You could imagine this being big news: Elon Musk said his xAI artificial intelligence startup has acquired the X platform, which he also controls, at a valuation of $33 billion, marking a surprise twist for the social network formerly known as Twitter. “The combination values xAI at $80 billion and X at $33 billion,” the billionaire wrote Friday in a post on X. The value of X is $45 billion when including $12 billion of debt, he said, describing the purchase as an all-stock transaction. The deal gives the new combined entity, called XAI Holdings, a value of more than $100 billion, not including the debt, according to a person familiar with the arrangement, who asked not to be identified because the terms weren’t public. Morgan Stanley was the sole banker on the deal, representing both sides, other people said. For Musk, the deal streamlines his businesses and solidifies the relationship between the former Twitter and xAI, which has used information from the social network to hone its chatbot. The deal also offers a resolution to X’s other backers following months of uncertainty over the state of their investment as Musk’s changes led to an exodus of users and advertisers. I mean! - This is the biggest mergers-and-acquisitions deal in the world so far in 2025, which has so far been a rough year for M&A. [1]
- There is an obvious symbiosis between modern artificial intelligence companies, with their demand for real-world training data, and social media companies, with their consumer distribution and huge corpora of natural-language posts. [2] So we’ve seen social media companies striking licensing deals with AI companies, and of course Meta Platforms Inc. is building its own AI models. But this seems to be the first large-scale acquisition of a big social platform by a big AI competitor. [3] You could imagine it setting a precedent, launching a rush of acquisitions, being a valuation benchmark, that sort of thing.
- Elon Musk is a famous guy who has done a lot of weird M&A stuff, and this is weird Elon Musk M&A stuff.
- In particular, there is an obvious conflict of interest: Musk owns most of X, he owns most of xAI, there is a lot of overlap among the other shareholders, but the overlap is not total. There is surely some xAI shareholder who does not own any X, and some X shareholder who does not own any xAI. [4] If you are, say, an xAI shareholder who thinks that xAI is a promising world-changing company and X is a broken social media company, you might be aggrieved to see the company you own wasting $45 billion to bail out another company just because Musk owns it. You might be annoyed that Musk was on both sides of the deal, that he controls both companies, that he used the same bankers on both sides, that no one negotiated independently on behalf of xAI’s shareholders. (Or X’s. William Cohan asks: “Will Linda Yaccarino be the C.E.O. of the combined entity? Did she even know it was happening?”) You might draw parallels to the time when Tesla Inc. bought SolarCity Corp. and Tesla shareholders sued, objecting that Musk was using Tesla’s money to bail out his investment in SolarCity. They lost that lawsuit but at least they tried.
- If you’re an X shareholder you are surely elated, but I suppose you are theoretically allowed to be aggrieved that your company took the first deal that Musk offered. What if X is really worth more than $45 billion? Cohan asks: “Did anyone know X was for sale? Was the company shopped around before deciding to sell out to xAI?” Absolutely not. Imagine!
- Just as like an M&A mechanics matter, what exactly does it mean that Musk tweeted “xAI has acquired X in an all-stock transaction”? The deal signed? It closed? Often when two companies agree to a $45 billion merger, there is some delay between signing and closing. The typical main causes of delay — getting a target shareholder vote, lining up financing — are probably not relevant here, because (1) Musk controls the vote at both companies [5] and (2) it’s an all-stock deal so it doesn’t require any money. [6] Still you might expect some time between the announcement of the deal and closing. Big acquisitions normally have to wait for antitrust clearance from US and foreign regulators, for one thing.
But my heart isn’t in any of this. Nobody cares. Musk has absolute control of (1) xAI, (2) X and (3) US government regulators. If he wants to smush X and xAI together, no one will complain, and it doesn’t mean anything. Surely Musk isn’t required to file forms or get regulatory approvals anymore. He is not required to abide by merger best practices or zealously protect the interests of minority shareholders, in part because all his companies have left Delaware, but in larger part because none of the shareholders of his private companies complain about anything he does. If you are giving money to a private Musk venture, it’s because you trust Musk to make decisions, not because you care about corporate formalities. X/xAI is the largest M&A deal so far in 2025 in the sense that it values X at, you know, 33/80 of the value of xAI, and Musk announced that xAI is worth $80 billion, and all of that is reasonably plausible though not clearly validated by arm’s-length transactions with economically motivated counterparties. xAI last raised money in December at a $51 billion valuation; it was reportedly looking to raise more at a $75 billion valuation last month. X raised money at a $33 billion equity valuation earlier this month, but (1) some of that money came from Musk and (2) I wonder if this deal was already in the works at the time? In hindsight, X’s funding round looks a little like it might have been designed to validate a merger price. In any case, it’s barely an M&A deal? Two companies that were owned by the same person (and some slightly non-overlapping friends) and shared employees and data and revenue and, you know, a name, are now one company. They were informally one company before, and they are formally one company now, and no money changed hands. It feels like a silly technicality to call this a big M&A deal. It is perhaps more interesting that they were ever separate. Elon Musk bought Twitter (and renamed it X) with a plan to do some AI to it. He then spun up a separate company called xAI, raising new money from investors to fund it but promising to give some of it to X. The companies were closely linked. Why start a new one? Why not just do the AI stuff within X? I don’t know, but when Musk closed the Twitter deal, Twitter looked to be a bit of a shambles. He had just fought a lawsuit calling it a big fraud, advertisers were fleeing, he was getting rid of all the employees and his bankers were having a tough time selling the debt he used to finance it. It was possibly not the most auspicious vehicle for ambitious AI hiring and fundraising. Just start a new one, free of the Twitter baggage. Investors and employees wanted AI, so put AI right in the name. It is sort of remarkable that an AI startup that Musk spun out of the damaged husk of Twitter — two years ago! — has gotten big enough to come back and be the acquirer in the latest Musk reshuffling. Those are, I suppose, the two critical elements of the Musk Mars Conglomerate: - Every new idea gets its own company: Each new Musk company can present a clean investment case to investors, [7] can incentivize employees with direct ownership of their work, and can give Musk full control and upside in the project.
- Nothing is permanent, and if one of the companies succeeds while another one has some temporary struggles, the winner can subsidize — or buy — the loser.
People are worried about FX liquidity | A theme in modern financial markets is the move from human trading to electronic trading. In the olden days, if you wanted to buy or sell a bond, you’d pick up a telephone and call a trader at a securities dealer (generally a big bank). [8] She would give you a price, you would trade with her, and you’d pay her a markup for doing the trade. In exchange for the markup, she would (1) try to fill all your orders at competitive prices and (2) take you out to dinners and drinks and sporting events. Sometimes, the market would crash and you would be desperate to offload your bonds. When that happened, you could call her, and she might buy the bonds from you at a pretty good price, despite the panic all around her. Why would she do that? Various reasons: - She is a fearless trader at a powerful bank with a big balance sheet, she thinks the market will turn around, and she thinks this is an attractive entry point. She will buy your bonds now, while everyone is panicking, to bet on a recovery. She has the balance sheet and risk appetite and patience to do that. Traders do make a lot of their money in volatile times. She is giving you what you think is a good price for you, because she thinks it is a good price for her.
- You have been a good customer over the years, paying her lots of big markups, and she wants to give you good service now so that you continue paying her big markups in the future.
- You have been to a lot of sporting events together, she likes you personally, and it would be awkward for her to listen to you weeping and cursing over the phone if she said no. She buys your bonds out of some sense of personal obligation.
- She has some moral sense that the social role of a bond dealer is to provide liquidity in crisis situations, so she steps up when the market is crashing to preserve her bank’s reputation as a good citizen that helps its customers in tough times.
I think that I have listed those reasons in roughly declining order of importance, and I am not sure that the fourth one was ever a thing at all, but people do sometimes talk about it. This system had a lot of overhead: The trader gets paid a lot of money, the tickets to sporting events are expensive, the bank requires a large balance sheet to hold all the bonds that she buys and sells, regulation makes that balance sheet expensive, phone calls are a fairly slow way to trade, and every so often a bank trader would blow herself up by taking on too much risk. Over time, trading tends to get more electronic: There are electronic trading platforms where you can submit an order to buy or sell bonds, and anyone can take the other side of that trade. Sometimes the person on the other side of the trade will be the bank trader whom you would previously have called on the phone. But as electronic markets get bigger, it is easier for other firms to compete. These firms have electronic algorithms that decide what prices they will pay for bonds, so they save money on traders. Because trading is electronic, they can buy and sell more quickly, so they do not have to hold lots of bonds for a long time and can have smaller balance sheets. A handful of people with a fast computer can compete with banks with huge balance sheets and armies of traders. They don’t need to buy you dinner, because they trade electronically and often anonymously. They just have to give you the best price. Over time these electronic trading firms can often out-compete the big banks, so they can come to dominate a market and the banks retreat. And then sometimes the market crashes and you want to sell your bonds, so you send an order to an electronic platform. Will the electronic trading firms buy your bonds at a good price? Well, no, probably not: - They do not have huge balance sheets, they are rigorously rational, and they keep their costs low in part by not taking huge swings that might blow them up. If markets are erratic, they can just turn off their computers and stop trading.
- Their markups aren’t that big, they often trade anonymously, and they might not even have a concept of “good customer who pays big markups.” They don’t care about your loyalty.
- They’ve never gone to a sporting event with you and don’t talk to you on the phone, so if you weep and curse it isn’t their problem.
- Whether or not bank traders have some sense of themselves as liquidity providers in crises, electronic trading firms surely have even less of that sense.
This is called “liquidity illusion.” When times are good, electronic trading lets you trade faster and with lower costs than the old system of calling a bank on the phone. But when times are bad, you might miss the bank traders, who at least had to listen to you weep. This is a pretty generic description but you see versions of it in many markets. US stocks perhaps most famously, Treasury bonds, a little bit in corporate bonds, and now here is Bloomberg’s Alice Atkins on foreign exchange: The global arena for foreign exchange appears on the surface to be a giant pool of cash in constant churn, generating $7.5 trillion of transactions per day — more than all other major asset classes combined. But lately key names from the industry including the likes of Citigroup Inc., Deutsche Bank AG and XTX Markets Ltd. have begun to fret about just how deep the pool really is. They fear a proliferation of trading platforms and the widespread use of automation are creating the illusion of market depth, while a retreat of big institutions has actually drained liquidity away. The symptoms are subtle, including things like rising trade rejections, declining volumes at key venues, and volatility in the gaps between what buyers offer and sellers ask. But they all point to what’s known as liquidity mirage, a phenomenon that can ramp up risk for market participants without them even knowing. “Liquidity on the face of it might look quite robust,” says Mark Meredith, head of FX e-trading and algo execution at Citigroup. “But actually in extreme events it’s more and more fragile.” … Jeremy Smart, global head of distribution at XTX, one of the market’s top providers of liquidity, says moments like the carry-trade selloff highlight changes at the heart of the system. Volumes on the two primary inter-dealer markets, EBS and LSEG (formerly Refinitiv), have been in steady decline for years as banks move to offset more of their client flow internally, while regulation has also reduced the ability of many dealers to take on risk. That leaves non-bank market makers playing an increasingly critical role. In stress events, banks can’t offset their various client flows against each other, Smart says. Instead, they turn back to the primary market and non-bank market makers, only to find they can’t absorb all of the risk. “There’s a limited number of takers on the other side of the market,” he says. “That does mean that when you get very directional markets, you get over-sized moves.” What is the solution? Well, if you ask a bank, the solution is “pay more markups to the nice bank traders who take you out to sporting events”: To some firms the fragmentation, rising competition and resulting liquidity dangers simply underline the value of a more old-school approach. Banks including Deutsche Bank, UBS Group AG and SEB Corp. maintain that, despite the move toward e-trading, clients who come direct and honor fewer relationships will see better liquidity, particularly during times of stress. “More is not better,” says Benedict Carter, global head of FX e-trading at Deutsche. By working with trusted partners “you can lower your technology cost, get better liquidity and negotiate outcomes. You can make sure they are there for you when times get tough,” he says. I do wonder a little what the contractual “there for you when times get tough” provisions are. The banks are businesses too. To a certain sort of person, the promise of crypto was not some utopian vision of financial empowerment, but this: - An entrepreneur could invent some flimsy project, and a token to go along with it.
- Institutional investors (venture capitalists, hedge funds, crypto celebrities, etc.) could buy the token from the entrepreneur, lending their names and reputations to the project and making people think it was legitimate. They could also tweet about how world-changing the project was and how nice it was that everyday investors could get in on it.
- The token would go up as retail investors bought it.
- The institutional investors would sell their tokens to the retail investors.
- There was no disclosure: The entrepreneur didn’t have to disclose anything about the project, the institutional investors didn’t have to disclose that they were selling, and no one even needed to disclose if the entrepreneur was paying the institutional investors to promote the token, as sometimes happened.
- Maybe the project would change the world! Or maybe it would collapse and the retail investors would lose all their money, but the institutions, who bought cheap from the entrepreneurs and sold at pumped-up prices to retail, would make a profit either way.
That is, crypto was arguably an alternative way to do stock offerings with these distinctive features: - You didn’t need a real business, since nobody cared if the stock was worthless.
- You could sell the stock privately, and the buyers could immediately resell to the general public.
- Securities laws did not apply.
It always seemed to me, and until recently to the US Securities and Exchange Commission, that this analysis couldn’t really be right. Surely if you sell tokens promising economic interests in some project, those tokens are securities and you need to follow securities laws? Surely if you sell them to institutional investors who plan to turn around and resell immediately to retail, those institutions are “underwriters,” are liable for any misstatements about the platform, and have to disclose their involvement? And to be fair a lot of people in crypto seemed to agree: Many venture capitalists who invested in crypto projects did due diligence and held onto their tokens for months or years before reselling them, to avoid accusations like “you are just pumping this token and dumping it on retail investors.” Not Galaxy Digital Holdings though: Michael Novogratz’s Galaxy Digital Holdings will pay $200 million in penalties over the investment firm’s role in promoting the failed Luna cryptocurrency, as part of a settlement with the New York Attorney General. ... The settlement includes an undiscounted monetary penalty of $200 million, payable in instalments until 2028. Galaxy noted a legal provision of $166 million to cover the fine in its full-year results on Friday, noting the impact of discounting. The investment firm was accused of violating rules in promoting an asset without disclosing its intent to sell it, according to a filing published by the NYAG. Galaxy did not admit or deny wrongdoing as part of the deal, the filing said. Here is the filing. Basically, per the New York Attorney General, “Galaxy bought more than 18.5 million Luna from Terraform in October 2020 at $0.22 per token, a nearly 30% discount to Luna’s then-market price of $0.31 and began endorsing Luna.” Terraform gave Galaxy the discount because it wanted Galaxy to promote its token. Novogratz tweeted about it regularly, and Galaxy falsely “marketed Terra and Luna’s purported use on the Chai payment platform, stating that Luna provided a real-world use case for crypto.” (Terraform told Galaxy that Luna was widely used on the Chai payment platform in Korea, but that was false and “Galaxy did not take sufficient steps to determine its veracity.”) Novogratz got (and tweeted) a Luna tattoo when the token hit $100. Meanwhile Galaxy was getting monthly releases of Luna tokens and selling them on without telling anyone; it ended up making more than $100 million by selling most of the Luna tokens it had pumped. Then Luna collapsed to essentially zero. “Do Kwon and Terraform, the creators of Luna, deceived us and many other prominent institutional investors,” Novogratz said in a statement, which seems true, but of course being deceived by Terraform was very profitable for Galaxy. If it is profitable to be deceived, you might go looking for opportunities to be deceived; certainly you won’t do due diligence to make sure that the tokens you are pounding out to retail investors do what they say they do. The incentives are bad! This is just exactly, exactly, exactly what securities laws are supposed to regulate. Now they’re gone! Oh well. Elsewhere in “fraud is legal now” news, President Donald Trump commuted Carlos Watson’s prison sentence and pardoned Trevor Milton. Watson ran Ozy Media, which impersonated a customer on a due diligence call; Milton ran Nikola Corp., which rolled a truck down a hill to pretend it had an engine. One possibility is that if you do hilarious high-profile fraud, Donald Trump will find that amusing and will pardon you. There are alternate readings. Meanwhile “Charlie Javice was found guilty of defrauding JPMorgan Chase & Co. in its $175 million acquisition of her student-finance startup, Frank, following a six-week trial.” I thought Javice’s fraud was also pretty funny — she sold JPMorgan a fake email list for $175 million! — so there is hope for her; perhaps Trump just hasn’t gotten around to pardoning her yet. It is probably bad that politically connected people can do fraud with impunity but. You know. In like 1900 the way a stock exchange worked is that there was a building or a room or a tree where people met to trade stocks. If you were in New York, you met at a New York stock exchange, possibly though not necessarily the New York Stock Exchange. (You could have met at the American Stock Exchange, which was also in New York.) If you were in Philadelphia or Chicago or San Francisco, you met at the Philadelphia or Chicago or San Francisco stock exchanges. A big company’s stock might trade in more than one place, but if you were in Chicago you could not stroll onto the New York Stock Exchange floor to trade stock, because New York was far away. In 2025 the way a stock exchange works is that there is a computer in New Jersey and you send an electronic messages to that computer to buy or sell stocks. Other people also send their electronic messages to that computer, and you can trade with them. There are a bunch of different stock exchanges, and you might want to send your messages to different computers to trade different stocks; for a big trade, you might want to send messages to all of the computers. But the computers are in New Jersey. The reason for this is basically that stock trading is very fast, and you want to be able to quickly send messages back and forth between the different exchanges, so that you can monitor and profit from price differences on each exchange. An exchange located outside of New Jersey would be too far away from all the other exchanges; it would take too long for messages to travel back and forth. Obviously there are exchanges outside of New Jersey: There are important futures exchanges in Chicago, and there are stock exchanges in, like, Italy and Japan. If you trade financial products globally, you will send messages back and forth between those exchanges and the US stock exchanges in New Jersey. But the US stock exchanges are in New Jersey. For historical and marketing and product-differentiation reasons, some of the computers in New Jersey have geographically misleading names like “Nasdaq PHLX” (the PHL is for Philadelphia) or, for that matter, “New York Stock Exchange.” One was called “NYSE Chicago” until this weekend; now it is called “NYSE Texas.” The advantage of NYSE Texas is that, if you want to list your stock on a US stock exchange, but you want to signal that you are not a New York City coastal elite, listing on NYSE Texas allows you to put the word “Texas” in your press release. Guess who? President Donald Trump’s media venture is forging a symbolic link to Texas, becoming the first company to list on the New York Stock Exchange’s upstart outpost in the Lone Star State. Trump Media & Technology Group Corp. will list warrants tied to its shares on the new NYSE Texas while maintaining a primary listing on the Nasdaq. The move — along with plans to switch its corporate headquarters to Florida — is part of efforts to align the company with states that are friendly to conservative values, according to Chief Executive Officer Devin Nunes. “We’re part of a growing movement to take our business to states that value free enterprise and personal freedom,” Nunes said in a statement. Adding Trump Media to another exchange carries more symbolic weight than practical effect, with the difference unlikely to be noticed by most investors or users of the social media site. … The plan to start NYSE Texas was announced in February. It will be a fully electronic exchange. Unlike the New York Stock exchange, there is no physical trading floor in Texas as the systems and data center are located outside the state. Okay. Trump Media’s stock is primarily listed on Nasdaq (whose computers are in Carteret, New Jersey), but now it is also listed on NYSE Texas (Mahwah). Bloomberg has a new news quiz game called Pointed, which has some almost Learned League-style tactical gameplay that I enjoy. You can play it here. Rocket to Buy Mortgage Firm Mr. Cooper in $9.4 Billion Deal. Is UBS’s ‘deal of the century’ starting to sour? CK Hutchison Shares Fall as Li Ka-shing Mulls Ports Deal Delay. Hedge funds fight over one of Europe’s largest lingerie brands. Strategy Ends Quarter With Biggest Bitcoin Purchase of Year. The Trump Family Advances Its All-Out Crypto Blitz, This Time With Bitcoin Mining. In Crypto’s Darkest Corner, a Suicide Becomes Just Another Meme Coin.In Japan, an Iceless Lake and an Absent God Sound an Ancient Warning. 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! |