Weekend Update - W2635

About this episode

▶ Explore this week’s Tape — live, sortable, drill-down →

The Bid Was the Business

PayPal spent the summer being priced by a buyer rather than by a business, and on Friday it found out which of the two it actually had. Advent and Stripe abandoned an approach that had run to roughly fifty-three billion dollars, a deal that would have ranked among the largest leveraged buyouts ever attempted¹. The stock gave back thirteen percent². Nothing about the company changed on Thursday night. What changed was the number of people willing to pay for it.

Price what walked away. The Cash Flow Memo has PayPal at about six and a half times trailing free cash flow, near a fifteen percent free-cash-flow yield, on close to seven billion dollars of trailing cash generation³. Then the memo does something to PayPal it does not do to anyone on the leaderboard: it takes the name off the ranked list entirely, because an operating-cash-flow method overstates cash for a business that holds customer funds⁴. So take the haircut. Take a generous one. You are still looking at a company the public market prices in single digits, that a buyout firm and a strategic spent a summer trying to take private, and that as of Friday has no bid above the tape.

Six and a half times is only a price if somebody can fund it. A fifty-three-billion-dollar take-private does not get funded out of equity. It gets funded out of the leveraged-loan and private-credit market, which spent this same week being asked for something else. CNBC reported Broadcom in talks over a chip financing package running to seventy or eighty billion dollars, forty-five senior and thirty-five junior, with Blackstone and Apollo among the firms discussed as putting up the money⁵. One deal, one week, aimed at chips that have not shipped.

Larger than the entire PayPal buyout.

Those are not literally the same firms, and the comparison is an argument rather than a report. They are the same appetite. Private credit is a finite bid, and for most of the last decade its natural habitat was the PayPal trade: mature, cash-generative, unloved, with enough coverage to carry debt. That capital now has a shorter, better-paying alternative underwriting infrastructure. When the marginal lender has somewhere better to be, the marginal buyer of cheap cash flow stops turning up, and a multiple that looked like a mispricing turns out to be the clearing price.

Salesforce is the control. It gained twenty-three percent on Thursday, its best day since 2020, for producing evidence that its cash flow survives the technology everyone assumed was coming to kill it⁶. Nobody had to finance that. The equity market re-rated it in a single session, at no cost to anybody. PayPal’s re-rating needed a sponsor, a lender and six weeks, and it did not survive the stock rising into the offer. A leveraged bid with no premium left in it is an expensive way to own what you were already looking at.

Marcus’s column below is on Micron, which has the opposite complaint: the market will pay for the cash flow, just not for more than one year of it.

What changes the read. The near test is not PayPal’s next print. Enrique Lores took over in March and is running a standalone plan⁷; that is a two-year story and the tape will price it as one. The test that matters is Wednesday and the weeks after it: whether the Broadcom package clears anywhere near its reported size, and where the junior tranche prices⁸. If seventy or eighty billion dollars of AI paper places easily, the bid underneath every cheap cash-flow name in the memo stays where it went. If the junior tranche struggles, capital comes back to businesses that already generate cash, and this is the first name it re-prices. The frame breaks if a strategic, rather than a sponsor, pays a real premium for a mature cash-flow business in the next two quarters. That would say the equity market is the marginal buyer again, and none of the above matters.

Wall Street’s consensus on PayPal: a cheap stock that just lost its catalyst. The catalyst was never PayPal’s to lose. It belonged to the credit market, and the credit market is busy.

The Tape — W2635

Universe of 94 cashflow-memo names, snap dates 2026-08-21 → 2026-08-28. 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

Micron is the one name in the top ten where the forward multiple and the trailing multiple are arguing, and the forward one is winning. The tape has MU at a 6.1x forward P/E against 92.8% NTM revenue growth — that combination is the market underwriting one enormous year and then a cycle turn. The trailing side says the cash has not arrived yet: a 2.6% FCF yield, because capex is consuming almost everything the memory cycle generates. Consensus reads a single-digit forward multiple on a semiconductor as cheap. It is a duration bet, and the duration is one year. The test is whether capex converts to free cash flow before pricing rolls. Thesis breaks if the NTM growth estimate starts coming down while the capex commitments stay fixed.

Telltales Yield — Bottom 10

This Week’s Reporters

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

91 of 92 ranked-eligible names ranked. 1 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable).

Source: cashflow-memo master2026-08-28.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation.

The Issue — This Week's Brief

The Cashflow Memo

The Week the AI Build Went on Credit

Oracle and Amazon burned tens of billions to build it. Salesforce sells it for 13 times free cash flow, Palantir for 125.

The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the companies in the Cash Flow Memo. About 14 minutes. No filler.

Download the memo at telltales.us. Hunt, Jason, and Mike are back Wednesday on episode 2636.

Chapter markers

* Time | Segment

* 0:00 | Cold open — the build stops paying for itself

* 0:45 | Theme — Who’s paying for the build: Oracle, Broadcom, Amazon

* 4:45 | Deep dive — Salesforce vs. Palantir

* 8:45 | Rapid-fire — Lantheus, PayPal, and the forward week

* 11:45 | Close — Consensus Watch

* 12:40 | Disclaimer

Full transcript

Cold open

Ava: Three companies, three different answers to one question: who actually pays for the AI build? Oracle’s answer is that the customer does — the VA raised its contract ceiling by $17 billion[^news-orcl-va-20260820]. Broadcom’s answer is that it will co-sign, reportedly putting its balance sheet behind up to $80 billion of somebody else’s borrowing[^news-avgo-debt-20260821]. Amazon’s answer is that it will just pay — 2 million more GPUs, announced Thursday[^news-amzn-gpus-20260827]. And on the other side of all that spending, two companies sold AI software this week to enormous applause, at multiples almost 10 times apart. Somebody is wrong about what this revenue is worth.

Ava: Telltales Weekend Update. I’m Ava Cabot, with Marcus Graham at the cashflow desk.

Theme — Who’s paying for the build

Ava: Start on page 2 of the Cash Flow Memo, where Oracle, Broadcom, and Salesforce all sit. Three companies, three completely different answers to the same question this week: who is actually paying for the AI build?

Ava: Oracle’s answer is that somebody else is. The Department of Veterans Affairs raised the ceiling on Oracle’s health-records contract by up to $17 billion last Thursday, taking the whole deal from just under $10 billion to roughly $27 billion, and extending the work through 2031[^news-orcl-va-20260820]. The VA’s stated reason, quoting the modification: unanticipated complexities slowed software deployments, which resulted in the contract’s ceiling being reached sooner than originally planned[^news-orcl-va-20260820]. The agency runs 164 medical facilities and now hopes to finish all of them by 2031[^news-orcl-va-20260820]. More money, more years, same job. And the market has already filed its opinion on how that trade is going: Oracle closed the week around $151 against a 52-week high of $346[^memo-orcl-price-20260828]. That is more than half the company gone in under a year, while the order book got bigger. Marcus — what does Oracle look like from the cash side right now?

Marcus: Oracle is spending almost as much on capital equipment as it books in revenue, and that is the whole Oracle story at the moment. Trailing 12 months, capex ran about $56 billion[^memo-orcl-capex-20260828] against $67 billion of revenue[^memo-orcl-rev-20260828]. Free cash flow came in at negative $19 billion[^memo-orcl-fcf-20260828]. So the multiple isn’t the right frame on this name — there isn’t one, the denominator is negative, and that’s the cost of the build rather than a flag. What actually prices Oracle is whether the contracted revenue lands on schedule. The VA contract is a tell on exactly that, and it isn’t a flattering one. The ceiling went up because the deployment went slow.

Ava: Broadcom’s answer is different. Broadcom will co-sign for you.

Ava: CNBC reported a week ago Friday that Broadcom is in talks over a chip financing deal of upwards of $70 to $80 billion, aimed at supporting AI companies — Anthropic among them[^news-avgo-debt-20260821]. Look at the shape of it. $45 billion senior, $35 billion junior, and Blackstone and Apollo among the firms in talks to put the money up[^news-avgo-debt-20260821]. The cash is theirs. What Broadcom is putting in is its credit. And on Tuesday OpenAI publicly called Broadcom’s custom AI chip a significant advance[^news-avgo-openai-20260825], which is a generous thing to say about a supplier you would like to keep co-signing for you. Broadcom reports Wednesday[^earn-avgo].

Marcus: Vendor financing isn’t a dirty word by itself. GMAC was vendor financing, and its loss rates held under 1% through the Depression. Structure decides it. What’s different here is that Broadcom is writing a guarantee rather than making a loan, so nothing lands on its books at inception — by design. Going into the print the memo has Broadcom carrying about $65 billion of total debt[^memo-avgo-debt-20260828], so a $70-80 billion guarantee is larger than everything the company has actually borrowed. It pays the shortfall rather than the notional, so read it as a ceiling. What I’d watch Wednesday is whether that guarantee shows up in the commitments footnote at all. It has never appeared in Broadcom’s own releases.

Ava: And Amazon’s answer is the oldest one in business. Amazon just pays. AWS said Thursday it is deploying an additional 2 million Nvidia GPUs — Blackwell Ultra and Rubin — across its data centers through 2028. That’s Amazon’s own newsroom, not a leak[^news-amzn-gpus-20260827]. No financing structure, no warrant, no consortium. They are writing the check.

Marcus: It’s the largest check in corporate history and almost nobody frames it that way. Amazon’s trailing-12-month capex is about $173 billion[^memo-amzn-capex-20260828]. Not a guide — spent. Free cash flow is negative $8 billion as a result[^memo-amzn-fcf-20260828], and against an enterprise value near $3 trillion[^memo-amzn-ev-20260828] that is a rounding error. Which is the point. Amazon is the only one of these three that needs nobody’s permission to build. Oracle needs the contract to land on schedule. Broadcom needs the credit market to stay open.

Deep dive — Salesforce vs. Palantir

Ava: Two companies sold AI software this week. Both beat, both raised, both got rewarded. And the market is paying almost 10 times more for one of them than the other — for revenue that is arriving right now, in both cases.

Ava: Salesforce first, back on page 2 — the company AI was supposed to kill. Wednesday’s print: revenue $11.3 billion, up 11%, adjusted earnings of $5.90 a share[^news-crm-q2-20260826]. Benioff announced Claudeforce, an expanded partnership putting Anthropic’s Claude inside the Salesforce platform[^news-crm-claudeforce-20260826], and raised the full-year outlook to about $46 billion[^news-crm-q2-20260826]. The stock rose 23% Thursday — the best day since 2020, and the second-biggest in the company’s history[^news-crm-surge-20260827].

Ava: And underneath the headline, the number that actually settles the argument. Agentforce and Data 360 together are now at nearly $3.9 billion of annual recurring revenue, up more than 210% year over year. Agentforce on its own passed $1.5 billion, up more than 240%[^news-crm-arr-20260826]. That is not a pilot. That is a product line.

Ava: Palantir is the opposite trade — the company AI was supposed to make unstoppable. Revenue up 93%[^memo-pltr-growth-20260828]. And the Pentagon formally designated the Maven Smart System a program of record, which means Maven stops being a pilot with a contract ceiling and becomes a budget line, with the Army taking over the contracting[^news-pltr-maven-20260825].

Ava: And the money behind that designation has a shape worth hearing. The Pentagon’s initial Maven contract in 2024 was worth up to $480 million. The program ceiling rose to $1.3 billion in 2025. The request now is $2.3 billion over the next five years[^news-pltr-maven-20260825]. Marcus — which of those two is the market getting wrong?

Marcus: The surprising one is Salesforce, and not because of the print. Salesforce is the company AI was supposed to kill — the SaaSpocalypse trade — and going into Wednesday the memo had it at about 13 times trailing free cash flow at a roughly 8% free-cash-flow yield, Q1 10-Q confirmed[^memo-crm-priorqtr-20260828]. We re-anchor when the Q2 10-Q files. That is a distressed multiple sitting on $15 billion of trailing free cash flow[^memo-crm-priorfcf-20260828]. The market spent a year pricing this business as a melting ice cube, and this week the melting ice cube sold AI and grew.

Ava: A year of that thesis. One quarter to unwind it.

Marcus: Palantir is the mirror image, and the business is genuinely excellent — I want that on the record before the multiple. The memo has it at about 125 times trailing free cash flow[^memo-pltr-evfcf-20260828], on roughly $3.5 billion of trailing free cash flow[^memo-pltr-fcf-20260828], growing 93%[^memo-pltr-growth-20260828]. Program-of-record is real and it’s durable; budget lines survive administrations in a way pilot contracts don’t. But at that multiple you aren’t paying for Maven. You’re paying for the next several Mavens, arriving on schedule. I’d call that more likely than not, and a long way from certain.

Ava: Marcus — how much of Palantir is that one customer?

Marcus: Enough that it’s the right question to ask. US government revenue was $809 million in the quarter against total revenue just under $2 billion[^news-pltr-govt-20260825], so call it a bit over 40%, and it’s the faster-growing half. Concentration in a defense budget line is a different animal from concentration in one enterprise account — the budget line is stickier and it moves slower. What it is not is diversified. For the multiple to hold, the commercial side has to keep running triple digits, and that’s the half with real competition in it.

Ava: Triple digits. Forever. With competition.

Marcus: Same week, same product category, same customer budgets getting bigger. One of them is priced like the AI story already ended badly, the other like it cannot end badly at all. My read is that Salesforce carries the smaller error term — you’re paying an ordinary multiple for cash that already exists, and the AI revenue sits on top as optionality. With Palantir the cash has to show up before the multiple makes sense. What I’d watch on both is the same number: net revenue retention. That’s where the AI attach either compounds or it doesn’t.

Ava: Almost 10 times apart, on the same page of the memo, in the same week, selling the same thing to the same budgets. One of those two prices is going to look ridiculous in two years. Nobody in the market can tell you which one.

Rapid-fire

Ava: Two takeover stories this week, opposite endings, and the memo tells you why.

Ava: Lantheus. Curium is buying the entire company for $102.50 a share in cash, plus up to $12 a share in contingent value rights, and Lantheus filed its preliminary merger proxy Wednesday — the document that sets up the shareholder vote and the Nasdaq delisting that follow if it closes[^news-lnth-proxy-20260826]. What is Curium buying? The memo has Lantheus at about 12 times trailing free cash flow, at an 8% free-cash-flow yield[^memo-lnth-evfcf-20260828]. Plus a pipeline that just cleared the FDA — TAUKLARIFY, its tau PET imaging agent for Alzheimer’s evaluation, approved two weeks ago[^news-lnth-fda-20260814]. The stock closed around $100 against $102.50 in cash[^memo-lnth-price-20260828]. The market is not arguing with this one.

Ava: PayPal is the other ending. Bloomberg reported Friday that the Advent and Stripe consortium has dropped its pursuit — a bid that had reached $60.50 a share, about $53 billion, which would have ranked among the largest leveraged buyouts ever attempted. The board had called the earlier offer inadequate. And the reason the buyers walked is almost funny: PayPal shares had already climbed more than 40% this quarter on a strong second-quarter print, which pushed the market value up toward $53 billion on its own. The company got too expensive by being too good. Shares fell about 13% Friday[^news-pypl-bid-20260828].

Ava: And here is the part that ties this whole episode together. The memo has PayPal at 6.5 times trailing free cash flow, at a 15% free-cash-flow yield[^memo-pypl-evfcf-20260828]. That is the highest free-cash-flow yield in the entire memo universe. Somebody offered $53 billion for it and then walked away, because it got too good. That is what the market pays for cash that already exists — in the same week it paid 125 times trailing free cash flow for cash that mostly doesn’t yet[^memo-pltr-evfcf-20260828].

Ava: Forward week. Broadcom reports Wednesday, and that’s the one that matters — consensus around $3.22 a share on roughly $29 billion in revenue, though the financing question is bigger than the print[^earn-avgo]. Snowflake, also Wednesday[^earn-snow]. Five Below, Wednesday[^earn-five]. And Oracle reports the following Tuesday, September 8[^earn-orcl].

Close

Ava: Wall Street’s consensus this week: AI is coming for software, so own the picks and shovels and avoid the applications. Salesforce just had its best day in six years by selling an application. Consensus was early, at best.

Ava: Which brings it back to where we started. Three ways to pay for the build — Oracle bills the customer, Broadcom co-signs, Amazon writes the check. Two of the three are running free cash flow negative to do it: Oracle at negative $19 billion trailing[^memo-orcl-fcf-20260828], Amazon at negative eight[^memo-amzn-fcf-20260828]. And the cheapest cash flow in the universe couldn’t find a buyer at any price. Somebody’s arithmetic is wrong here, and it tends to show up in the free cash flow line first. That’s why we publish the Cash Flow Memo. Download it at telltales.us.

Ava: On Wednesday’s show, episode 2635, Hunt said he doesn’t see how an investor can not own SpaceX. Mike didn’t disagree so much as price it. You can’t get a large margin of safety here by traditional methods, because you have to credit them for something nobody has done yet — so he underwrote it on the terrestrial data centers alone, and treated the space-based ones as option value on top. He’d wanted $80 a share; it touched $100. His answer was to buy some anyway and hold your nose, at a size small enough that being wrong doesn’t cost you the portfolio[^ep-e2635]. Hunt, Jason, and Mike are back Wednesday on episode 2636.

Ava: If we got something wrong, or there’s a name you want covered, send it through the Substack. Every one of them gets seen.

Ava: The show is produced entirely with AI tools, and both voices you’re hearing are AI-generated.

Disclaimer

Ava: This podcast is intended for informational purposes only. You should always do your own work to determine if an investment is suitable for you. The views expressed on this podcast are the host alone and do not constitute an offer to sell or a recommendation to purchase, or a solicitation of an offer to buy any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the host nor any of their employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness, or completeness of this information. The host and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future, and may or may not hold positions in the securities mentioned.

Sources

* Amazon. (2026, August 27). AWS and NVIDIA expand collaboration to deploy 2 million additional GPUs [Press release]. About Amazon. https://www.aboutamazon.com/news/aws/aws-nvidia-2-million-gpus-ai [^news-amzn-gpus-20260827]

* Broadcom debt deal expected to reach upwards of $70 billion, sources say. (2026, August 21). CNBC. https://www.cnbc.com/2026/08/21/broadcom-debt-deal-expected-to-reach-upwards-of-70-billion-sources.html [^news-avgo-debt-20260821]

* Lantheus Holdings, Inc. (2026, August 14). Lantheus announces FDA approval of TAUKLARIFY (florquinitau F 18 injection), an F18-labeled tau PET imaging agent for Alzheimer’s disease [Press release]. https://lantheusholdings.gcs-web.com/news-releases/news-release-details/lantheus-announces-fda-approval-tauklarifytm-florquinitau-f-18 [^news-lnth-fda-20260814]

* Lantheus Holdings, Inc. (2026, August 26). Preliminary merger proxy statement [Form PREM14A]. U.S. Securities and Exchange Commission. https://www.stocktitan.net/sec-filings/LNTH/prem14a-lantheus-holdings-inc-preliminary-merger-proxy-statement-d07f5a9c0aee.html [^news-lnth-proxy-20260826]

* OpenAI says its Broadcom custom chip is a winner. What does that mean for Nvidia? (2026, August 25). CNBC. https://www.cnbc.com/2026/08/25/openai-says-its-broadcom-chip-is-a-big-advance-what-about-nvidia.html [^news-avgo-openai-20260825]

* Palantir’s Maven is now an official Pentagon program of record. Here’s what guaranteed budget dollars are worth. (2026, August 25). The Motley Fool. https://www.fool.com/investing/2026/08/25/palantirs-maven-is-now-an-official-pentagon-progra/ [^news-pltr-maven-20260825] [^news-pltr-govt-20260825]

* PayPal deal talks end as Advent, Stripe group abandons acquisition effort. (2026, August 28). Bloomberg. https://www.bloomberg.com/news/articles/2026-08-28/advent-stripe-consortium-is-said-to-drop-pursuit-of-paypal [^news-pypl-bid-20260828]

* Salesforce, Inc. (2026, August 26). Salesforce delivers record second quarter fiscal 2027 results [Press release]. https://www.salesforce.com/news/press-releases/2026/08/26/fy27-q2-earnings/ [^news-crm-q2-20260826]

* Salesforce, Inc. (2026, August 26). Salesforce delivers record second quarter fiscal 2027 results [Press release]. Salesforce Investor Relations. https://investor.salesforce.com/news/news-details/2026/Salesforce-Delivers-Record-Second-Quarter-Fiscal-2027-Results/default.aspx [^news-crm-arr-20260826]

* Salesforce and Anthropic expand partnership as Benioff responds to SaaSpocalypse concerns. (2026, August 26). CNBC. https://www.cnbc.com/2026/08/26/salesforce-anthropic-partnership-claudeforce.html [^news-crm-claudeforce-20260826]

* Stock market today, Aug. 27: Salesforce surges 23% on Anthropic partnership and Q2 earnings beat. (2026, August 27). The Motley Fool. https://www.fool.com/coverage/stock-market-today/2026/08/27/stock-market-today-aug-27-salesforce-surges-23-on-anthropic-partnership-and-q2-earnings-beat/ [^news-crm-surge-20260827]

* VA boosts EHR modernization contract with Oracle by $17B. (2026, August 20). Nextgov/FCW. https://www.nextgov.com/modernization/2026/08/va-boosts-ehr-modernization-contract-oracle-17b/415548/ [^news-orcl-va-20260820]

Internal data

Internal data is provided on a best efforts basis.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit <a href="https://telltales.substack.com?utmmedium=podcast&utmcampaign=CTA1">telltales.substack.com