implied that they should host TikTok (so that he can make a deal for US investors to buy it), and doing what Trump wants is more important than following the law? I don’t know.

We have talked about similar problems before, in the context of the Foreign Corrupt Practices Act, a US law that prohibits bribery but that Trump has also announced he will not enforce. There, I noted that most companies still seem to comply pretty strictly with the FCPA, because most companies don’t want to break the law, even if the law is not currently being enforced by the Department of Justice. If you break a federal law you can probably get in trouble with someone, and it is generally best policy not to break the law. (The TikTok situation is maybe a little different. It’s not like you have to pay bribes to foreign officials to satisfy Trump; it’s possible that Apple does have to host TikTok.) 

In any case. Any time a public company does something bad, or arguably bad, there is a second-order problem, one that I discuss frequently around