One niche financial service is getting rid of margin requirements. There are many financial products that require mark-to-market collateral. You enter into a foreign exchange forward contract in which you agree to pay me 100 euros in three months in exchange for 100 US dollars. If tomorrow the euro appreciates, so it would take $1.10 to buy €1, you will have a mark-to-market loss on this trade: You have agreed to pay me €100 for something that, today, is worth only €90.91. You have a loss of almost €9. You might decide not to pay me — after all, you can get $110 for your €100 elsewhere, so why pay me €100 for $100? — or you might be unable to pay me, if you have done a lot of these trades and they have moved against you and you’re bankrupt. And so it is traditional in many contexts for me to require collateral from you: If the trade moves against you by €9, you have to post €9 to a collateral account so I can make sure you’re good for it. (Depending on the type of product and our bargaining power, if the trade moves in your favor by €9, I might have to post €9 of collateral to you.)

You might find that annoying for various reasons. One is that it is just administratively annoying to have to have money move back and forth each day as exchange rates move. Another reason is that you have other uses for your money, and having to post some of it to me as collateral is costly. A third reason is that you might be a big levered bettor on foreign exchange rates, and if those rates all move against you at once you will face big margin calls and have to come up with a lot of money you don’t have. So it would be convenient, for you, to have a product that didn’t require collateral. You pay me €100 in three months, I give you $100 in three months, and whatever happens in between is none of our business. No collateral, just a contract that pays off in three months.

There is demand for that sort of thing, so people supply it. How do they supply it? You could imagine various answers, but here’s a simple one:

1. You come to me for a foreign exchange forward, and I give it to you — I agree to pay you $100 for €100 in three months — without collateral.
2. I turn around and go to a big bank and enter into an offsetting forward, in which I agree to pay the bank €100 for $100 in three months.
3. The bank does require collateral from me.
4. I post collateral to the bank, and don’t take any from you.
5. I charge you a fee for doing this transformation.

That’s a good service! It has a … very obvious problem? Most of the time, this service looks something like me renting my balance sheet to you and making essentially credit decisions. If you are creditworthy — if at the end of the three months you will almost certainly give me the €100 — then this all works out fine. My exchange-rate risk is hedged: I give you $100 for €100, the bank gives me €100 for €100, and net nothing happens. All I am doing is dealing with the cash collateral for you, but at the end of the trade that goes away; posting the cash collateral costs me something, but I charge you a fee to more than cover that cost.

But then some of the time the euro appreciates a lot against the dollar, all my clients are doing this trade, and I get huge margin calls from my banks and run out of money.

Is this the first blow-up of the Trump tariffs? Bloomberg’s Donal Griffin, Greg Ritchie and Meg Short reported Friday about Argentex Group Plc., a London foreign exchange brokerage:

The company announced Friday it had agreed to sell itself to rival IFX (UK) Ltd. for about 2.5 pence per share, 94% below where the stock traded a week ago, while Chief Executive Officer Jim Ormonde stepped down with immediate effect. ...

Argentex entered into derivatives agreements with banks including London-based Barclays and New York-based Citigroup, agreeing to post collateral — or margin — for its trades. This is a buffer against potential losses that banks can seize and sell if bets begin to sour.

Typically, a FX brokerage would then demand similar levels of margin from its own clients to protect itself. Yet Argentex pursued zero-zero lines with some of its corporate customers, declining to ask them for either an initial deposit of margin or so-called variation margin in response to how the trades fared, the people said.

Zero-zero lines — a reference to both the zero initial margin at the start of a trade and then the zero variation margin throughout its lifetime — have been in use by London’s FX brokerages since at least the early 2010s, according to one senior market participant, who requested anonymity. The strategy has proliferated in recent years, led in part by Argentex, the person said.

The bulk of FX brokerages in the UK now offer zero-zero lines to their corporate clients, said Daniel Kinnear, a veteran of the currency-sales business who now runs EAKO Capital in London. This allows them to compete against rival brokerages as well as the banks and financial-technology firms that make up the cutthroat industry, he said.

Yet zero-zero lines can cause a potentially fatal mismatch. While brokers like Argentex are on the hook for demands for margin from their lenders, they cannot issue offsetting margin calls to their clients.

“The key risks relate to the funding gap that this can create between the credit extended to the corporate clients and the trading lines that the FX brokerage has with its own liquidity providers,” said Kinnear, who previously worked at Wall Street banks including Deutsche Bank AG, Barclays and Nomura Holdings Inc. “FX brokers need to ensure that they have access to sufficient liquidity — either through their own equity or debt facilities — to fund any margin calls.”

Now, there are a lot of specific things that could go wrong here, but schematically, at a high level, those trades really should work out. If Argentex hedged its client trades with banks, and if the clients mostly didn’t blow up, then everything is in some approximate sense fine: The client trades will eventually settle, Argentex will settle with the banks, the results will offset and Argentex will have earned its normal profits. There is a liquidity mismatch — Argentex needs money now, and will get the client money later — but no essential economic problem.

Still, in volatile markets, there is perhaps nothing worse you can say than “it’s just a liquidity problem.” When you get a bunch of margin calls and don’t have enough money to meet them right now, but you’re definitely good for the money in the long run, the natural thing to do is to go to investors and say “look, our business is economically valuable but we need cash right now, could you lend us the money.” And the natural thing for them to say is “sure, we’ll lend you the money to meet your margin calls, but in exchange we want 100% of the business and your shareholders get nothing.” (“Give you a Snickers bar in exchange for 100% of the equity,” is how I characterized it, when FTX was in approximately this position.) And then you say “no no no you don’t understand, there is no solvency problem at all, our trades are fine, it’s just a liquidity mismatch.” But (1) that is rarely true and (2) even if it is, no one will believe you.

(Here is one exception. In early 2021, Robinhood Markets Inc. was in approximately this situation: It had a lot of clients doing meme-stock trades, the stock clearinghouse demanded more collateral, Robinhood didn’t have the cash and couldn’t take it from clients, so it had to go out and raise billions of dollars of equity to meet its collateral calls. That equity was somewhat expensive, but not “100% of the company for a Snickers bar” expensive, because it was obviously true that Robinhood had a pure liquidity problem and that the meme-stock frenzy would be good for business.) 

Anyway here is Argentex’s announcement of the IFX deal, which includes an immediate “secured bridging loan of £6.5 million.” The announcement suggests that Argentex did its best to get more than a Snickers bar for shareholders, but nobody — not even existing shareholders — was all that interested:

In recent days the Argentex Board began exploring potential options to secure additional funding and entered into discussions with a number of parties regarding expression of interest in Argentex. The Board also engaged with certain of Argentex’s leading institutional shareholders to explore the possibility for an urgent fundraising but concluded that this option was not viable in the circumstances.

On 22 April 2025, Argentex announced that it had experienced a rapid and significant impact on its near term liquidity position, driven by, inter alia, margin calls linked to its FX forward and options books. Argentex stated that it had taken a number of steps to preserve cash and increase the collateral received from its counterparties, but that if material volatility in currency markets persisted then Argentex's financial liquidity position, if not strengthened in the near term, would be significantly stretched. …

Argentex also announced a further deterioration in its liquidity position on 23 April 2025 and that it needed to secure immediate financing for Argentex to provide it with immediate working capital flexibility to assist with its near-term liquidity needs. …

Prior to agreeing the terms of the Acquisition, Argentex evaluated other expressions of interest in Argentex in order to ensure that the position of Argentex Shareholders was optimised based on the options available to it and concluded that the Acquisition was the most attractive for Argentex Shareholders.

Yes right if you are in the business of shielding customers from margin calls, it is pretty natural that you will get blown up by margin calls.

One of my most controversial claims in this column is that it is easy to keep a list of who has how many dollars. I wrote in 2019:

If you have U.S. dollars in a bank account at JPMorgan Chase & Co., and I have U.S. dollars in a bank account at JPMorgan Chase & Co., and I want to send you 100 of my dollars, what we do is I tell JPMorgan to subtract 100 from the number of dollars in my bank account and add 100 to the number of dollars in your bank account. This gets dressed up in a lot of procedures, because it would be bad if JPMorgan got the math wrong or if it moved money from one account to another without getting the proper authorizations, but as a matter of, like, computer science, it is dead easy. JPMorgan keeps a list of people and how many dollars they have, and you and I both trust JPMorgan to keep that list (that’s what it means that we bank there!), and so we just tell JPMorgan to update the list to reflect the transaction between us.

And then some software people objected that in fact this is very hard, that keeping a list of dollars synchronized among many computers for many purposes is a difficult problem of computer science, that my assumption that this should all work seamlessly and simply ignores the hard work of thousands of dedicated software engineers who come to work every day to make banking look straightforward, that blockchain solves this. I tweeted that “every problem that seems dumb and easy to people outside a discipline seems incredibly deep to the people inside it, may be the lesson here,” and also that “(1) there can be deep and difficult issues involved in reliably maintaining and modifying data at large scale and (2) I’m entitled to assume that there aren’t.” One should, in thinking about banking, be able to abstract away from the computational difficulties of maintaining a big ledger. There were banks for hundreds of years before computers were invented! Somehow they kept lists.

I think about this a lot, especially when a bank messes up its ledgers. Fine fine fine, it is hard.

A closely related problem is maintaining a company’s payroll records. How hard is that? Would you believe: impossible? I assert that at some level obviously you should not believe that: Companies kept track of how much they had to pay their workers for many decades before the invention of B2B SaaS payroll startups. On the other hand, can you find someone at a payroll startup who will talk your ear off for an hour about how managing payroll is actually the hardest problem in computer science? Sure, probably. 

Anyway we talked last month about Rippling’s lawsuit against Deel. People Center Inc. (doing business as Rippling) is a payroll startup, and it sued Deel, another payroll startup, alleging that Deel “cultivated a spy to systematically steal its competitor’s most sensitive business information and trade secrets.” Rippling caught the alleged spy, who then hid in a bathroom, tried to flush his phone down the toilet, and fled. Good stuff.

Last Thursday, Deel countersued Rippling. Here is the countersuit, which amusingly does not deny any of the spy stuff. Instead it is stuff like this:

Haunted by his previous failures, and now fueled by suffocating jealousy at his inability to fairly compete with Deel in the marketplace, Rippling’s co-founder and CEO, Parker Conrad—who was investigated and penalized by the SEC, and exiled from his own former company, Zenefits, for flouting the law—has fallen back on his old playbook: cheating. 

And:

Conrad grew up on New York City’s Upper East Side, where he earned “mediocre” grades at an expensive and prestigious all-boys preparatory school. Despite his mediocrity, Conrad was then admitted to Harvard University.

Predictably, Conrad then failed out of Harvard, which he described as an “incredibly humiliating and shocking experience”—because he apparently did not attend any classes “for like a year.” Despite Conrad’s forced leave of absence from Harvard, he was eventually readmitted, once again showcasing his privilege.

And even claims of a counterspy:

And incredibly—given Rippling’s recent “spying” allegations—Deel has information to believe that Rippling’s strange obsession with Deel had led Rippling to place an insider at Deel, essentially allowing it to eavesdrop on Deel’s internal communications without Deel’s permission.

But my favorite bit is this claim about the impossibility of payroll processing:

By way of background, building any kind of payroll engine is an extremely difficult task, and some of the more established companies still use the same historic mainframe they have always used to process payroll, and build their own manufacturing plants to make the parts to keep it running. Alternatively, smaller local companies have their own engines just for local payroll processing.

To date, however, on information and belief no one has been able to build a large-scale payroll engine to process payroll on a global scale. Indeed, these likely cannot actually be built without significant advances in quantum computing.

However, Rippling never lets facts get in the way of a good scam. On information and belief, Rippling has been falsely representing to its current and prospective customers that Rippling is running its payroll services on its own platform. 

I would not take any of the claims in these lawsuits as facts; these guys really seem to hate each other! Still. I love that if you meet with the chief financial officer of a normal company and ask “are you able to keep track of all of your employees globally and pay them?” she will say “yes,” and if you ask “is that the hardest problem that you face as CFO” she will say “no we just have a payroll department it’s fine,” and if you ask “do you have your own manufacturing plant to make the parts to keep your payroll system running” she will say “what, no, it’s just software, why would we have a manufacturing plant, I don’t even understand what you are talking about.” But if you meet with the head of a payroll software company and ask him “are you able to keep track of all of your customers’ employees globally so they can pay them” he will be like “man that is the great unsolved problem of computer science, we are pretty sure that it violates the laws of classical physics but we have our best scientists working on it and with a few billion dollars of infrastructure investment we’re hoping to crack it by 2050, but certainly our competitors will never figure it out.”

There are, what, four kinds of crypto projects?

1. Serious projects designed to build useful stuff by selling tokens that are securities under US law;
2. Serious projects designed to build useful stuff by selling tokens that are not securities under US law;
3. Jokes;
4. Frauds.

What is the legal status of each of these? Well, not legal advice, but:

1. Selling securities tokens to build useful stuff was, until a few months ago, essentially impossible in the US. The Securities and Exchange Commission, under Gary Gensler, demanded that crypto projects strictly comply with securities law, and that is hard for crypto projects to do. Also Gensler’s SEC took a maximalist — though, to my eye, largely correct — view of which tokens were securities. (Almost all of them.) Under the Trump administration, though, this has changed. It’s not crystal clear how it has changed — a crypto task force is working on it — but at the very least the new SEC (1) thinks many fewer tokens are securities and (2) wants to work with crypto issuers to make it more plausible for those tokens that are securities to comply with the law.
2. Selling non-security tokens to build useful stuff is fine. A few months ago this meant mostly Bitcoin; now it is a bigger category.
3. Selling joke tokens to do nothing is also fine? Memecoins, the new SEC has explicitly ruled, are not securities, but I think even under the old SEC regime it was hard to argue that they were securities. (If you buy a memecoin you are not investing in a business, so it’s not a security.) This is a weird situation — the law favors useless crypto projects over useful ones — but both securities law and crypto are pretty weird so, you know, fine.
4. Selling fraudulent tokens was clearly illegal under the old regime: Fraud is illegal, and the SEC took the position that pretty much all crypto fraud is securities fraud, and brought enforcement cases. [1] Under the new regime, though, the analysis is harder. For one thing, if most crypto tokens are not securities, then most token fraud is not securities fraud; even if crypto fraud is illegal, the SEC can’t bring a case. (And who else would?) For another thing, uh, look around? There are a lot of Trump-related crypto projects. Are they … you know what, never mind, I don’t know where I was going with this paragraph, forget it.

I guess it was possible that the new SEC would not bring any crypto fraud enforcement actions: Crypto fraud might no longer be the SEC’s problem. (I guess it was possible that the new SEC would not bring any fraud enforcement actions at all; I wrote in January: “What if the US Securities and Exchange Commission never brings another enforcement case? What if it’s like ‘ahh, go ahead, do all the fraud you want’?”)

But in fact the new SEC has disclaimed this analysis: Despite concluding that crypto tokens might not be securities, it has nonetheless expressed interest in policing them for (securities) fraud. Somebody has to, so why not the SEC?

And so last week the SEC brought a very garden-variety fraud case against some crypto guy:

The Securities and Exchange Commission [Tuesday] charged Ramil Palafox for orchestrating a fraudulent scheme that raised approximately $198 million from investors worldwide and for misappropriating more than $57 million of investor funds.

According to the SEC’s complaint, Palafox’s company, known as PGI Global, claimed to be a crypto asset and foreign exchange trading company. From January 2020 through October 2021, Palafox offered and sold PGI Global “membership” packages, which he claimed guaranteed investors high returns from PGI Global’s supposed crypto asset and foreign exchange trading and offered members multi-level-marketing-like referral incentives to encourage them to recruit new investors. However, as the complaint alleges, Palafox misappropriated more than $57 million in investor funds to buy Lamborghinis, items from luxury retailers, and for other personal expenses. He also used the majority of the remaining investor funds to pay other investors their purported returns and referral rewards in a Ponzi-like scheme until its collapse in late 2021. …

“Palafox used the guise of innovation to lure investors into lining his pockets with millions of dollars while leaving many victims empty-handed,” said Laura D’Allaird, Chief of the Commission’s new Cyber and Emerging Technologies Unit. “In reality, his false claims of crypto industry expertise and a supposed AI-powered auto-trading platform were just masking an international securities fraud.”

Yeah no of course, pretend to do AI-powered crypto trading and actually steal all the money for Ponzi and Lambos, a tale as old as time, or at least as old as 2017. But it’s a little interesting that the current SEC crypto regime signed off on a complaint saying this:

PGI Global’s membership packages were investment contracts and thus constituted securities within the meaning of Section 2(a)(1) of the Securities Act and Section 3(a)(10) of the Exchange Act. As investment contracts, PGI Global’s membership packages involved: (1) the investment of money; (2) in a common enterprise; (3) with a reasonable expectation of profits to be derived from the efforts of others. See SEC v. W.J. Howey Co., 328 U.S. 298-99 (1946).

This is an easier case under Howey than some of the Gensler enforcement actions, but it is still a statement from the current SEC that fraudulently managing people’s crypto investments is securities fraud, even if the crypto investments are not necessarily securities.

The timeline is something like:

• In April 2022, Elon Musk decided to buy Twitter Inc., and quickly signed a deal to buy it for about $44 billion.
• Subsequently, the prices of social media companies fell, Twitter in particular looked overpriced, and Musk regretted his deal and tried to back out.
• Nonetheless the deal closed in October 2022, and Musk promptly renamed Twitter “X.”
• One month later, in November 2022, OpenAI launched ChatGPT and the modern artificial intelligence boom got started in earnest.

If ChatGPT had launched a year earlier, it would not have been obvious that Twitter was an AI company. I mean, it wasn’t an AI company at all. But it had some nice features for an AI company: It had data (lots of tweets with real-time information and jokes and natural language), and it had distribution (lots of people typed in the box on Twitter’s website or app every day, making them a ready audience for an AI chatbot in that box). It just needed to buy a lot of processing power and hire and inspire and motivate a bunch of top AI researchers and put them to work. Twitter’s existing managers in April 2022 were probably not the people to do that, but Elon Musk arguably was. So a thesis like “Elon Musk should acquire Twitter, buy a lot of chips, hire a lot of AI researchers, and put them to work on a hard-core mission of turning Twitter into a leading AI lab with a 12-digit valuation” would have sounded a little crazy, but not that crazy, and turned out to be basically correct.

Except, again, the whole “become a leading AI lab with a 12-digit valuation” concept didn’t really exist in October 2022. That was more of a November thing.

Anyway you do gotta hand it to him:

Elon Musk’s XAI Holdings is in talks with investors to raise roughly $20 billion in funding for his newly combined artificial intelligence startup and social media business, according to people familiar with the matter.

If completed, the prospective deal would be the second-largest startup funding round of all time, according to data provider PitchBook, trailing only OpenAI’s $40 billion financing earlier this year. The transaction would value the company at more than $120 billion, said one of the people, all of whom asked not to be identified because the information is private.

XAI didn’t immediately respond to requests for comment.

Announced in March, XAI Holdings was created from the combination of X, formerly Twitter, and Musk’s AI venture, xAI. The new funding could be used to pay down some of the debt that Musk took on when he converted Twitter into a private company, later renaming it X, one person said.

That debt has been weighing on X, Bloomberg previously reported. In March alone, X paid about $200 million in debt-servicing costs related to its buyout, according to people briefed on the matter. The firm’s annual interest expense by the end of 2024 was more than $1.3 billion.

We have talked about that debt a few times: For a while it seemed to be worth about 50 cents on the dollar, and I have wondered about the thesis that made investors want to buy it. It turns out the correct thesis was “Twitter is an AI company, and AI companies can raise unlimited billions of dollars of equity right now, and those equity investors will pay off the debt.” 

Here is a New York Times story that begins with a middle-aged suburban couple being kidnapped from their $240,000 Lamborghini Urus, and ends by at least strongly implying that the Lamborghini was bought by their 18-year-old son with the proceeds of a $243 million Bitcoin theft, and that the kidnappers were after that Bitcoin. 

There is a lot of stuff here. The son “was secretly a member of the Com, also referred to as the Comm or the Community, an online network of chat groups that has its roots in the hacking underground of the 1980s and functions as a kind of social network for cybercriminals or aspiring ones”? The $243 million was stolen from an early Bitcoin investor by calling him up, pretending to be a crypto exchange representative, and asking him to “download a program that would provide additional security,” which promptly stole all his crypto?

But I do think that the main lesson here — and nothing in this newsletter is either legal or parenting advice — is that, if your teenaged son is really into Minecraft and one day he comes home from school in a Lamborghini that he just bought, you need to sit him down and have a serious conversation with him about the risks of stealing a lot of Bitcoin. Those risks are serious — not just legal risks but physical violence — and he has definitely stolen some Bitcoin!

Trump’s China Tariffs Set to Unleash Supply Shock on US Economy. Demand slump fuelled by Trump tariffs hits US ports and air freight. America Inc. Slashes Spending as Tariff Uncertainty Swirls. Shein Hikes US Prices as Much as 377% Ahead of Tariff Increases. Goldman Sachs Offers Advice on Tariffs to Countries Scrambling to Please Trump. Global Race to