Money Stuff: Google Needs More Money

You could tell a stylized story of the stock market in three phases. Historically, the point of the stock market was to raise money for companies to do stuff. You wanted to start a railroad, you needed money to build tracks, so you went out and sold stock to investors to raise the money. People bought stock to fund companies, and when the companies did well, they were able to pay nice dividends to their investors.

The stock market of the 21st century is different, in at least three ways:

1. A lot of the biggest and most successful modern companies are not particularly capital-intensive. They’re, like, websites. You don’t need that much capital to build a social media site, or a ridesharing platform, or a software-as-a-service business. These are businesses with low marginal costs that can scale up without raising new capital.
2. To the extent they do need to raise a lot of money, they often do it early and in private markets. “Private markets are the new public markets,” I often say: There’s a lot of money in private markets, it is easier and more pleasant to stay private, and so companies tend to fund their growth in private markets before going public. 
3. Now companies tend to return money to investors by buying back stock rather than by paying dividends.

I wrote back in 2015 that “the public stock markets are increasingly about capital return rather than capital raising”: The main thing that public companies do in the stock market is buy back stock, not sell it. “Equity retirements have been consistently greater than issuances,” said the Federal Reserve back in 2017, and since then net equity issuance by US companies — the amount of stock they sell to raise money, minus the amount they buy back — has remained consistently negative. Meta Platforms Inc., which has a market capitalization of $1.5 trillion, has sold a total of about $11.2 billion of stock in its history; it bought back about $23.6 billion of stock last year. Building a website doesn’t cost very much money, but if it’s the right website it generates a lot of money.

But perhaps now we are entering into a third phase:

1. A lot of the biggest and most successful companies now are enormously capital-intensive. They are artificial-intelligence hyperscalers, and their business model is like “build nuclear power plants and orbital data centers and massive chip fabrication facilities.” After years in which the cutting edge of the economy was nearly zero-marginal-cost software, now the cutting edge of the economy is extremely capital-intensive, uh, software.
2. The private markets can provide many tens of billions of dollars to those companies, but there’s some limit.
3. Even if the successful companies redirect their huge profits from stock buybacks to capital expenditure, that still might not be enough to fund the AI buildout. 

In some ways it is a return to the old story, in which companies go to the stock market to raise money to fund their growth. The novelty is that now they’re doing that at trillion-dollar valuations.

We have talked a lot about SpaceX, the launch-rockets-into-space-to-power-enterprise-AI-software company, and its plans to do an initial public offering of perhaps $75 billion. That would be the largest IPO ever, and might kick off a wave of other similar AI mega-IPOs for Anthropic (which filed confidentially for an IPO this week) and OpenAI. But it’s not just the private companies. Bloomberg News reports:

Google parent Alphabet Inc. is raising $80 billion through a package of equity offerings, including an investment deal with Berkshire Hathaway Inc., as the company races to fund its ambitious artificial intelligence spending plans.

The undertaking includes a $40 billion so-called at-the-market program to sell shares from time to time beginning in the third quarter, according to a statement Monday. The company will also offer $30 billion in underwritten offerings of shares and mandatory convertible preferred stock, as well as the $10 billion deal with Berkshire.

Together, the transactions represent one of the largest equity deals of all time — and they bring an unexpected twist to a blockbuster year for initial public offerings.

It’s rare for a large public company to raise this much equity, but the economics of the AI business have pushed Google and its peers to get creative. The company has embarked on an unprecedented spending spree to build the infrastructure that it needs to develop cutting-edge artificial intelligence models and meet demand from customers who want to buy its chips to fulfill their own AI ambitions.

Google is trying to capitalize on a growing appetite for its homegrown AI chips, known as tensor processing units, or TPUs. They have become a key alternative to Nvidia Corp.’s market-leading processors in an industry that requires tremendous amounts of computing power.

“AI is driving an expansionary moment for Alphabet,” the company said in the statement. “By scaling its investments, the company seeks to expand its foundational infrastructure to support the significant growth opportunity ahead.”

Alphabet “bought back $62 billion of stock in 2024 and $46 billion in 2025,” notes Barrons. it had $133 billion of net income and $165 billion of operating cash flow in 2025, along with $91.4 billion of capital expenditures. This is a company that runs a giant lucrative business that can pay for a lot of data centers. But still not enough data centers.

I assume that somewhere at Alphabet there’s a finance employee who modeled out their need for $80 billion, but I like to think that the main reason for that number is to be slightly bigger than the (expected) SpaceX deal. “Biggest IPO ever” is a somewhat arbitrary accolade: We are in a new world, and in a sense everyone is starting over. A bunch of giant AI companies will do 11-digit equity raises this year; Alphabet and SpaceX are two of them. One is an IPO and one isn’t, but these are both giant companies racing to raise money to build all the data centers they can.

The big stock-market theme in the last few years has been the AI trade, as investors have tried to buy stock in the expected winners of the AI boom and rotate out of the expected losers. This has pushed up the valuations of the big AI companies, but in that abstract, disconnected, second-phase-of-the-stock-market way: It’s not like the companies are getting the money, directly, from their new valuations. But now Alphabet will.

Elsewhere in equity capital markets: 

Elon Musk’s SpaceX is negotiating to pay razor-thin fees to Wall Street firms handling its IPO — but banks are still likely to rake in about $500 million from the record-setting market debut.

Musk’s space and artificial-intelligence conglomerate is negotiating to pay less than 0.75% for the $75 billion it aims to drum up in an initial public offering this month, according to people with knowledge of the matter. Even at that low spread, it will likely amount to one of the biggest fee events ever for Wall Street firms that arrange public listings. …

The slim margin that banks are offering Musk has implications for a slate of blockbuster IPOs in the works for this year. That could force investors and analysts to temper expectations for Wall Street’s earnings as OpenAI and Anthropic PBC also lay the groundwork to tap equity markets in the months ahead.

Yeah, look, charging 3% for IPOs is fine when an IPO is a bespoke vanity project, but if there are six big companies raising $100 billion a year in public equity, the fees will have to come down.

Meanwhile in debt capital markets, the great financial innovation of the AI age is probably the put option on computer chips. It goes like this:

1. Building and deploying artificial intelligence is a huge infrastructure project; it requires building giant data centers filled with computer chips, and new electric power plants to run them.
2. Sure you can raise equity. But the traditional way to fund giant infrastructure projects involves a lot of debt financing: You fund a data center by borrowing money from insurance companies looking for steady cash flows.
3. Those lenders want steady cash flows.
4. The projects expect steady cash flows: The big AI companies will happily enter into giant multi-year leases for not-yet-built data centers, because computing power is one of the main constraints on their growth and they’ll take a much as they can get.
5. So the project can go to lenders and say “we have a 10-year lease with Anthropic that will pay us $10 billion per year, so we can easily cover your $9 billion a year in interest payments.” Good, good.
6. But this has just shifted the problem: Now the lenders are effectively lending the money to Anthropic, because they are relying on Anthropic to make the lease payments.
7. Equity investors love Anthropic, but lenders might worry. Anthropic “is still a startup and lacks the earnings to support such a large debt deal”; it doesn’t have the steady cash flows that will soothe the lenders. AI is all fairly new, and who knows if Anthropic will even exist in five years. OpenAI and some other giant AI companies are in the same boat.
8. “What else you got,” the lenders might reasonably ask the developers of the project. 
9. “Well, look,” say the developers, “Anthropic might not exist in five years, because the AI business is competitive and fast-moving and maybe someone else will be the winner. But the point is that computing power is the constraint on everyone, so even if Anthropic disappears, someone else — whoever the winner is in the AI race — will still want these chips in this data center. So if we lose our current tenant, we’ll easily be able to replace it with another tenant paying the same rate.”
10. This is not a bad argument, but it leaves the lenders underwriting the credit of an unknown tenant. “Some yet-to-be-founded AI company will provide your cash flows” is not soothing to lenders. “What else you got?”
11. The developers might say: “Even if we can’t find a tenant, we can sell the chips for a pretty high price, because, again, computing power is the enduring constraint of AI and somebody will always want it. This is not just a loan backed by cash flows; it’s backed by collateral (the chips).”
12. This again is not a bad argument, but it might still worry the lenders. Who says the chips will be valuable? Equity investors in the AI boom are telling an up-and-to-the-right story in which demand for AI will only increase and the chips will be in demand forever, but what do the lenders know about that? A world in which Anthropic stops making lease payments is very plausibly a world in which the chips have become worthless. “What else you got?”
13. “Fine,” say the developers. “We actually have a giant existing investment-grade public company involved in this deal, and that company is also very bullish on this whole AI thing. That company will promise to buy the chips at a fixed price, as last resort. So if everything goes wrong, we can put the chips to the guarantor and get money back to repay your loans. Happy?”
14. So the whole complicated structure is, at bottom, guaranteed by a giant investment-grade company. That’s something the lenders can understand.

We talked about a version of this — a project named “Beignet” — in October, when Meta was providing the guarantee (and also leasing the data center). And last week, Bloomberg’s Paula Seligson, Scott Carpenter and Silas Brown reported on a new deal for Anthropic:

Apollo Global Management Inc. and Blackstone Inc. are working to bring additional investors into a roughly $36 billion debt financing deal to help Anthropic PBC build out its AI infrastructure.

The debt will be used to purchase Google’s custom chips called TPUs, or tensor processing units, which Anthropic will then lease, according to people with knowledge of the matter. Broadcom Inc., which helps Google develop the chips, is backstopping payments on the largest portions of the transaction, said the people, who asked not to be identified because the information is private. …

In this type of structure, a special-purpose vehicle, or SPV, borrows the money and also receives an equity investment. That cash is used to buy the chips, which are then leased to a customer. The debt is backed predominantly by the lease payments, along with the unknown long-term value of the chips. …

Because Anthropic is still a startup and lacks the earnings to support such a large debt deal, the financing relies on Broadcom’s strong credit profile. But the structure allows both companies to stay at arm’s length from the transaction. …

A key feature of the deal is Broadcom providing a “residual value support” agreement, the people said. That means that if Anthropic fails to make the lease payments for a certain period of time, the SPV will sell the chips to pay back the debt investors. If the value of the chips doesn’t make the debt investors whole, then Broadcom will make up the shortfall for 100% of the value owed to the A1 and A2 investors.

That puts the A1 and A2 notes’ credit profile in line with Broadcom, which has an investment-grade rating, the people added. This type of structure means that Broadcom’s own ratings and balance sheet shouldn’t be affected by the transaction.

There is a sort of arbitrage, or at least market segmentation, here. This put option is very valuable to the SPV, to Apollo, to Anthropic, to the buyers of the debt and to the ratings agencies rating the debt: It puts an investment-grade floor on the value of the (A1 and A2) bonds. In the worst-case scenario, these bonds convert into unsecured obligations of an investment-grade company, [1]  so they can get an investment-grade rating. So the SPV can, in effect, pay Broadcom well for the put: Broadcom’s investment-grade balance sheet is what makes the whole AI machine go, so it can charge a high price to rent out that balance sheet.

But the put option is cheap for Broadcom to manufacture, in several respects:

• Presumably Broadcom actually thinks that the TPUs will be valuable in the future state of the world; it is not very worried about chip prices falling and having to pay out on the put. Credit investors are naturally nervous, but Broadcom is naturally bullish, so it makes sense that the credit investors would put a high value on buying the put and Broadcom would put a low value on selling it.
• There is a related IBGYBG element: In a world in which the AI boom fades and demand for computing chips collapses, Broadcom is going to have problems beyond this residual value guarantee. [2]
• The guarantee doesn’t affect Broadcom’s investment-grade balance sheet. A residual value guarantee is generally not counted as a liability for accounting purposes if it is not “probable,” meaning more likely than not, to pay out. Again, presumably Broadcom is bullish enough on chip prices not to worry about it.

So from the bondholders’ and ratings agencies’ perspective, this is (at worst) debt of Broadcom, and thus investment-grade. But from Broadcom’s and its accountants’ perspective, this is not debt of Broadcom, so its credit is not hurt by taking on more debt.

We talked last month about compute futures: The big commodities futures exchanges (CME, ICE) are launching trading markets on the future price of computing power. This is an effort to solve the same problem with market prices rather than debt guarantees: If you can lock in the price of compute in 10 years, you don’t need to buy a put from Broadcom. (Even if you don’t lock it in, just seeing the market price of compute 10 years out might get you more comfortable with the collateral value.) Financing the AI boom depends on getting lenders comfortable that the chips used in that boom will retain their value, so financial innovation is busy figuring out how to do that.

Elsewhere in debt capital markets: “For leveraged finance practitioners, artificial intelligence is the only game in town.”

Look, I’m not going to sit here and say Andrew Left is innocent. He was charged with securities fraud, and convicted yesterday, for a sort of pump and dump: He was an influential investor, he said publicly that he was short (or long) some stock, his followers on social media sold (or bought) the stock, and he would close out his position at a profit while still publicly saying he was short (or long). There were some bad messages suggesting that he was intentionally taking advantage of his influence, that his statements were not motivated by honest conviction and deep research but were just designed to make a quick buck by getting retail investors to trade. He might have committed a certain amount of light securities fraud. Everything is securities fraud, though, so I can’t really hold that against him.

What I will say, though, is that the victims of his fraud, called by prosecutors to testify against him, basically said that (1) they bought shares in dodgy cannabis companies, (2) Left said publicly “yo these companies are frauds,” (3) their stocks went down, (4) they lost money and (5) Left is mean. This drives me nuts? Left frequently was right: He would call out pretty vaporous companies, some of which were in fact sued by regulators for fraud, some of which in fact went more or less to zero, and he was prosecuted for that. People do not like short sellers because they are rude, and that seems to have been what got Left in trouble.

But not just short selling! One of the charges against Left was that he told his followers to buy Nvidia Corp. stock in November 2018. I would be retired if I had bought Nvidia stock in 2018. “We see $165 before we see $120,” he tweeted, when Nvidia was trading at $143.64. In fact Nvidia never again fell below $120, and it hit his $165 price target within two weeks. Also since then it has been Nvidia. It’s up about 6,200% since his tweet. [3]

Basically the criminal case against Andrew Left was that he tweeted honest correct opinions about stocks, but for bad reasons. Instead of doing deep fundamental research and coming to a reasoned defensible conclusion, he was like “ehh if I say this company is bad, people will believe me and it will go down and I’ll make a quick buck.” The reason people believed him is that he had a track record of correctly identifying companies that were bad. And then, I suppose, he started lazily cashing in on that reputation. But his instincts remained good! When he said companies were bad, often they were. 

Bloomberg’s Erik Larson reports:

In Left’s case, the government took issue with how quickly Left closed out his trading positions after publicly criticizing or boosting companies. That practice has long been polarizing, and Left’s case marked one of few times that the issue has faced trial.

Frank Zhang, an accounting professor at the Yale School of Management, said the verdict will have a chilling effect on short sellers, because “it will scare them into silence.”

“This sets a dangerous precedent for short sellers, who now fear that publishing negative research and exiting trades quickly will trigger federal audits and market manipulation charges,” Zhang said.

We have talked about this. As I wrote when Left was first charged, it makes a lot of sense for him to close his positions quickly:

Citron [Research, Left’s investment vehicle] is in this model a service provider to the market: It informs the market that Company X is a fraud, the market saves a lot of money …, and Citron gets a reasonable cut of the money immediately. Citron is not a long-term investor that has to ride Company X all the way down, and it doesn’t have to make 100% of its profits on any trade contingent on being 100% right about Company X. In the long run, all of its profits are contingent on being right enough about enough of its trades that people keep listening to it and prices keep going down.

The most efficient use of a short seller’s time and capital is to identify frauds, short them, point them out publicly, watch them drop, take an immediate profit and move on to the next trade. If he is usually right, he will make money by making the market more efficient; if he is usually wrong, people won’t listen to him. But if he’s usually right, he will also make people pretty angry.

“Prediction markets” is, mostly, a fancy way to say “sports gambling.” But it is a special kind of sports gambling, one that is regulated by the US Commodity Futures Trading Commission. In some ways this is uncomfortable, but in other ways not: The CFTC has historically regulated commodity futures trading, and people used to think that commodity futures trading was gambling. What is a wheat futures contract if not a bet on the future price of wheat?

The traditional answer is: Wheat futures are for hedging. Farmers can sell futures to lock in the price of wheat before they plant, so they can have predictable income. Bakers can buy futures to lock in the price of wheat next year, so they can have predictable costs. Oh of course in modern markets hedge funds and retail investors can buy and sell wheat futures without having any preexisting economic exposure to wheat prices, as pure speculative bets. But those speculative bets — even if they make up most of the volume — are epiphenomena on the real economic purpose of futures markets, which are to allow real industrial producers and consumers to hedge their risks. The speculators provide liquidity and price discovery and efficiency to a market with a real economic purpose.

You don’t have to (1) believe that or (2) care about it; you are perfectly free to think either “no, wheat futures speculation is bad and should be banned” or else “meh, wheat futures speculation is fun gambling and that’s good enough for me.” But the notion of an “economic purpose” of futures markets is important to a lot of people, including the CFTC itself, because the alternative (gambling) is sort of embarrassing. [4]

But prediction markets are different, in that they are sports gambling. Is the purpose of sports gambling to hedge real economic risk? Noooooooooooo. Noooooooooooo. No it is not. You can tell this from a second’s introspection: You have never bet on a sports game, or met anyone who has bet on a sports game, to hedge some existing economic risk. People do not bet on sports to make their cash flows steadier. People bet on sports to make their cash flows more variable. [5] Sports gambling is the opposite of a hedge. It is gambling. You can tell from the name. There is nothing confusing or mysterious or controversial or difficult about this.

Nonetheless, it would be convenient for a lot of people if the purpose of sports gambling was in fact to hedge real economic risk. Then sports gambling — prediction markets — would just be a regular sort of commodity derivatives trade, no different from wheat futures if you think about it, and the rapid rise of federally endorsed sports gambling would not be embarrassing.

And, of course, it is possible to hedge real economic risk with sports gambling. Sports are a big business! (You can bet on them.) A basketball team might make more money if it makes the playoffs; maybe its owner should bet against the team to hedge the risk of not making the playoffs. (Controversial!) Or: A basketball team might have to pay bigger bonuses to players if it makes the playoffs; maybe its owner should bet on the team to hedge the bonus risk. Or: A shoe company might pay a lot of money to a basketball player to wear its shoes, but will not get much value for that money if the player has a bad season; maybe it should bet against his performance to hedge its investment. Or: A bar near the stadium will sell more beer if the team makes the playoffs; maybe it should bet against the team to hedge its beer-selling risk. The CFTC has said in a court filing:

“Whether LeBron James will score 20 or more points in a game” or “whether the combined score in the annual Ohio State-Michigan football game will be over or under 50.5 points” … can be associated with potential economic consequences — how many LeBron James jerseys are sold, how many viewers stay tuned into the Ohio State-Michigan game, ticket prices for future events, and so on.

And, you know, fine, whatever. I’m not going to talk anyone out of this. “The real purpose of sports betting is to hedge beer sales, and all of the retail gamblers are just an epiphenomenon that provides liquidity to economic hedgers.” Okay. Do you hear yourself?

Anyway literally seven people have sent me this promotion from a New York bar offering free beer and food if the Knicks win on Wednesday night, hedged with a $5,000 bet on the Knicks on Kalshi. Sure! 

Because I am a grump, though, I have to point out that the guy is not going on Kalshi to hedge a preexisting economic risk. He’s going on Kalshi to hedge a sports bet! He bet everyone at his bar free food and beer if the Knicks win, and he’s hedging that bet on Kalshi. 

SpaceX Staffers Prep for Multimillion-Dollar Windfalls by Pushing for VIP Terms. How Deutsche Bank learned to stop chasing America. Gold replaces US Treasuries as world’s top reserve asset, ECB says. Silicon Valley’s $140 Billion Tax Break Is Going Mainstream. Hedge Fund Arrowpoint Hires 10 Trading Teams, Seeks More Capital. Castlelake: the private credit lender taking a run at easyJet. Aspiration Partners’ co-founder Joe Sanberg gets 14-year fraud sentence.  Endowment effect. University of California Professors Are Begging Schools to Reinstate the SAT. Trump Appoints Loyalist Bill Pulte as Acting Spy Chief. 

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!