Ten years ago I might have told a story like this. There are a bunch of companies in the world. As the economy grows, their profits will grow. If you buy their stocks, you will get a share of that growth. If you are a normal person saving for retirement, that’s a good thing to do: You want your savings to grow, and owning a share of growing corporate profits in a growing economy is a good way to do that. [1]

The easiest way to do that is to just buy all the stocks in proportion to their market capitalizations, for instance in an index fund. Maybe you can do better than that: Maybe you can buy the companies that will grow a lot and avoid the companies that will go bankrupt. But that requires luck and skill and, crucially, it is zero-sum: If you buy the good companies and avoid the bad ones, that means someone else is buying the bad ones and avoiding the good ones, oops for them. Similarly, you could buy lottery tickets or day-trade stock options to get more than your share of economic growth. Same problem, though: Maybe you can get some extra return by winning the lottery, but that’s not a viable plan for everyone. 

Conventionally, owning a broad slice of the stock market gives you “beta”: It gives you exposure to the market return, which over the long term is mostly positive. Everything else — trying to beat the market by picking the right stocks or options or lottery tickets — gives you “alpha”: It gives you returns to skill, it gives you outperformance (or underperformance), it gives you something other than the broad market return. Beta is generally positive: Everyone can get rich owning the stock market. Alpha is generally zero-sum: Anyone’s outperformance requires someone else’s underperformance.

There’s one other thing that doesn’t fit neatly into these categories. Some companies are not public; they don’t trade on the exchange, and their stocks do not contribute to the market return. Some companies are early-stage tech startups in someone’s garage; others were taken private in leveraged buyouts by private equity firms. Institutional investors can invest in those companies through private equity or venture capital funds. Is that alpha or beta? Some of both. It stands to reason that the smartest private investors would cherry-pick the best companies, help run them well, and get higher returns than the broad average of the public markets. But also, you know, these are companies, they participate in economic growth, the goal is for them to one day (re)join the public markets, and they will have some correlation to the public market return.

Even in 2016, the lines were blurring a bit. I started writing that “private markets are the new public markets” in 2015: Startups were staying private longer and raising more money than they used to. Canonically the term “unicorn,” referring to a billion-dollar startup, was coined in 2013. Back then, a tech company that grew to a $1 billion valuation before going public was rare and magical. By 2016, Uber was a quinquagintacorn with lots of individual investors. 

One other component of this story is about fees. Beta, in 2016, was virtually free. You could buy a broad stock market index fund that charged fees of between 0.02% and 0.05% of assets. Alpha, though, was expensive. It’s hard to beat the market, and people who could do it — hedge fund managers, etc. — could charge fees on the order of 2% of assets and 20% of profits, or much more if they could really beat the market. And then private equity and venture capital funds also tended to charge fees on the order of 2-and-20, because they were basically in the alpha business. In 2016.

Today, I would tell a different story. This story has two main components, “AI” and “private markets are the new public markets.”

The first part of the story is that, when you think about owning your share of future economic growth, you are … probably? … thinking about owning your share of the artificial intelligence boom. Like, the maximalist vision of AI is that, in the very near future, all human economic output will be done by AI, and whoever owns the AI will live in leisure off the AI profits, until the AI kills us. But even a more modest vision of AI is something like “we’re gonna build $10 trillion of data centers in the next few years.” A big chunk of the stock and bond markets are now devoted to raising money for the AI buildout, a big chunk of expected earnings growth is coming from AI, and when people worry about an AI slowdown, the stock and bond markets go down. I wrote last month:

“The artificial intelligence trade,” in some broad sense, increasingly is the economy. [An investor’s] high-level dashboard is a list of like 50 economic factors, and the top factor on the list is like “global economic growth,” and the second is probably “AI stuff.” And your dashboard shows you some positive correlation to the AI factor. And you can turn the dials until the knobs pop off, but you’re not going to get that correlation down to zero.

Ten years ago, if you were like “I’m going to invest in generative AI,” that would have been a very strange and differentiated trade, and people would have looked at you funny and thought you were a kook, and over time you would have made an absolute fortune. (You could have gotten into OpenAI at a valuation of “nonprofit.”) Now, if you are like “I’m going to invest in generative AI,” duh! That’s all anyone is doing! Here’s a Financial Times story about a law firm that wanted to sell its AI business for billions of dollars! Every company is an AI company, etc. etc. etc.

And, conversely, if you are just like “ehhh I don’t know much about investing, I’m just gonna click the default option on my 401(k),” you are also probably getting tons of AI. Tons of AI is just the default option. AI in 2016 was esoteric alpha; AI in 2026 is bog-standard beta. 

The second part of the story is that money is increasingly concentrated in private markets. The unicorns and decacorns and quinquagintacorns of 2016 seem quaint. OpenAI and Anthropic, two of the leading AI labs, are (1) still private and (2) probably worth more than a trillion dollars each. SpaceX, another AI company, went public this year at a $1.77 trillion valuation. There is no particular reason to think that, if you want broad exposure to economic growth, buying the stock market is the way to get it. Anthropic and OpenAI aren’t even in the stock market.

Meanwhile, a lot of the financing for the AI boom comes from private credit and infrastructure funds. The leveraged buyout firms of the olden days have grown into the giant “alternative asset managers” of today, with investments in private credit, real estate, infrastructure, etc. These managers run trillions of dollars for endowments and pensions and (especially) insurance companies and (increasingly) ordinary retirement savers. I have written that, in the early days of private equity, it was essentially about alpha, about spotting undervalued companies and beating the market and being smarter than everyone else, but that can’t scale forever. The returns to being smarter than everyone else are limited; alpha is rare and zero-sum. If you’re running trillions of dollars, your returns are coming mostly from exposure to broad economic growth.

More specifically, your returns are coming from data center financing. Today Bloomberg’s Preeti Singh reports on Blackstone Inc. president Jon Gray and his push into AI financing:

Gray, who’s long been the lead candidate to one day replace Chief Executive Officer Steve Schwarzman, is now making AI the biggest bet of his career. Already, he’s amassed a $200 billion data-center empire and is building a cloud venture with Google. Blackstone has taken stakes in labs like OpenAI and lent against Broadcom Inc. and Nvidia Corp. chips. And it’s committing capital to power plants and data-center cooling equipment. It all shows that the world’s largest alternative investment manager is leaning into AI every way it can.

Even in property, where Gray cut his teeth, AI is now a driver of many deals, such as leasing offices to Anthropic or launching a data-center real estate investment trust. Blackstone is also embedding AI into the hundreds of real-economy businesses the private equity giant controls. And the technology is slowly reshaping the firm’s own operations: a model now weighs in on proposed investments.

Every private markets firm is investing in some piece of the AI boom. But while others have shied away from certain corners of the market, Blackstone has been unequivocal about the opportunity the technology represents, with Gray calling the investments "the single most important thing" underway at the firm.

Of course it’s the single most important thing at Blackstone! How could it be otherwise? There are plenty of small hedge funds and middle-market private equity funds where AI is not the most important thing; plenty of people can make a nice living in pest control. But when you get to a certain scale, you have to invest in the thing with capacity.

The story includes some history of Blackstone’s AI bets: 

In 2023, Gray initially balked at a pitch by dealmaker Jas Khaira, then part of the freewheeling “tactical opportunities” team. Khaira wanted to lend to a former cryptocurrency miner that was now using fast chips, known as graphics processing units, to crunch data for AI models.

“He said, we want to do this multibillion-dollar financing on these GPUs, with an offtake contract from Microsoft. And these guys were ex-crypto,” Gray recalled. “I was like, ‘What?’ But it turned out to be a great opportunity.”

The customer was CoreWeave, a nascent “neocloud” company, and the deal helped kickstart asset-backed financing based on AI chips.

In 2023! In 2023, “let’s do GPU financing” was an idea that came from Blackstone’s “freewheeling ‘tactical opportunities’ team,” and that was met with skepticism at the highest levels. In 2026, “let’s do GPU financing” is like 4% of the economy.

One other component of this story is about fees. If you run a trillion dollars and get broad exposure to economic growth, but that exposure comes in the form of private equity and private credit, what kind of fees do you charge? “Relatively high fees,” is the basic answer: Maybe not 2 and 20, but not 0.05% either. I wrote in April that, “at some level, the rise of private equity and private credit and ‘alternative asset management’ generally is about shifting ownership of the economy to vehicles that charge more fees.” The economy is the AI boom now, and the AI boom has a lot of private financing.

The article also features “AI-generated images of Jonathan Gray portraying him in the 1800s and 1900s,” to drive home, apparently, this point:

“As an investor these are interesting times, because there’s just an enormous amount of transformation underway,” Gray said in an interview — warming to the same theme that he hit in his AI-enabled address to chief investment officers in September. “In the 1880s you would have lived in a world of wood, horses and candles,” he said. “And then, by the 1900s, you were in a world of steel and railroads and electricity.”

“We are in one of those moments.”

Since the railroad boom of the 1900s, the US has developed a whole infrastructure of public financial markets that provide deep and liquid pools of capital and the opportunity for ordinary savers to own the broad market basically for free. Like, that was a big deal. It is just sort of interesting that we’re not really using it for the AI boom.

Elsewhere in private-market beta. Let’s say you own a bunch of private credit loans, it’s some Wednesday in mid-October, and the following things have happened, or not happened:

• None of the companies that have borrowed money from you have announced any news or put out any financial updates today.
• None of them have any payments due today, though they all made their most recent payments last month or last quarter or whenever.
• You have not bought or sold any of the private credit loans today, and you’re not aware of anyone else buying or selling any of those loans.
• But public loans — publicly traded bonds and broadly syndicated loans — to other companies have had a rough time, and are trading several points lower than they were yesterday.

Your loan portfolio was worth $100 yesterday. How much is it worth today? Two possible answers are:

1. Nothing has happened to your portfolio — no news, no financial updates, no problems, no trades, no nothing — so nothing has changed and the loans are worth $100.
2. These loans are loans to companies, and you can look at other loans to other companies for clues about what these loans are worth. Obviously there is some correlation between your portfolio of private loans to AI companies and a basket of public loans to AI companies. If the market is paying 2% less for public AI loans today than it was yesterday, then it stands to reason that, if you had to sell your loans today, I would get $98. Or some other number; perhaps you have reason to believe that your loans are safer or riskier, better or worse collateralized, more or less leveraged to AI worries, blah blah blah; perhaps the answer is $97 or $99. But there is relatively high-frequency public data about debt prices, and you own debt, and surely that high-frequency public data gives you some information about the value of your debt.

The second answer seems pretty obviously right to me, but arguably that is a cultural matter. You can see why some people would like the first answer. For one thing, it is kind of simple and intuitive: There’s no news about your borrowers, so why should the value of your loans change? Who cares about other news about other companies? What does that have to do with anything?

For another thing, if you are running an alternative asset manager, it is very helpful for you to have a low correlation to public markets. After all, people can buy public markets cheaply; they are paying you a lot of money for something else, for alpha, for uncorrelated returns, for high returns with low volatility. “Our private-credit loans perform well even when public debt markets go down” is a really good pitch. Assuming that your loans are correlated to public markets, and then jerking their prices around to reflect jumpy public markets, undermines the pitch. Just keep them at $100.

Traditionally private credit managers do not mark all their loans to market every day, and now Apollo does, and people don’t like it. Bloomberg’s Olivia Fishlow reports:

Monroe Capital’s Ted Koenig isn’t buying that private credit investments can be accurately valued every day.

“Daily marks in illiquid private assets,” he said, “is an illusion of precision, not accuracy.”

The comments come two weeks after Apollo Global Management Inc. began rolling out daily valuations across its roughly $850 billion credit business, a move that is fueling debate at the CAIS Alternative Investment Summit in Beverly Hills this week. Koenig didn’t name Apollo in his remarks at a panel Wednesday, instead taking issue with the idea of daily private credit marks more broadly. …

“To have actual marks, you need financial statements, you need other inputs that you don’t receive on a daily basis,” Koenig added. “There’s just no way to do this.”

His fellow panelist Drew McKnight, co-CEO of Fortress Investment Group, also questioned how meaningful such frequent updates could be without a deeper assessment of individual borrowers.

“If they are just marking the book towards a leveraged loan index, you can do that daily theoretically just on a spread basis,” he said, “but it wouldn’t show you if the loan is at par unless you’re doing the actual analysis.”

Right, if your private credit book is fundamentally uncorrelated with a leveraged loan index, then it makes no sense to mark it based on moves in the leveraged loan index. It would be weird if that were true, though. If your private credit book is perfectly correlated with the leveraged loan index, then you can easily mark it based on moves in that index, but then you aren’t really earning your fees, are you? If it’s somewhere in between, the marks are at least a little helpful.

We have talked from time to time around here about a shift in investment banks’ trading business. A decade ago, a bank’s trading division traded: It would buy stocks and bonds from customers who wanted to sell them, sell stocks and bonds to customers who wanted to buy them, and keep a little cut for itself. These days, though, a bank’s trading division is increasingly in the business of lending: Its customers want to borrow money to buy stocks and bonds, and the bank lends them the money. 

Part of the reason for this shift is that trading has gotten more efficient: Customers can trade stocks and bonds with each other electronically, without paying banks a big markup to connect them. Part of the reason is that big hedge funds and proprietary trading firms now compete in a lot of the trading businesses that the banks used to do. Part of the reason is that, after 2008, banks are more regulated and more conservative about risk; lending against stocks and bonds, rather than owning them directly, allows the banks to “move up the capital structure.” Part of the reason is that there is a lot of demand for leverage in markets these days. Part of the reason is that the banks’ customer base has shifted: These days, hedge funds and proprietary trading firms are bigger, more institutional and more diversified, so they want (and get) more leverage.

These are all vague qualitative points, and I don’t want to overstate any of it. Banks’ trading businesses still do a lot of trading. Still, this week Bloomberg’s Todd Gillespie reported on an arguable tipping point:

Goldman Sachs Group Inc.’s lending to hedge funds beat trading in the bank’s equities division for the first quarter on record, a change that comes as major investors step up leverage while riding volatile markets.

The Wall Street firm posted $3.37 billion in equities-financing revenue in the third quarter, a 96% increase from a year earlier, it said Tuesday. Equities intermediation, where Goldman traders act as middlemen to match buyers and sellers of stocks, pulled in $3.08 billion, a 52% jump.

The rise in so-called prime brokerage, where banks offer services such as financing, custody and matchmaking funds and investors, has been a key driver of equities revenue across banks. Highly levered funds have grabbed headlines this year as volatility, particularly in technology stocks, has driven much of the surge in trading revenue.

Goldman’s stock trading business is now mostly a stock financing business.

The Financial Times reports that “Britain’s financial watchdog is planning to become an ‘agentic regulator’ by harnessing AI to automate much of its supervisory and enforcement activities”:

The Financial Conduct Authority will next week unveil its response to the latest frontier AI models in a speech by Jessica Rusu, chief data, information and intelligence officer. …

Another of the review’s recommendations was for the watchdog to “build and adopt an AI-enabled agentic supervisory model” to stay on top of financial services activities as they are increasingly carried out by autonomous agents with less input from humans. ...

“Becoming an agentic regulator is part of our overall ambition of being a smarter regulator,” she said. “As we consider the various ways we can use that for intelligence gathering and case optimisation, we have been able to look at ways to embed that in authorisations, supervision and enforcement work.”

The regulator is deploying AI in ways that allow its officials to oversee and direct its activities, but Rusu said it aimed to go “further than that, to automate wherever we can ... and to intervene at scale”.

Part of me thinks it would be very weird to have an “agentic regulator” in any meaningful sense. Like, if you get an email from the regulator saying “we think you did insider trading,” and you explain yourself, and the enforcement staff evaluate your explanation and ask follow-up questions and decide to charge you, and your lawyer submits a defense, and the regulator’s internal tribunal considers the charges and the defense and finds you guilty and fines you a lot of money and sends you an invoice, and all of that — the email, the enforcement staff, the follow-up questions, the charges, the tribunal, the decision, the penalty— is done by agentic AI, you’d find that unnerving. “Can I talk to a human,” you would ask, but no, agentic regulator.

But of course Rusu is looking ahead to a world of agentic finance, and in that world this is fine. Like if your AI agent gets an email from the regulator saying “we think you did insider trading,” and your agent explains itself to the agent’s regulator, and there’s some back and forth and a decision and a penalty, and it all happens in 20 seconds with no humans in the loop, that is kind of an efficient regulatory system. Whereas if a human regulator asks your AI agent to explain its suspicious trading, that will be hopelessly slow and prone to misunderstandings. “My trading was based on 10 billion signals that are too subtle for you to understand, human,” your agent will huff.

I have to say, I would love to see the evolution of financial regulation and enforcement when it’s just trading agents trying to pull a fast one and regulatory agents trying to stop them. With human financial institutions and human regulators, it takes years to write new rules and more years to find exploits around them. With AI it could be minutes. It will be weird to write my column, though. “The agentic securities regulator fined an agentic trading firm $1 billion for … doing something too subtly bad for a human to understand.” I guess I’ll need an agent to write the column.

Anthropic Revenue Jumps to $16.8 Billion in Quarter Before IPO. US Pump-and-Dump Crackdown Aimed Abroad Meets Homegrown Protests. A Swiss Banking Giant Thinks the Unthinkable: Moving Out of Switzerland. Credit Suisse Bankers Lose Final Bid to Win Back Bonuses at Swiss Top Court. Credit Suisse secured $800mn from insurers in Greensill settlement. The man who would lead McKinsey. Gas Exporter Venture Global Wants to Get In on AI Power Boom. The Fierce Battle to Control How Americans Buy and Sell Their Homes. Stubborn Sellers Are the Last Thing Propping Up Home Prices. OpenAI Alleges Ex-Employee Took Confidential Files to Apple. He Crafted a Crypto Law for Trump — as His Company Got Millions From a Crypto Giant. “We found our people. We don’t have to be embarrassed or hide the fact that we have a plane and a yacht and five homes. We can finally be open about it.” Farmed salmon now being raised on chicken.

If you'd like to get Money Stuff in handy email form, right in your inbox, please subscribe at this link. Or you can subscribe to Money Stuff and other great Bloomberg newsletters here. Thanks!