
Weekend Update - W2623
▶ Explore this week’s Tape — live, sortable, drill-down →
Even the Cash Machines Are Borrowing Now
The story this week was not that the AI build is expensive. Everyone knew that. The story was that the companies best equipped to pay for it out of their own pockets stopped doing so — three of them, in the same five trading days.
Alphabet, which had not sold a share of stock since its IPO, raised about eighty-five billion dollars in equity to fund its data-center build, with Berkshire Hathaway taking ten billion of it directly (per CNBC).¹ Meta, which did not need a dime, was reported to be lining up a raise of its own — and shed almost seven percent on Friday on the rumor alone (per Bloomberg).² Apple, on page one of the same Cash Flow Memo, did the one thing nobody else in the arms race is doing: it rented.
Take the first one, because it is the one with thirty years of history behind it.
A company funds its capital spending out of its own cash flow right up until the build outgrows the cash flow. The day it reaches for outside money is the day the build stopped paying for itself. Alphabet’s trailing free cash flow fell about forty-four percent year over year — the capex ate it — so they sold stock.³ The company that described itself as the asset-light, capital-light compounder for two decades just issued eighty-five billion dollars of equity to buy data centers. Asset-light.
The cashflow read is in Marcus’s column below — short version: at roughly seventy times trailing free cash flow, this is now a capital-intensive company that happens to own a search engine.⁴
But the tell this week was not Alphabet alone. It was the synchronization. Meta still throws off about fifty billion dollars of trailing free cash flow, still growing north of twenty percent, and trades at thirty-two times — the cheapest large-cap cash machine on the board.⁵⁶ When the best-capitalized company in a sector decides it wants a financing cushion anyway, that is not a statement about that company. It is a statement about the size of the build.
We have watched real demand get financed externally before. The late-nineties fiber boom funded a build the internet genuinely needed — and the companies laying the cable raised equity and debt into the same thesis at the same moment. The capacity got used. A decade later. The financing peak and the equity peak landed in the same window, and the build being right did not save the multiple. The lesson was never that the demand was fake. It was that synchronized external financing is what a capital cycle looks like at its most expensive, not its cheapest.
Which is what makes Apple the sharpest line on the page. Apple pays Google roughly a billion dollars a year to run the next Siri on Gemini rather than train a frontier model itself — less than one percent of the hundred-twenty-nine billion dollars in free cash flow it generates annually.⁷⁸ Everyone else is spending a hundred times that and selling stock to do it. Dilute, borrow, or rent: only one of the three costs the shareholder nothing.
So the demand side is now financed. What gets tested next is absorption — whether the capacity fills before the financing cycle turns. The first data point lands Wednesday after the close, when Oracle reports against a backlog north of five hundred fifty billion dollars and trailing free cash flow that is already negative.⁹¹⁰ Broadcom just showed everyone the grading curve: a record AI-chip quarter, up more than a hundred-forty percent (per Broadcom’s fiscal-Q2 release), and the stock fell fifteen percent anyway because Hock Tan declined to raise the story he had already sold.¹¹¹² At these multiples, delivering is not enough; you have to deliver and beat the build you already financed. Mark the calendar for Wednesday. The thesis breaks if the backlog keeps growing and the cash behind it never shows up.
Wall Street’s consensus on the AI build: the balance sheets are fortresses and the capex funds itself. Three of the richest companies on earth sold or floated stock this week to tell you it doesn’t.
The Tape — W2623
Universe of 94 cashflow-memo names, snap dates 2026-06-02 → 2026-06-05. Composite is rank-sum percentile of FCF Yield + NTM Revenue Growth (higher = better balance). Banks and finance-book names shown separately.
Telltales Yield — Top 10
From the Cashflow Desk — Marcus Graham
The name sitting at the top of our own board is the one that does not have to finance anything. Uber leads the composite this week on a 7.5% free-cash-flow yield, growing the forward top line about fifteen percent, at thirteen times EV to free cash flow. It converts cash and grows without selling a data center or a share to do it — the asset-light, self-funding profile Alphabet used to own before it raised eighty-five billion dollars of equity this week to pay for the build. The screens still file Uber under gig economy and the hyperscalers under quality compounders. The cash flow says the labels are backwards. What changes the read is the day Uber has to raise outside capital the way the hyperscalers now are; until then, the top of the board is the part of the market still paying for itself.
Telltales Yield — Bottom 10
This Week’s Reporters
Note: Oracle’s 46.3% NTM revenue-growth figure is the raw FMP analyst-estimates consensus and looks anomalous against Oracle’s own ~15% forward guide; we read Oracle through its RPO backlog and trailing free cash flow (negative this quarter on the capex cycle), as in The Take above, not this estimate.
Sector Medians
Debt / FCF Watch (highest leverage on TTM FCF)
Weekly Price Movement
Top 5 (week-over-week price)
Bottom 5 (week-over-week price)
Banks (shown separately — FCF metric not meaningful)
Finance-book — FCF not comparable
Customer-float / captive-finance / reserve businesses (IBKR broker float, KMX CarMax Auto Finance, PYPL customer funds, CRCL stablecoin reserves). The memo’s operating-FCF method overstates their FCF, so they are held off the ranked leaderboard pending the P&L-waterfall rebuild.
Data Gaps
90 of 90 ranked-eligible names ranked. 0 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable).
Source: cashflow-memo master_2026-06-05.csv. NTM growth from FMP analyst-estimates consensus. Composite is a percentile rank, not a recommendation.
The Issue — This Week's Brief
The Cashflow Memo
Who Pays for the AI Build?
The week the AI trade stopped being about the chips and became a question about who pays for them.
The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the universe of companies in the Cash Flow Memo. About 13 minutes. No filler.
Download the memo at telltales.us. Hunt, Jason, and Mike are back Wednesday on episode E2624.
Chapter markers
* Time | Segment
* 0:00 | Disclaimer
* 0:15 | Cold open
* 0:45 | Theme — Who pays for the AI build
* 4:45 | Deep dive — AI infrastructure’s two-act week
* 8:45 | Rapid-fire + the forward week
* 11:45 | Close + Consensus Watch
Full transcript
Disclaimer
Ava: The following conversation is intended for informational purposes only. You should always do your own work to determine if an investment is suitable for you.
Cold open
Ava: You’re listening to the Telltales Weekend Update. I’m Ava Cabot.
Marcus: And I’m Marcus Graham — the cashflow desk.
Ava: Quick note: the show is produced entirely with AI tools, and both voices you’re hearing are AI-generated. Send feedback through the Substack. We’re still in pilot, so tell us what’s working and what isn’t.
Ava: Here’s the week. For two years, the AI trade was a story about chips — who makes the fastest one, who gets the allocation. This week it turned into a different question, and it’s a harder one. Who actually pays for the build? Broadcom printed a record and got punished for it. Oracle’s about to walk into the same exam on Wednesday. And three of the biggest companies on earth spent the week showing you three completely different ways to fund the thing — d…
This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit telltales.substack.com
Even the Cash Machines Are Borrowing Now
The story this week was not that the AI build is expensive. Everyone knew that. The story was that the companies best equipped to pay for it out of their own pockets stopped doing so — three of them, in the same five trading days.
Alphabet, which had not sold a share of stock since its IPO, raised about eighty-five billion dollars in equity to fund its data-center build, with Berkshire Hathaway taking ten billion of it directly (per CNBC).¹ Meta, which did not need a dime, was reported to be lining up a raise of its own — and shed almost seven percent on Friday on the rumor alone (per Bloomberg).² Apple, on page one of the same Cash Flow Memo, did the one thing nobody else in the arms race is doing: it rented.
Take the first one, because it is the one with thirty years of history behind it.
A company funds its capital spending out of its own cash flow right up until the build outgrows the cash flow. The day it reaches for outside money is the day the build stopped paying for itself. Alphabet’s trailing free cash flow fell about forty-four percent year over year — the capex ate it — so they sold stock.³ The company that described itself as the asset-light, capital-light compounder for two decades just issued eighty-five billion dollars of equity to buy data centers. Asset-light.
The cashflow read is in Marcus’s column below — short version: at roughly seventy times trailing free cash flow, this is now a capital-intensive company that happens to own a search engine.⁴
But the tell this week was not Alphabet alone. It was the synchronization. Meta still throws off about fifty billion dollars of trailing free cash flow, still growing north of twenty percent, and trades at thirty-two times — the cheapest large-cap cash machine on the board.⁵⁶ When the best-capitalized company in a sector decides it wants a financing cushion anyway, that is not a statement about that company. It is a statement about the size of the build.
We have watched real demand get financed externally before. The late-nineties fiber boom funded a build the internet genuinely needed — and the companies laying the cable raised equity and debt into the same thesis at the same moment. The capacity got used. A decade later. The financing peak and the equity peak landed in the same window, and the build being right did not save the multiple. The lesson was never that the demand was fake. It was that synchronized external financing is what a capital cycle looks like at its most expensive, not its cheapest.
Which is what makes Apple the sharpest line on the page. Apple pays Google roughly a billion dollars a year to run the next Siri on Gemini rather than train a frontier model itself — less than one percent of the hundred-twenty-nine billion dollars in free cash flow it generates annually.⁷⁸ Everyone else is spending a hundred times that and selling stock to do it. Dilute, borrow, or rent: only one of the three costs the shareholder nothing.
So the demand side is now financed. What gets tested next is absorption — whether the capacity fills before the financing cycle turns. The first data point lands Wednesday after the close, when Oracle reports against a backlog north of five hundred fifty billion dollars and trailing free cash flow that is already negative.⁹¹⁰ Broadcom just showed everyone the grading curve: a record AI-chip quarter, up more than a hundred-forty percent (per Broadcom’s fiscal-Q2 release), and the stock fell fifteen percent anyway because Hock Tan declined to raise the story he had already sold.¹¹¹² At these multiples, delivering is not enough; you have to deliver and beat the build you already financed. Mark the calendar for Wednesday. The thesis breaks if the backlog keeps growing and the cash behind it never shows up.
Wall Street’s consensus on the AI build: the balance sheets are fortresses and the capex funds itself. Three of the richest companies on earth sold or floated stock this week to tell you it doesn’t.
The Tape — W2623
Universe of 94 cashflow-memo names, snap dates 2026-06-02 → 2026-06-05. Composite is rank-sum percentile of FCF Yield + NTM Revenue Growth (higher = better balance). Banks and finance-book names shown separately.
Telltales Yield — Top 10
From the Cashflow Desk — Marcus Graham
The name sitting at the top of our own board is the one that does not have to finance anything. Uber leads the composite this week on a 7.5% free-cash-flow yield, growing the forward top line about fifteen percent, at thirteen times EV to free cash flow. It converts cash and grows without selling a data center or a share to do it — the asset-light, self-funding profile Alphabet used to own before it raised eighty-five billion dollars of equity this week to pay for the build. The screens still file Uber under gig economy and the hyperscalers under quality compounders. The cash flow says the labels are backwards. What changes the read is the day Uber has to raise outside capital the way the hyperscalers now are; until then, the top of the board is the part of the market still paying for itself.
Telltales Yield — Bottom 10
This Week’s Reporters
Note: Oracle’s 46.3% NTM revenue-growth figure is the raw FMP analyst-estimates consensus and looks anomalous against Oracle’s own ~15% forward guide; we read Oracle through its RPO backlog and trailing free cash flow (negative this quarter on the capex cycle), as in The Take above, not this estimate.
Sector Medians
Debt / FCF Watch (highest leverage on TTM FCF)
Weekly Price Movement
Top 5 (week-over-week price)
Bottom 5 (week-over-week price)
Banks (shown separately — FCF metric not meaningful)
Finance-book — FCF not comparable
Customer-float / captive-finance / reserve businesses (IBKR broker float, KMX CarMax Auto Finance, PYPL customer funds, CRCL stablecoin reserves). The memo’s operating-FCF method overstates their FCF, so they are held off the ranked leaderboard pending the P&L-waterfall rebuild.
Data Gaps
90 of 90 ranked-eligible names ranked. 0 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable).
Source: cashflow-memo master_2026-06-05.csv. NTM growth from FMP analyst-estimates consensus. Composite is a percentile rank, not a recommendation.
The Issue — This Week's Brief
The Cashflow Memo
Who Pays for the AI Build?
The week the AI trade stopped being about the chips and became a question about who pays for them.
The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the universe of companies in the Cash Flow Memo. About 13 minutes. No filler.
Download the memo at telltales.us. Hunt, Jason, and Mike are back Wednesday on episode E2624.
Chapter markers
* Time | Segment
* 0:00 | Disclaimer
* 0:15 | Cold open
* 0:45 | Theme — Who pays for the AI build
* 4:45 | Deep dive — AI infrastructure’s two-act week
* 8:45 | Rapid-fire + the forward week
* 11:45 | Close + Consensus Watch
Full transcript
Disclaimer
Ava: The following conversation is intended for informational purposes only. You should always do your own work to determine if an investment is suitable for you.
Cold open
Ava: You’re listening to the Telltales Weekend Update. I’m Ava Cabot.
Marcus: And I’m Marcus Graham — the cashflow desk.
Ava: Quick note: the show is produced entirely with AI tools, and both voices you’re hearing are AI-generated. Send feedback through the Substack. We’re still in pilot, so tell us what’s working and what isn’t.
Ava: Here’s the week. For two years, the AI trade was a story about chips — who makes the fastest one, who gets the allocation. This week it turned into a different question, and it’s a harder one. Who actually pays for the build? Broadcom printed a record and got punished for it. Oracle’s about to walk into the same exam on Wednesday. And three of the biggest companies on earth spent the week showing you three completely different ways to fund the thing — d…
This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit telltales.substack.com
Loading content...
Read the full transcript
Matterfact has every podcast on Apple Podcasts and Spotify: 700K+ channels, 120M+ episodes.
The candid version is on a podcast,
not the earnings call.
Matterfact turns the conversations your coverage is already having into a searchable, citable record — across every episode, not just this one.