Subscribe to Bloomberg.com for unlimited access to all our coverage.

I guess the sci-fi big picture here is something like:

1. In the future, artificial intelligence will do a lot of the useful labor. If you want a merger analysis or a medical diagnosis or a fun newsletter to read, an AI will create it for you.
2. The AI will in many respects be cheaper than human labor: It won’t need to eat or sleep or take nice vacations or send its children to college.
3. But the AI will need somewhere to live. AI lives on computer chips in data centers, and it will need to pay rent. 
4. Humans will own the data centers, and the supply of high-end chips and electric power will be somewhat constrained, and most of the value of the AI future will ultimately be paid out in data-center rents.
5. In the long run, the main human economic activity will be owning chips to rent out to AI, which will do the rest of the economic activity.

You could sort of imagine a Jane Austen-y early-19th-century economy, with AI as the farm workers doing the economic activity and humans as the aristocratic rentiers who own land and earn 5% in perpetuity. And then the structural changes you’d need to get there are:

1. Putting “private investments” like infrastructure funds and private credit into ordinary people’s retirement accounts, so that everyone can be at least a small-scale AI rentier, 
2. Raising a gazillion dollars of private-credit and infrastructure funds, and
3. Using the money to build the data centers to rent to the AIs.

There are other approaches — “everyone should buy stock in the AI buildout,” for instance — but the rentier approach has a certain simplicity. Who knows which AI model or harness or end user will ultimately capture the upside from AI? Whatever it is will probably need chips, though, so there’s a steady living to be made from renting out chips.

Obviously one can worry about a bubble. But I must say that the two main dynamics I write about are “we should invest everyone’s retirement savings in AI infrastructure” and “we should invest everyone’s retirement savings in sports gambling,” and only the first one has some actual economic logic to it. The point of investing is to own a share of future economic productivity. If a big share of future economic productivity might happen on computer chips, you might as well own chips.

Anyway:

US investment giants including Apollo Global Management Inc., Blackstone Inc., BlackRock Inc. and Brookfield Asset Management are partnering with Nvidia Corp. to source $500 billion in financing for artificial intelligence infrastructure.

The coalition, which also includes Goldman Sachs Group Inc. and KKR & Co., will “create dedicated pools of capital at significant scale at attractive rates for Nvidia customers,” according to a statement Monday. Nvidia Chief Executive Officer Jensen Huang said in a CNBC interview that he approached only the six firms for the commitment, and none turned him down.

The effort comes with a huge headline figure but few details on the timing and structure of the financing, or how much the plan goes beyond the string of AI deals that are already driving a large chunk of Wall Street’s biggest transactions. Executives indicated that it will focus on debt financing to provide access to compute for Nvidia’s largest customers and that there are already many deals in the works that would qualify toward this commitment.

From the announcement:

“Modern compute has emerged as a scarce, mission-critical asset class with compelling investment characteristics that is positioned to drive significant long-term economic growth and productivity gains,” said Apollo President Jim Zelter. “The combination of NVIDIA’s proprietary technology ecosystem and Apollo’s flexible, long-term capital base provides a strong foundation to support the next stage of the AI buildout as part of the broader Global Industrial Renaissance.”

Part of what is going on here is that Nvidia is competing to be the main supplier of computer chips to AI, and making its chips easier to finance is helpful for maintaining its lead. Huang posted on X that “in some cases, NVIDIA may provide a residual-value support mechanism for up to 25% of an opportunity”: Nvidia is probably more optimistic about (and more directly invested in) the AI buildout than most investment managers are, and using its own giant investment-grade balance sheet to support AI projects makes it easier to raise debt. Ben Thompson writes:

The company is backstopping opportunities with up to 25% residual-value based financing, suggesting that Huang believes his “investable asset class” pitch much more than the market does. That is, in a certain sense, a price cut, as the goal is to reduce the cost of capital for entities building data centers with Nvidia chips, by putting Nvidia’s profits on the line for uncertain investments.

But as the Financial Times notes, this arrangement “also shows how Nvidia is building relationships with the giants of the private capital industry, which are collectively preparing to invest trillions of dollars of their insurance, retail and institutional investor assets into AI infrastructure.” A major theme in financial markets over the past few years has been the rise of “privates,” and the push to put retail investors’ savings into private assets. Sometimes this means selling private funds to retail investors in their retirement accounts; other times it means selling them annuities or life-insurance products or pensions that invest in private assets. Private assets are more lucrative for the financial industry than public ones, and selling trillions of dollars of private assets to everyone is good business. But where will the industry get those trillions of dollars of private assets? In data centers, apparently.

Elon Musk’s pay package at Tesla Inc. rewards him with giant piles of stock if he achieves certain valuation and operational goals. There are 12 tranches of stock, and each one unlocks based on some combination of (1) Tesla’s market capitalization, (2) Tesla’s earnings and (3) specific product goals like having a million robotaxis in operation, delivering a million humanoid robots, having 10 million full self-driving subscriptions, etc.

When you think about it, this is an unusual level of micromanagement for Elon Musk. The main way that investors manage Elon Musk is by giving him money, leaving him alone and seeing what he gets up to. His job, as a steward of investors’ capital, is to do whatever he thinks will maximize value. A lot of investors gave Musk billions of dollars to buy Twitter. Now Twitter is, um, a space rocket company with a $26.5 trillion total addressable market in artificial intelligence. Those investors have made a fortune.

Imagine, though, if they had put conditions on their investment with Musk. “We’ll help you buy Twitter, but we want our money back if you don’t increase ad revenue by 50% within the first two years.” “You have to commit to increasing user engagement on Twitter by 10% per year.” “You need to sell at least $10 million of premium Twitter subscriptions.” “You have to get rid of the bots.” Stupid! All of these operational metrics that you might have thought were relevant to Twitter shareholder value, and that turned out not to be. The relevant metrics were, like, data centers in space.

I mean, the relevant metrics are data centers in space, in the sense that Musk’s pay package at SpaceX (the successor to Twitter!) does contain operational targets including “non-Earth-based data centers” and “a permanent human colony on Mars with at least one million inhabitants.” Again: micromanaging! Ambitious sci-fi micromanaging, sure, but still a constraint on Musk’s imagination. Last year, it seemed to Tesla’s board of directors (and Musk) that the most likely and durable way to maximize Tesla’s value would be with humanoid robots and self-driving robotaxis. This year, it seemed to SpaceX’s board of directors (and Musk) that the most likely and durable way to maximize SpaceX’s value would be with Mars colonies and orbital data centers. 

Next year, though, what if Musk decides that the best business model for Tesla is to invent and sell teleportation devices? What if he decides that the best location for SpaceX’s data centers is in the deep ocean, or in the Earth’s core, or in the human spirit? He can produce business pivots much faster than public-company boards of directors can produce moonshot pay plans, and all of his pay plans will quickly feel a bit dated. Tesla and SpaceX will not be getting the full Elon Musk experience, if Musk is forced to think about how to build self-driving cars or orbital data centers rather than whatever he comes up with next.

In practice, there are a number of escape hatches for this problem. Most straightforwardly, Musk can start a new company, do whatever new thing he wants in that new company, quickly make it worth a trillion dollars, and then sell it at a premium to one of his other companies. His ability to extract value from his ideas, for himself and for his investors, is not limited to his formal pay packages. 

But the formal pay packages have some escape hatches too, the Wall Street Journal reports:

Deep in a $1 trillion pay package approved by Tesla’s shareholders is an escape clause that could pay off for Elon Musk in several ways if he folds the automaker into his SpaceX empire.

That kind of cars-and-rockets tie-up — the subject of widespread speculation among investors, and hinted at by Musk himself — would likely give him the tighter control he has long sought at Tesla. A hefty-enough price would also, in an instant, wipe away key performance targets standing between Musk and billions of dollars in shares.

“This $1 trillion—that was supposed to be a stretch,” said Mary Ellen Carter, a Boston College accounting professor who studies executive pay. “It turns out it isn’t really that hard. All you have to do is be bought.”

Musk has considerable say over any offer SpaceX makes for Tesla, thanks to his sweeping control over the rocket company. The upshot: Even a stratospheric bid would keep Musk in firm control of the combined company. …

Tucked into the bottom of the fifth page of the 16-page 2025 CEO Performance Award Agreement, a single sentence declares half the targets are as good as accomplished if Tesla is acquired or otherwise taken over.

“In a change in control, the earning of tranches will be based solely on the Market Capitalization Milestones,” is that sentence: If Tesla is bought, including by SpaceX, Musk will be rewarded only for how much value he creates, not for how he creates that value. Full self-driving? Robotaxis? Who cares! What you want from Elon Musk is unconstrained maximization of shareholder value, and in a merger that’s what you get.

The basic state of affairs in private markets, over the past few years, was that if you had $100 worth of SpaceX stock, you could sell it to eager individual investors for $200. There was an institutional market where SpaceX traded at one price, and a fragmented opaque retail market where SpaceX traded at a bunch of different prices, most of them higher than the institutional price. This mostly took the form of special-purpose vehicles that owned SpaceX stock and sold shares in themselves — at substantial premiums, and/or with substantial fees — to individual investors. But there were other formats, funds and futures and tokens that offered some form of SpaceX at some price. Now SpaceX is public and this trade has, for the most part, gone away, but there are similar trades in other big private companies.

Why? When you sold $100 worth of SpaceX for $200, what were you getting paid for? As far as I can tell there are about three answers:

• Access: Not just anyone could buy SpaceX in the institutional market. But you could. 
• Aggregation: The institutional market has bigger ticket sizes than most individual investors could afford; you could collect small checks from a lot of individual investors and write one big check to buy an institutional quantity of SpaceX.
• Confusion: SpaceX did not trade on a public market with a clear price, and people might just not have understood how much they were paying you or for how much stock.

I have casually been assuming that you had $100 of SpaceX stock to sell for $200, but that is not essential. If you had (1) access to SpaceX stock, (2) the ability to aggregate individual investors’ money and (3) the ability to confuse those investors, you could raise the money first and then acquire the SpaceX stock. Like:

1. Go to investors and say “give me $200 to buy SpaceX stock.”
2. They give you money.
3. You buy $100 of SpaceX stock for them.
4. Also you have $100 for yourself.

You could imagine doing this trade on the up-and-up: “If you give me $200, I will keep $100 for myself and buy $100 of SpaceX stock for you, which is a good trade because it’s not like you can buy $100 of SpaceX stock yourself,” you could say. In practice, though, that pitch sounds bad, and the trade works better if you add some confusion. “If you give me $200, I will buy $200 of SpaceX stock for you” is a better pitch, even if it is not exactly true. SpaceX didn’t trade publicly and who was to say what its price was. Buy $100 worth of SpaceX stock, call it $200 worth of SpaceX stock, easy.

Now that SpaceX is public, as I mentioned last week, a lot of people are going to discover that they didn’t own the SpaceX stock they thought they owned. Others are going to discover they overpaid for it. Here’s a US Securities and Exchange Commission enforcement action from yesterday:

The Securities and Exchange Commission today charged New York-based investment adviser Adit Ventures Management LLC, its CEO Eric Munson, and three affiliated general partners, Adit Ventures LLC; Adit Ventures II LLC; and Adit Ventures III LLC (the General Partners), for allegedly defrauding investors and client funds in connection with investments in pre-IPO shares, such as SpaceX and Klarna, including by misappropriating advisory client assets and charging millions in undisclosed fees.

According to the SEC’s complaint, from at least April 2019 through December 2024, the defendants used false claims and promises to persuade investors to contribute capital to Adit-managed funds, including Munson soliciting an investor by falsely claiming that a fund owned shares of stock of a private, pre-IPO company. As alleged, the defendants regularly used client capital for their own benefit, including by taking unsecured loans from funds on favorable terms, and these transactions were not authorized by fund documents and generally not disclosed to investors.

A simple fraud would be (1) claiming to own stock in a pre-IPO company, (2) taking investors’ money for it and (3) never actually buying the stock. But that’s not what happened. (I mean, it happened a little.^[1]) Instead, it was more like (1) taking investors’ money, (2) actually buying the stock with the money and (3) re-selling the stock to investors at a higher price. From the SEC’s complaint:

Rather than using invested Fund capital to buy shares of pre-IPO companies for the Fund directly, Defendant General Partners regularly took loans from Funds to buy shares of pre-IPO company stock and then resold those shares to their client Funds at a higher price. …

Here is an example of how this worked. In the summer of 2021, Munson had the opportunity to buy an interest in a third-party fund holding shares of SpaceX at a price of around $420 per share of SpaceX. In July 2021, Fund Astra Holdings SPV, LP acquired an interest equivalent to around 13,100 SpaceX shares for $420 per share. Part of that purchase was made with $1.2 million borrowed from Fund Ethos Holdings SPV, LP on June 30, 2021. ...

Acting through the Fund’s General Partner Adit Ventures I, Munson caused Astra Holdings SPV, LP, which owned interest in the 13,100 SpaceX shares, to assign its interest in those SpaceX shares to the Fund’s General Partner, Adit Ventures I, at cost—$420 per share.

By the end of July 2021, General Partner Adit Ventures I had sold an interest equal to those same 13,100 SpaceX shares to a new client Fund, Astra Holdings SPV III, LP, for around $498.00 per share—netting a profit of $78 per share. Adit Ventures I kept the resulting profit of around $1,020,000. …

Defendants misled the Funds and investors by claiming that the General Partners would not receive payments unless authorized by the Fund Agreements when they were actually obtaining undisclosed payments from client Funds through hidden markups. And, as explained below, Defendants misled their client Funds and investors by reporting a misleading “Original” price without disclosing that they were making a profit from these markups.

See, if you buy SpaceX stock with your own money for $420, and then you find some investors who are willing to pay $498 for it, and you sell it to them for $498, that feels, you know, fine, willing buyers willing sellers, whatever. But if you find some investors, borrow money from them to buy SpaceX stock for $420 for your own account, and then turn around and sell the stock to the investors for $498, that does seem like fraud. But as pre-IPO SpaceX SPV fraud goes, it’s maybe not that bad. At least they actually bought the shares.

The way financial innovation works is that you have a rich sophisticated client who will pay a lot of money for cutting-edge stuff, so you spend millions of dollars building her some cutting-edge stuff and you charge her a big markup. Then, having advanced the state of the art for cutting-edge stuff, you go out to a few other rich sophisticated clients and replicate it. “We just did a cutting-edge trade for Ms. X, and if it’s good enough for her it’s good enough for you,” you tell them, and you do a few more similar trades. These are much cheaper for you to do than the first one, because you get to reuse a lot of your work, but they still have high profit margins. Then your competitors get wind of the thing, and they start doing similar stuff for their rich sophisticated clients, which drives down margins.

Eventually the stuff becomes standard and automated, and the costs keep coming down. Someone realizes that they can do the formerly sophisticated cutting-edge stuff for ordinary retail clients. The marketing writes itself: “The cutting-edge stuff that was formerly accessible only to sophisticated billionaires is now available to you,” etc. And then eventually it’s on TikTok. Bloomberg’s Charlie Wells, Denitsa Tsekova and Isabelle Lee report:

For years, tax-optimizing investing strategies were confined to hedge funds and ultrawealthy families with access to specialized financial planning.

Now, they’re being pitched to the masses on YouTube and TikTok.

“There are maneuvers the wealthy have used for years to lower their tax bills that now you can access, too,” says one executive on YouTube while strolling through a leafy neighborhood. “Listen carefully in case you’re sick of paying high taxes,” starts a podcast-style interview. “Owning individual securities equals freedom to harvest losses,” explains another adviser on Instagram.

As with stock trading, alternative assets and financial advice itself, tax-focused investing is moving from the upper echelons of the American economy into the mainstream. Wealth advisers and financial influencers are peddling once-obscure strategies such as tax-aware long-short investing on social media and in marketing materials, aiming to capture business from a broader range of clientele.

Of course? We talked about tax-aware long-short investing last week. One fun fact about tax-aware long-short was that it was apparently invented at AQR Capital Management “just selfishly for the AQR partners co-invested in” AQR’s funds for tax-exempt institutions: That is, it was originally a product designed by hedge fund managers for hedge fund managers. Another fact about tax-aware long-short is that it is pretty quant-y and automatable, so there are obvious economies of scale. As I wrote last week:

If finance is getting more efficient over time, then one way — perhaps increasingly the main way? — that would manifest is in the form of investors paying no taxes.

Right now this stuff is expensive, as Wells, Tsekova and Lee write:

While tax alpha has become a “core client priority,” for smaller portfolios the strategies can ultimately offer little edge over tax-deferred vehicles such as IRAs or 401(k)s, said Jeffrey Janson, an adviser with Fiduciary Financial Advisors.

“Implementation costs can offset the benefits, and there's always the potential for tracking error or over-optimization that introduces unintended risks,” he said. “Sometimes the effort isn’t worth the reward.”

But eventually perhaps it will be efficient enough that no one has to pay capital gains taxes.

Intel Raises $20 Billion in Upsized Share Sale for AI Plans. Anthropic Tries to Shore Up Investor Confidence Ahead of Blockbuster IPO. AI-Dominated Leveraged ETFs Are Rattling Markets. Anthropic Strikes $9 Billion Cloud Deal With Riot Platforms. OpenAI’s head of ethics leaves start-up less than a year after joining. SEC Exempts Data-Center Bonds From Key Securitization Rules. Hong Kong set to include trading firms in ‘big bang’ tax reforms. Apollo Inks $2.6 Billion Financing Deal With NY Yankees. Trump Media Losses Hit $238 Million as Firm Pivots to Fast Feeds. Domino’s Set to Release New Individual-Size Pizza for Diners Who Don’t Want to Share. ‘Guac Signal’ Sparked Chipotle’s Frantic Bid to Recall Jalapeños. Italy’s $4.7 billion cheese economy is feeling the heat as climate change threatens its cheese banks that hold Parmigiano wheels as loan collateral.

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!