Friday charts

SpaceX is in talks to borrow $40 billion from banks and asset managers to finance its next giant purchase of NVIDIA GPUs.

Not to be outdone, Broadcom is seeking to borrow as much as $100 billion. The money would fund a special-purpose vehicle to buy Broadcom chips and lease them to Anthropic and OpenAI.

Oracle is also said to be asking to borrow billions — from that same group of lenders, for partly the same customers, and mostly the same purpose (buying more chips).

All told, AI-related borrowing is forecast to hit $570 billion this year. 

The dramatic rise in bond yields over the past month suggests this is testing the outer limits of the bond market. “The AI boom is the main cause of soaring long-term interest rates,” Paul Krugman says. 

The boom is testing the limits of equity investors, too. 

OpenAI is reportedly hoping to raise another $30 billion in equity capital, despite having raised $122 billion six months ago.

Anthropic is expected to ask investors for $100 billion in their IPO as soon as next month.

It won’t be the last time they ask. Morgan Stanley estimates that AI infrastructure will require $1.5 trillion in external financing by 2028.

(Fun fact: 2028 is just 14 months away!)

With luck, the giant investment in AI will usher in an age of abundance, in which our material needs are easily met and work is optional because super-intelligent robots do everything for us.

In the meantime, though, we live in a world of scarcity.

“All that money going into construction of data centers and purchasing foreign-produced semiconductors comes at the expense of investment in everything else,” Krugman says. “The AI boom is crowding out everything.”

This is new.

Historically, the concern has been that government borrowing crowds out private-sector investment. 

Now, it’s almost the opposite: private-sector investment in AI is making it harder for the government to borrow.

Over the past 12 months, hyperscaler borrowing has equaled an incredible 14.1% of new Treasury borrowing. That’s up from 0% as recently as two years ago.

That won’t stop the government from borrowing, of course. They’ll just have to pay more for it and run bigger deficits.

But it will stop a lot of other borrowing. And slow the rest of the economy.

Until now, the investing debate around AI has been about how much abundant intelligence could grow the economy and how far it might make the stock market go up.

Now, investors are asking how much it could shrink the non-AI economy and how far it could make the non-AI stock market go down.

Let’s see if a few charts can help us figure it out.

Crowded construction:

Construction of data centers is driving up interest rates, wages, and input costs, making construction of anything else much less economic.

The input costs:

In the long run, AI should be disinflationary. But not until all the data centers are built.

AI is making it harder to move:

With US 30-year mortgage rates up to 7.3%, almost everyone who wants to sell one house to buy another will have to refinance at a higher rate.

Market cap is crowded at the top:

The top five stocks in the S&P 500 are worth the same as the bottom 434 stocks. A record.

Earnings are crowded at the top, too:

Just two firms, Micron and NVIDIA, are expected to account for one-third of the S&P 500 earnings in 2026. Amazing.

The Last of Us: 

Once Anthropic and OpenAI have IPO’d, Berkshire Hathaway will be the last remaining non-tech stock in the top 15 of the S&P 500.

2028:

Coatue estimates that in 2028 AI capex will be 50% bigger than US single-family mortgages and nearly three times the national defense budget.

You have to choose, apparently:

Since the launch of ChatGPT, tech spending in the US has been a mirror image of non-tech spending.

Crowding-in:

Paul Krugman notes that the current dynamic of AI crowding out non-AI investment is the opposite of what happened in the dotcom boom. “During the 90s boom America attracted very large inflows of investment from abroad, which effectively financed the tech boom even as other investment rose.”

Historic times:

As a percentage of GDP, the investment in AI is expected to be the largest ever: more than three times bigger than building the highway system and seven times bigger than electrification.

Let’s hope it’s that much more productive, too.

Have an uncrowded weekend, non-AI readers.

— Byron Gilliam

_{Brought to you by:}

Meridian 2026 is Stellar's annual gathering for the institutions, fintechs, and developers putting financial infrastructure onchain.

Join them October 28-29 at Convento do Beato in Lisbon for two days on tokenization, payments, and what it takes to run them in production.

Register today with Blockworks10 for 10% off.