This website uses cookies

Read our Privacy policy and Terms of use for more information.

Hey {{first_name|Investor}} -

The giants still have plenty of money. Their AI buildout is eating the cash-flow cushion that used to make borrowing costs irrelevant to them, and the dividing line now is how much flexibility each one has left.

Last Sunday I ended on a question. Alphabet brought in about $70 billion of outside money in a single quarter and finished it with $242.5 billion of cash and marketable securities. Why would one of the richest companies in the world raise money at all? This issue is my attempt to answer that, and to test my own view that the giants are the safe end of the AI trade.

The math changed this year

For most of the last decade, the biggest tech companies paid for data centers out of the cash their businesses threw off, with plenty left over for buybacks.

That cushion is shrinking. In July, Goldman Sachs projected the hyperscalers would spend about $750 billion on CapEx this year, meaning buildings and equipment, against roughly $778 billion of operating cash flow (Goldman, via Reuters, July 29). If that holds, about 96 cents of every dollar the core businesses bring in goes straight back into the buildout. Goldman has since raised its 2026 spending estimate to about $800 billion and sees $1.2 trillion in 2027 (Goldman, via Bloomberg, September 25), so the July ratio is the gentler version.

They've also stepped up borrowing. Amazon, Alphabet, Meta and Oracle issued about $194 billion of bonds this year through July 7, up from roughly $108 billion in all of 2025 (Reuters analysis of LSEG data, July 29). In September, Goldman projected hyperscaler bond issuance of about $420 billion in 2027 (Reuters).

Lenders are still lending, and they're charging a bit more for it. The extra yield these companies paid on new bonds, compared with their existing ones, rose to a median 0.12 percentage points this year from about 0.02 in 2025 (Reuters, July 29). Bonds they already sold keep their fixed rates. The higher price applies to the next one.

None of that is a crisis. These are some of the strongest borrowers in the world. But it means the giants now lean on the capital markets in a way they didn't 2 years ago, and that's where rates bite.

Why would a giant raise money it doesn't seem to need?

Alphabet is the clearest example. In June it sold common stock, plus mandatory convertible preferred stock that pays a 6.25% dividend and turns into common shares around May 2029. It also agreed to sell $10 billion of stock to Berkshire Hathaway, and set up a $40 billion program to sell shares into the market over time. With the underwriters' options, the whole package reached $90 billion (Cleary Gottlieb, Alphabet's legal adviser).

Part of that $90 billion is capacity, though, and the money actually received is smaller. In the second quarter Alphabet received about $49.6 billion from the stock sales and $20.3 billion from senior notes, and it hadn't sold any shares through the $40 billion program by June 30 (Alphabet Q2 release). That $49.6 billion plus $20.3 billion is the "about $70 billion."

The company says the money is for general corporate purposes, including CapEx to scale AI infrastructure. It also says about $30 billion of the $40 billion program is expected to cover taxes for employees whose stock awards vest this year (Alphabet, June 1). So a big slice of the headline replaces cash Alphabet would otherwise have spent on employee taxes.

And the $242.5 billion needs a footnote: $87.1 billion of it is marketable equity securities, meaning stakes in other companies (Alphabet Q2 call). Not all of it is cash Alphabet could spend on data centers.

Here's my reading of the rest. Alphabet guided to $195 billion to $205 billion of CapEx this year and a significant increase in 2027. Raising equity while the stock was high, alongside debt, lets it fund that without stretching its balance sheet. That's a strong company acting carefully. It's also a company preparing to spend more than its business generates, which is the part I want to track.

4 signs a giant's borrowing starts to matter

Strong companies borrow all the time, and borrowing alone tells you little. These are the signals I'd watch for the point where a giant has fewer choices.

1. Free cash flow is negative for a full year. That means cash from the business isn't covering capital spending. The company has to draw on cash it already has, sell assets, raise money or slow down. One negative quarter can be timing. A negative full year is a pattern.

2. Commitments move outside the headline numbers. Data centers financed through partner-owned ventures, or leased instead of bought, still have to be paid for. They show up as lease obligations and guarantees instead of CapEx or company debt. The more of this there is, the more the headline figures understate the commitment.

3. Shareholder payouts shrink. Buybacks stopping, or dividends running ahead of free cash flow, show where the cash is going.

4. Financing conditions worsen while cash coverage weakens. Watch the credit spread (the extra interest above government bonds), the premium on new bond deals, interest costs against earnings, upcoming maturities and rating changes. A worse bond market alone matters less if cash flow is strong.

You can check all 4 from a quarterly report and a few bond-deal headlines. One sign on its own can be noise. Two or three together are worth paying attention to.

The rest of this issue is for Insiders.

Below the line: Alphabet, Amazon, Microsoft, Meta and Oracle run through the 4 signs, on the same measures, with a table. The one that still pays its own way, the one that stopped buybacks, the one with a 12-month cash shortfall, the one with a $13 billion guarantee on a data center it doesn't own, and the one that looks more like last week's middle than like a giant.

logo

This is where it gets interesting.

Insiders get the full research — the emerging patterns, the second-order effects, and the trends I think matter before they're obvious

Unlock the full post

A subscription gets you:

  • Weekly deep dives on trends before they're consensus
  • Operator-level research and frameworks behind every thesis
  • Full archive of every past issue and report