Telltales

by Top Mark Capital
  • 5.0 (6)
  • Investing
  • Updated Semiweekly

An investing podcast + substack for people who want to compound their wealth over the long run and don't mind sailing analogies telltales.substack.com

  1. 12h ago

    "I Don't See How An Investor Can Not Own SpaceX"

    Hunt, Mike, and Jason walk the Cash Flow Memo: oil and gas exhibits, the deficit math nobody wants to do, the Moderna/Merck cancer-vaccine economics, and the SpaceX position Hunt says he’ll probably regret saying out loud. The Cashflow Memo Key Takeaways * Hunt left Exhibit C’s supply/demand untouched: 10-12M bbl/d still clears Hormuz on tankers running with GPS off, near-month crude falls from $87 to roughly $81-82 while the ’27 strip barely moves ($75 to $73-74), so E&P underwriting stays anchored at $70-75 — the price EOG and Magnolia investors are already using. * Gas supply got revised up (’26 dry gas 109.5 Bcf/d, ’27 111.5, with 13 of the 18 Bcf/d added since 2021 coming out of the Permian and Waha back to $2 against a $2.70 Henry Hub), but LNG demand at 20.5 Bcf/d next year narrows the supply-demand gap to about 1 Bcf/d, which is what makes the $3.50 ’27 forecast work. * Exhibit A’s arithmetic leaves one lever: interest expense went from $400B in 2018 to $1T at a 3.5% average rate, defense sits near $1T and Social Security is demographically fixed, so Hunt’s path to a smaller deficit is extending Medicare to everyone and phasing out a $1.2T Medicaid program — and he expects Bessent to have to put a number on deficit reduction in a Monday press conference. * The Moderna cancer-vaccine readout is thinner than the reaction: endpoints met and a statistically significant recurrence reduction, but no hazard ratio and no p-value published, and the economics run to Merck, which shared costs 50/50, takes 50% of vaccine profit plus 100% of Keytruda, roughly 75% of the combined pie against a Keytruda patent cliff in early 2028. * Hunt’s I don’t see how an investor can not own SpaceX: 12M Starlink customers, a permitted 100M sq ft Texas fab aimed at making chips without ASML machines, and Memphis delivering 1.5 GW for $28B — leased out two-thirds to Anthropic and one-third to Google — while data centers stall in every other state; Mike’s counter is that traditional methods give you no margin of safety near $100 (they wanted $80), so size it small and hold your nose ahead of a Tesla/SpaceX merger he expects inside a year. Show Notes [00:29] Oil Markets & Hormuz Ten to twelve million barrels a day still move through Hormuz on tankers with their GPS switched off, which is why crude is drifting down rather than up. Near-month falls from $87 toward $81-82 while the ’27 strip holds near $73-74. [03:43] Natural Gas Production & Demand Dry gas production revised up to 109.5 Bcf/d for ’26 and 111.5 for ’27, with 13 of the last 18 Bcf/d of growth from the Permian. LNG demand at 20.5 Bcf/d next year is what closes the gap and supports $3.50. [05:59] National Debt & Deficit Interest expense went from $400 billion in 2018 to a trillion today at a 3.5% average rate. Hunt walks Exhibit A column by column and lands on Medicare and Medicaid as the only line item with real room. [11:22] Medicare/Medicaid Reform Discussion Jason and Mike stress-test extending Medicare to everyone: fraud and waste as the conservative sell, unwinding ACA underwriting rules, and killing 51 overlapping layers of management. [12:56] Moderna Cancer Vaccine Update The trial met its endpoints, but no hazard ratio and no p-value were published. Merck co-developed it pre-COVID on a 50/50 cost share and takes roughly 75% of the combined profit with Keytruda going off patent in 2028. [16:02] SpaceX Investment Thesis Hunt: I don’t see how an investor can not own SpaceX. Compute scarcity, 12 million Starlink customers, and a 100 million square foot chip facility permitted in Texas. Mike on why the valuation gives you nothing to hold onto. [23:47] NVIDIA / Google Compute Memphis put in 1.5 gigawatts for $28 billion, all NVIDIA equipment, leased two-thirds to Anthropic and one-third to Google. Why is Google leasing NVIDIA capacity when its whole plan was TPUs? [26:20] Meta Litigation & Social Media Rules Forty-nine state attorneys general and Meta appear to be converging on roughly $18 billion, shared with YouTube and TikTok. The agreed rules: two hours a day for children, a midnight-to-6am block, parent-only settings. [29:12] AI Agents & Next Week Agents are the story now, and they’re the middle segment next Wednesday. Get the Cash Flow Memo free at telltales.us — roughly 80 companies with updated financials, plus Exhibit A (federal finances), Exhibit B (natural gas), and Exhibit C (oil). New episode every Wednesday. Cashtags $ASML $GOOGL $META $MGY $MRK $MRNA $MSFT $NVDA $SNAP $SPCX $TSLA $TSM This post and the information herein are intended for informational purposes only. The views expressed herein are the author’s 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 author nor any of his 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 author 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. 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

    "I Don't See How An Investor Can Not Own SpaceX"

  2. 3d ago

    Weekend Update - W2634

    ▶ Explore this week’s Tape — live, sortable, drill-down → Twelve Times Free Cash Flow Bought One Company and Trapped Another Two of our companies signed deals this week, and the Cash Flow Memo prices them almost identically: about twelve times trailing free cash flow, roughly an eight percent free cash flow yield, page fifteen and page six¹²³⁴. One of them got bought at a premium. The other one had to be its own buyer. The multiple decided none of that. Leverage did, and the leaderboard the memo ranks on does not measure leverage at all. Start with what an acquirer is actually buying. Curium agreed to take Lantheus out whole — a hundred and two fifty a share in cash plus contingent value rights worth up to twelve dollars more, call it eight billion dollars, per the merger 8-K, with a close pushed out to the first half of 2027⁵. The agreement landed five days after the FDA approved Tauklarify, the company’s tau imaging agent for Alzheimer’s evaluation⁶. What Curium is buying is a business carrying about one turn of debt against its cash flow⁷. Not a multiple. Cash flow with nobody else’s name on it, which turned out to be the scarce thing this week. Charter ran the same arithmetic from the other side of the table. It closed a thirty-four and a half billion dollar acquisition of Cox and Liberty Broadband, took the largest internet and cable footprint in the country, and then took the acquired company’s name⁸. Same twelve times trailing free cash flow. Ten turns of debt against it⁹, on fifty-four billion dollars of revenue that is going backwards¹⁰. The market prices that equity under four times next year’s earnings¹¹, which is less a verdict on the business than an estimate of who ends up owning the cash it throws off. Ten turns. Charter has been at this altitude before. It filed Chapter 11 in March 2009 and came out that November having shed about eight billion dollars of debt, roughly forty percent of the stack, and some eight hundred and thirty million dollars a year of interest expense, per the company’s own release¹². That trade worked because broadband was still a growth business. Cut the debt, keep the subscribers, let the growth re-lever you back into solvency. The same trade is not available on revenue that shrinks, and the fix on offer this time is a bigger footprint bought with more paper. The California Public Utilities Commission approved the merger on August thirteenth and attached affordable-broadband and network-upgrade conditions to the sign-off¹³. A regulator collecting spending promises from a company whose free cash flow is committed a decade out. Everyone signed. The cashflow read is in Marcus’s column below; short version, the composite ranks cash yield and growth and stops there. Scroll down to the debt-to-free-cash-flow table in The Tape and read it as something other than a risk screen. It is a list of the companies in this memo that will be approached rather than approaching. The ones that fund the next move with equity because the cash is already spoken for. The ones that answer to a lender before they answer to an owner. Last week’s essay argued the AI buildout had turned into a credit trade. This week credit walked into a healthcare name and a cable name priced within a tenth of a turn of each other and decided which one was the asset. What changes the read. Lantheus is still trading under the hundred and two fifty in cash¹⁴, which is what an eighteen-month regulatory close does to a spread — the cash alone is barely above where the stock sat into the weekend, and the full package, contingent value rights paid out, is about fifteen percent above it. The test is the antitrust review of two radiopharmacy businesses combining, and the contingent value rights are where the argument over the Tauklarify ramp actually gets settled. On Charter, the date is mid-September, when the full Spectrum portfolio lands in Cox markets¹⁵. Watch what the integration gets funded with. Out of operating cash flow and the read holds. New paper against the same cash flow and ten turns becomes eleven, into a refinancing that has to underwrite a declining revenue line. Wall Street’s consensus on Charter: the largest broadband network in the country, at that price, has to be the cheapest thing on the board. It is the cheapest thing on the board because the equity stands last in line. Lantheus is this week’s quote on what it costs to stand first. The Tape — W2634 Universe of 94 cashflow-memo names, snap dates 2026-08-15 → 2026-08-21. 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 Salesforce is no longer being priced as software. It is being priced as an annuity, and the dashboard is where that shows up. Going into Wednesday’s print, the memo has CRM at a 7.8% FCF yield, roughly four times the 1.9% median for the tech names in this universe, against NTM revenue growth of 9.4%. That pairing does not belong on a software comp sheet. It belongs on a pipeline. Consensus is still arguing about whether the agent products re-accelerate the top line; the multiple stopped waiting for that answer some time ago and re-rated onto the cash. The test on the 8/26 print is current RPO growth, not the EPS line. If RPO tracks the revenue guide, the annuity read holds. If it steps up, the tape has been pricing the wrong company — call that 30%. 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 89 of 92 ranked-eligible names ranked. 3 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable). Source: cashflow-memo master_2026-08-21.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation. The Issue — This Week's Brief The Cashflow Memo Repriced Without a Print Nothing that repriced the memo this week came out of an earnings report. The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the 86 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 E2635. Chapter markers * Time | Segment * 0:00 | Cold open — four repricings, zero prints * 0:45 | Theme — the legal bill: Meta and Eli Lilly * 4:45 | Deep dive — two deals, same price: Lantheus and Charter * 8:45 | Rapid fire — Harrow, Celsius, Walmart, Tesla, forward calendar * 11:45 | Close — Consensus Watch and the Wednesday tease * 12:45 | Disclaimer Full transcript Cold open Ava: Nothing that repriced this memo this week came out of an earnings report. A California jury put a number on Meta[^news-meta-verdict-20260820]. Eli Lilly agreed to plead guilty and write a check to make a marketing case go away[^news-lly-zyprexa-20260821]. A rival radiopharmacy company agreed to buy one of our healthcare names outright[^news-lnth-curium-20260819], five days after the FDA approved its newest product[^news-lnth-fda-20260814]. And a $34.5 billion cable merger closed, creating the largest internet and cable company in the country[^news-chtr-close-20260820]. Four repricings. Zero earnings prints. Ava: Telltales Weekend Update. I’m Ava Cabot, with Marcus Graham at the cashflow desk. Theme — The legal bill Ava: Meta spent this week finding out what its product design is worth in a courtroom, and the number in the headline is not the number that matters. A California jury found the company liable on 75,000 separate violations of state consumer protection law, and a judge separately found that Meta’s platforms constitute a public nuisance. $375 million in civil penalties[^news-meta-verdict-20260820]. That is the part that is now settled. The open part is bigger. 29 state attorneys general are still in federal trial in California, alleging Meta deliberately designed Facebook and Instagram to be addictive, and they are asking for up to $200 billion in damages plus limits on the algorithmic feed[^news-meta-trial-20260818]. And a former Meta engineering director testified that Mark Zuckerberg ran a culture that treated child safety as secondary to growth[^news-meta-testimony-20260820]. Not a plaintiff’s lawyer. Their own engineering director. Marcus, what does a jury verdict actually cost? Marcus: The penalty isn’t the problem, the finding is. A public-nuisance ruling is a template every other state attorney general can now run without having to win the argument themselves, and it lands on a cash flow that has already been cut in half this year. The memo has Meta at 32x trailing free cash flow[^memo-meta-evfcf-20260821] on $43 billion trailing twelve, and that number is down about 73% from the prior TTM[^memo-meta-fcf-20260821]. Not because the business slowed. Because the buildout ate it. What I’d watch is which attorney general files next, and whether it lands before or after the capex guide. Ava: So the jury did the cheap part first. And the guide didn’t help either. Meta told the Street to expect $61 billion to $64 billion in the third quarter, midpoint $62.5 billion, below where the Street was sitting[^news-meta-q3outlook-20260820]. Meanwhile the company is cutting staff across WhatsApp, Instagram, and Reality Labs in another team r

  3. Aug 19

    This Time, the Hare Beat the Tortoise

    Hunt, Mike, and Jason go more than half healthcare this week: Moderna’s Merck-partnered mRNA cancer vaccine cleared phase 3 in melanoma and the stock doubled. Plus Hormuz at a standstill, $7 trillion of federal spending with only one line left to cut, and xAI putting agents on the shelf. The Cashflow Memo Key Takeaways * Hormuz traffic has gone from 15-20 ships a day to effectively zero and Hunt sees an indefinite stalemate, so with WTI at 82 last Thursday against a 2027 strip at 73, the backwardated 2027 price is the number to underwrite because it is the number operators use for their own capital decisions. * Gas printed under $3 front-month while the 2027 strip holds near $3.40, and with the Permian supplying two-thirds of US supply growth, new Gulf Coast pipelines narrowing Waha to roughly 50 cents under Henry Hub, and LNG feedgas above 18 Bcf/d this year, the 108.5 and 110.5 Bcf/d supply estimates for ’26 and ’27 both look too low. * On Exhibit A’s $7 trillion of FY26 federal spending, healthcare is the only line with real money in it (Medicaid alone is $600 billion, half federally funded), because defense is heading to $1.1-1.2 trillion and interest is locked at a 3.5% average rate with the 10-year at 4.6% - and Jason put Medicare improper payments near 6% with fraud between 3% and north of 10%. * Moderna doubled on a Merck-partnered phase 3 melanoma readout (Keytruda alone versus Keytruda plus a personalized 34-antigen mRNA vaccine) that met all endpoints on a press release with no data attached, validating a decade of platform work rather than a single drug. * The capital-allocation scoreboard flipped: BioNTech was the safer steward and still holds roughly $15 billion of COVID cash to Moderna’s $3 billion as of the March balance sheet, but the hare reached the cancer readout first, and Montana’s read is that the trial proves the technology rather than the drug, which is what makes Moderna’s nine other cancer studies matter. * xAI shipped off-the-shelf Grok agents that cut agent build time from months to about a day, 64% of OpenAI’s tokens are now consumed agentically rather than through human prompts, and that demand curve widens the compute gap further just as Nvidia backstops the physical shell of an OpenAI data center in Ohio and the governors of Pennsylvania, New York, and Texas impose one-year delays or audits on new projects. Show Notes [00:00:30] Exhibit C: Iran, Hormuz, and the Stalemate Traffic through the strait has fallen from 15-20 ships a day to near zero on projectile risk. Hunt sees no path to resolution and argues the 2027 strip at 73, not spot at 82, is the price that should drive investment decisions. [00:04:03] Exhibit B: Natural Gas and the Permian Problem Front-month gas under $3 against a 2027 strip near $3.40. New pipelines have pulled Waha to roughly 50 cents under Henry Hub, and rising associated gas means the published supply estimates are almost certainly too low. [00:05:51] Exhibit A: $7 Trillion and One Place to Cut Long rates are rising worldwide on deficits, not just in the US. Hunt walks the spending columns and finds healthcare is the only category with real savings available, with Medicaid at $600 billion the specific target. [00:10:58] Moderna vs. BioNTech: Page 15 The two mRNA COVID winners diverged on capital allocation - BioNTech kept roughly $15 billion, Moderna ran cash down to $3 billion. Then Moderna announced results and doubled. [00:12:44] The Phase 3 Melanoma Readout Jason lays out the Merck joint development study: stage 3 and 4 melanoma patients, surgical resection, then Keytruda alone versus Keytruda plus an mRNA cancer vaccine. All endpoints met, though no data was released with the press release. [00:13:51] How the Cancer Vaccine Actually Works Montana explains antigen identification via genetic sequencing, the 34 most prominent antigens per tumor, lipid encapsulation, and priming the immune system to catch recurrence before it becomes a tumor. [00:17:48] Nine More Cancers: Does the Platform Travel? Mike asks how investors should handicap the same approach across Moderna’s nine other studies. Montana’s answer: this proves the technology, not a drug - the question is whether the antigen flags stay identifiable. [00:20:07] MRD Testing and the Sequencing Layer Minimal residual disease testing as a monitoring tool, reading cell-free tumor DNA from a blood draw to guide treatment protocols before an end-of-treatment scan. [00:22:06] xAI Ships Agents Off the Shelf Grok agents cut build time from the months Mike and Jason spent to about a day. Jason’s read: AI is just software, and the endpoint is a personal assistant that ships on every phone. [00:25:25] The Token Gap Widens Agentic consumption is now 64% of OpenAI’s tokens versus human prompts. If an agent ships native on every handset, demand steps up again against compute that already cannot serve it. [00:27:43] Nvidia Backstops a Data Center Shell Nvidia is guaranteeing the building, not just the GPUs, on an Ohio site with OpenAI as end customer. Ben Thompson frames it as discounting by taking risk instead of cutting price. [00:28:48] Governors Hit Pause One-year delays in Pennsylvania and New York, an audit requirement in Texas, and a public-sentiment problem that Jensen Huang himself is flagging as a constraint on where the chips can go. Get the Cash Flow Memo at telltales.us and subscribe for a new episode every Wednesday. Cashtags $BNTX $MRK $MRNA $NVDA $PFE This post and the information herein are intended for informational purposes only. The views expressed herein are the author’s 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 author nor any of his 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 author 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. 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

    This Time, the Hare Beat the Tortoise

  4. Aug 16

    Weekend Update - W2633

    ▶ Explore this week’s Tape — live, sortable, drill-down → The AI Buildout Just Became a Credit Trade Broadcom put up the best AI guidance in the semiconductor industry and lost five percent for it. Nobody who read the release thinks that was a demand problem. The buildout has passed the point where equity markets fund it, and what prices the chip names from here is whether somebody else’s lender says yes. Start with the number that did the damage. Bank of America’s Tom Curcuruto put the financing requirement sitting behind the chip buildout at three-hundred-seventy billion dollars, and downgraded on XPU credit risk¹². Not on units. Not on the guide. Fifty-six billion of 2026 AI revenue, up a hundred and eighty percent³, was never in dispute. The bank asked who writes the checks and did not like the answer. Three-hundred-seventy billion dollars. That is the analytical move worth taking seriously, because it re-prices the whole complex. An order book is a claim on somebody else’s balance sheet. When the buyer is Alphabet, the claim is money good and the conversation stays boring. Alphabet’s own filings show AI purchase commitments going from three-hundred-thirty-two billion at the end of the first quarter to eight-hundred-eleven billion by the end of the second⁴, with capex guided to one-ninety-five to two-oh-five billion and free cash flow explicitly taken negative to fund it⁵. Alphabet can carry that. The question was never Alphabet. The question is the tier underneath: the neoclouds, the sovereign projects, the model labs whose GPU orders are contracted against capital they have not raised yet. The last time the equipment cycle outran its customers’ ability to pay, the vendors solved it themselves. Lucent and Nortel financed the competitive carriers buying their gear, booked the revenue, and carried the receivable. When the carriers could not refinance in 2001, the write-downs landed on the vendors’ own income statements. Ugly, and useful, because you could see it coming in the filings. The receivable line moved first. This cycle solved the same problem in a way that removes exactly that tell. The credit is not on the vendor’s books. It is syndicated out to private capital, which is what Nvidia’s five-hundred-billion-dollar third-party financing platform with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR is actually for⁶. Hunt, Jason and Mike took that apart on Wednesday’s show from the supply side. Read it from the credit side and it is the more interesting object: the vendor arranging the loan without holding the loan. Revenue converts to cash on delivery, and the duration risk sits with somebody whose marks nobody publishes quarterly. There is no receivable line to watch, because there is no receivable. Which is why Intel is the honest data point of the week. A company that needed twenty billion dollars to fund AI capacity went and sold common stock at ninety-five a share to get it⁷, upsized from fifteen. That is what it looks like when the credit window is not the cheapest window available to you. The cashflow read is in Marcus’s column below, short version, the memo has the whole complex priced as though the funding is settled. Page one of the Cash Flow Memo has Alphabet at roughly fifty-six billion of trailing free cash flow against a hundred and thirty-two billion of trailing capex⁸⁹. The buildout is already being paid for out of somebody’s balance sheet. The only live question is whose, and at what spread. What changes the read. The disclosure that matters on Broadcom’s next print is customer concentration, not the AI revenue guide. If the ten-K names the XPU customers and they are all investment grade, the BofA cut ages badly. If the concentration line stays vague, the three-hundred-seventy billion is not a modeling assumption, it is a gap. Nvidia reports Wednesday the twenty-sixth¹⁰, and the tell there is the same one: financed demand versus funded demand, and whether management will separate them out loud. Watch private credit spreads on data-center paper alongside the print. Those two things now move the semis together, and the thesis breaks the first quarter they diverge. Wall Street’s consensus on the AI financing question: the demand is real, so the money will be there. The demand was real in 1999 too. It was the money that stopped. The Tape — W2633 Universe of 94 cashflow-memo names, snap dates 2026-08-07 → 2026-08-14. 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 line on this table where two multiples describe two different companies. Forward P/E of 6.5 says the market has already called the top of the memory cycle. EV/FCF of 40.7 says the cash from this cycle has not arrived yet. Both can be true for another quarter or two, and NTM revenue growth of 92.8% is the only thing reconciling them. Consensus reads the 6.5 as cheap. It is not cheap, it is a peak-earnings multiple behaving the way peak-earnings multiples behave, and the composite here is carried by the growth leg, not the cash leg. The test on the next print is whether free cash flow converges toward earnings or the gap holds open on capex. I read it as roughly 60/40 that it closes. 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 90 of 92 ranked-eligible names ranked. 2 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable). Source: cashflow-memo master_2026-08-14.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation. The Issue — This Week's Brief The Cashflow Memo Who’s Paying For It The AI buildout outgrew its own cash flow, and this week the invoices got itemized. The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the 86 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 E2634. Chapter markers * Time | Segment * 0:00 | Cold open — the buildout outran the cash flow * 0:45 | Theme — who’s paying for it: Alphabet, Broadcom, Intel * 4:45 | Deep dive — Apple: Cook hands over the margin * 8:45 | Rapid fire — PayPal, Eli Lilly, Snowflake, Microsoft * 11:45 | Close — Consensus Watch and the forward week * 12:40 | Closing disclaimer Full transcript Cold open Ava: The AI buildout has officially outrun free cash flow. This week three companies showed you how they intend to cover the gap. Alphabet is covering it with commitments. Broadcom is covering it with somebody else’s credit. Intel covered it by printing $20 billion of new stock in a single week. And the largest company in the memo, which is not building any of it, is paying for it anyway, out of gross margin, in the same week it changed CEOs. Somebody always pays. This week the invoices got itemized. Ava: Telltales Weekend Update. I’m Ava Cabot, with Marcus Graham at the cashflow desk. Theme — who’s paying for it Ava: Alphabet on page 1, Broadcom on page 2, Intel on page 3. Three pages of the Cash Flow Memo, one problem. Start with Alphabet, because Alphabet stopped being a cash-generative company this quarter and told you so in writing. AI purchase commitments went from $332.4 billion at the end of the first quarter to $811 billion by the end of the second[^news-googl-commitments-20260814]. In one quarter. Capital expenditure guidance for the full year moved to $195-205 billion, and Alphabet said plainly that this takes free cash flow negative[^news-googl-capex-20260814]. And in the same week the world found out that Berkshire Hathaway spent the quarter buying it, lifting its stake 83% to nearly 106 million shares, about $37.8 billion, now a top-three Berkshire holding[^news-googl-berkshire-20260814][^news-googl-berkshire-stake-20260814]. Marcus, what does Buffett see that the free cash flow line doesn’t? Marcus: Alphabet stopped being a cash machine on purpose, and that’s a defensible thing to do once. The memo has capex running about $132 billion trailing twelve months[^memo-googl-capex-20260814] against roughly $56 billion of trailing free cash flow[^memo-googl-fcf-20260814]. Then they guided capex to $195-205 billion for the year[^news-googl-capex-20260814]. When you spend at that rate free cash flow goes negative, and that’s the cost of the build, not a flag. Which means the multiple isn’t the right frame on this name right now. What actually prices Alphabet is whether that commitment book converts into revenue that clears the depreciation it just bought. I’d hold that read until the December quarter shows an operating margin with the new asset base in it. Ava: There is one more line in Alphabet’s filings worth sitting with. More than 70% of net income last quarter came from investments in other companies, largely Elon Musk’s SpaceX[^news-googl-spacex-income-20260814]. Alphabet disclosed a 7.2% stake in SpaceX Class A stock — 551.2 million shares across Alphabet, XXVI Holdings and Google[^news-googl-spacex-stake-20260814]. The search company’s earnings quality this quarter is a rocket company. Now Broadcom, which had the opposite kind of week: the numbers

  5. Aug 14

    Sold Out Through 2028

    Hunt, Jason, and Mike walk the Cash Flow Memo through $90 oil that will not change oil-patch spending, a compute market sold out through 2028, and a healthcare desk that still found time for Eliquis. The Cashflow Memo Key Takeaways * Hunt’s $70-$90 oil range still holds at the top - crude is ~$90 and he does not see it running much higher - but Iran’s infeasible demands (US bases out, Hormuz cargo tolls, reparations, sanctions off) keep him closer to 90 than 80 for 6-12 months; ~20 ships/day still move (Iraqi crude allowed, Saudis can go Red Sea), producers still underwrite $70-75 so activity does not spike, and $90 oil maps to roughly $4.00-$4.20 US gasoline. * Industrial read-through is Caterpillar: data-center turbines plus reshoring and data-center construction more than offset lagging housing, and Hunt says tariffs, $90 oil, and the political noise are not changing what they hold or add; Mike wants more macro time on these calls now that Warsh is detoxifying Fed guidance. * Compute demand is still exponential: Jassy says AWS AI is sold out through 2026-27 and most of 2028, $200B of CapEx this year still will not close it, and recursive learning models by year-end are another multiple of inference per user - Hunt prefers Amazon, Google, and Microsoft as the compute owners and cannot make up his mind on Meta. * Supply cannot catch that demand in ’27-’28: New York’s >200 MW hold, Virginia saturation, and the Texas governor’s ERCOT audit all push off-grid; combined cycle is 3-4 years out, simple turbines (GE, Cat, Siemens) are sold out, SpaceX Memphis (~$30B, 2 GW, 1-1.5 year payout to Anthropic and Google, 4-5 GW ambition by end-’27) is the exception; Hunt would avoid CoreWeave even after it extended 2020 A100s through 2029 at a 25% price hike. * Chip and memory bottlenecks sit behind the power constraint - TSMC wafers plus packaging, Intel closer on packaging than lithography, memory prices hitting Apple - and Tesla/SpaceX’s Terafab is a first-principles end-run around ASML (particle-accelerator EUV vs tin-droplet lasers) that Huawei is already pursuing; healthcare: FDA closes the GRAS food-ingredient loophole, the childhood vaccine schedule goes 17 to 11 diseases (Japan’s MMR split was reversed after completion collapsed), Thermo Fisher is seeing early-stage equipment rebound, and the AZN-BMY rumor died even as Eliquis (one-third of BMY sales, largest Medicare Part D line) faces a 2028 patent cliff that Jason puts at ~$10B of Part D savings. Show Notes [00:00] Intro & Cash Flow Memo Download the memo at telltales.us; 30 minutes on energy, technology, and healthcare cash flows. [00:27] Iran, Hormuz, and $90 Oil Iran’s demands are not negotiable, so Hunt keeps oil closer to $90 than $80 for 6–12 months. About 20 ships a day still move; Iraqi crude is allowed through, and Saudi barrels can go Red Sea. [04:20] $70 Decisions and Caterpillar Producers still underwrite $70–75, so activity does not spike. Caterpillar is the industrial read-through: data-center turbines and reshoring more than offset weak housing. [07:47] AWS Sold Out Through 2028 Jassy says AWS AI demand is sold out through this year, next year, and most of 2028. $200B of CapEx will not close it, and recursive models by year-end add another multiple of inference per user. [10:37] The Grid Says Build Your Own Power New York’s hold, Virginia saturation, and the Texas governor’s ERCOT audit all push data centers off-grid. Combined cycle is years out; simple turbines are sold out. SpaceX Memphis (~$30B, 2 GW) is the exception. [14:45] A 2020 Chip Gets a 25% Hike CoreWeave extended Nvidia A100 rentals through 2029 at a 25% price increase. Hunt would still avoid the stock and own Amazon, Google, and Microsoft instead. [16:31] TSMC, Memory, and the Terafab TSMC wafers and packaging, plus a memory squeeze that hits Apple, sit behind the power constraint. Tesla and SpaceX are planning a Terafab that replaces ASML’s tin-droplet EUV with a particle accelerator. [22:15] Why It Won’t Run on Your Phone Frontier models need a rack of GPUs and have to run 24/7 even when the phone is offline. On-device AI will route requests, not replace the cloud. [24:11] Huawei and the China Supply Threat Locked out of the latest ASML tools, China is already on particle accelerators and e-beam. If the rest of the world stays chip-constrained, Huawei has a willing market. [26:12] Healthcare: FDA, Vaccines, and the Eliquis Cliff FDA closes the GRAS loophole; the childhood schedule goes 17→11 diseases. Thermo Fisher is seeing early-stage equipment rebound. The AZN–BMY rumor died; Eliquis is one-third of BMY and the largest Medicare Part D line, with a 2028 patent cliff Jason puts at ~$10B of savings. Subscribe for the weekly Cash Flow Memo walkthrough, download the memo at telltales.us, and join us next Wednesday. Cashtags $AAPL $AMZN $ASML $BMY $CAT $CRWV $GOOGL $INTC $META $MSFT $NVDA $SEI $TSLA $TSM This post and the information herein are intended for informational purposes only. The views expressed herein are the author’s 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 author nor any of his 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 author 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. 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

    Sold Out Through 2028

  6. Aug 10

    Weekend Update - W2632

    ▶ Explore this week’s Tape — live, sortable, drill-down → A note from the desk: this week’s update is landing a few days late. We hit a technical problem over the weekend that took our production pipeline down. Everything below is as of Friday’s close, August 7. We’re back on the normal cadence next week. ## The Public Market Just Quoted a Price on a Drug Pipeline. The Price Was Zero. Eighteen billion dollars of biotech changed hands inside seventy-two hours this week, and not one dollar of it was priced by the tape. That is the part worth sitting with. Not the premiums, not the CEO change, not the guidance cut. The fact that when a listed market and a private buyer looked at the same molecules on the same Monday, they came back with numbers that do not live in the same decade. The cleanest evidence is an instrument almost nobody will look at. Curium is paying a hundred-two-fifty a share in cash for Lantheus at closing, plus up to twelve dollars a share in contingent value rights tied to clinical milestones, with the deal closing in the first half of 2027.¹ Lantheus finished the week around a hundred and one.² Read that as the market does: the cash is money-good, and the twelve dollars of clinical optionality is worth approximately nothing. Approximately nothing. That is not a Lantheus fact. It is a quoted, tradeable, sector-wide statement about what the public market will pay for radiopharmaceutical milestones it has not yet seen work, and the answer is that it will not pay. Every discounted-cash-flow argument about pipeline value in this space now has a live market print arguing against it, and the print says zero. Meanwhile the buyers on the other side of that refusal are underwriting the exact same molecules at a discount rate the tape will not touch. Vertex beat the quarter, raised the year to roughly thirteen billion, and wrote a ten-billion-dollar cash check for Crinetics at eighty-five a share in the same week.³⁴⁵ It can do that because a franchise throwing off close to four billion of trailing free cash flow, marked in the Cash Flow Memo around thirty times, funds a decade-long option without going near the debt line.⁶⁷ Curium is private capital reaching the same conclusion through a different funding stack. Both transacted above where the listed market had the asset marked. The cashflow read is in Marcus’s column below; short version, Lantheus screens like a cash machine and now trades like a legal document. And then BioNTech, which is the one that ratifies all of it. Seven times trailing free cash flow, a fourteen percent yield, a market that has effectively stopped underwriting a future at all.⁸⁹ Management’s answer was not to argue. It halved the revenue guide, replaced the co-founder in the CEO seat, said it would close manufacturing sites affecting up to eighteen hundred and sixty jobs, and authorized a billion-dollar buyback.¹⁰¹¹¹² That is a management team agreeing, in cash, with the tape’s refusal to fund its own pipeline. Here is the second-order effect nobody put a number on this week. Every one of these transactions moves the option value of a drug pipeline off a public balance sheet and onto a private or strategic one, at the precise moment the public market has declared that optionality worthless. Public shareholders get the cash and forfeit the decade. That is not a premium — it is a settlement. Crinetics holders get eighty-five dollars today; Vertex gets whatever those molecules are in 2034. Lantheus holders get a hundred-two-fifty and a lottery ticket the market has already voided. The premium is the consolation prize for handing over the part that compounds. What changes the read. Watch the contingent value right itself. It is a listed instrument that will trade between now and the close in the first half of 2027, and it is the only continuous public quote on radiopharmaceutical milestone risk anyone will get. If it develops a real bid, the public market is re-underwriting clinical optionality and this discount closes. If it sits at nothing through the close, the price-setting has moved permanently off the tape, and the next listed rare-disease or radiopharma name to go strategic or private goes at a premium the screens never marked. The other test is BioNTech’s oncology data, which is the only thing that separates a rational retreat at seven times from a market that got it wrong. Wall Street’s consensus on biotech M&A: a premium is a win for shareholders. Ten billion in cash and a contingent value right nobody will bid for suggests the premium is what you get paid to stop owning the decade. The Tape — W2632 Universe of 94 cashflow-memo names, snap dates 2026-07-31 → 2026-08-07. 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 Lantheus ranks in the table above on cash flows that stopped setting its price this week. The composite reads it as a balanced cash generator, 8.2% FCF yield, top-five in the universe, and none of that is what clears the stock now. It is a legal document trading against $102.50 of cash at closing plus up to $12 a share in contingent value rights, and the tape is marking that $12 at approximately nothing. That is not laziness. Radiopharmaceutical milestones are clinical, not commercial, and the public market has never priced that risk well in either direction. The test between now and the first-half-2027 close is whether the CVR develops a real bid. A bid means someone is re-underwriting milestones the seller could not prove standing on its own. 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 85 of 92 ranked-eligible names ranked. 7 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable). Source: cashflow-memo master_2026-08-07.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation. The Issue — This Week's Brief The Cashflow Memo Weekend Update - W2632 Why a Q2 beat bought nothing this week, and $18 billion of biotech changed hands 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 E2633. Chapter markers * Time | Segment * 0:00 | Cold open — a good quarter bought nothing * 0:45 | Theme — the quarter is a receipt (AMD, CVS, Celsius) * 4:45 | Deep dive — page 15: Vertex, Lantheus, BioNTech * 8:45 | Rapid fire — ConocoPhillips, Harrow, and the forward calendar * 11:45 | Close, Consensus Watch, and the Wednesday tease * 12:40 | Closing disclaimer Full transcript Cold open Ava: A good quarter bought you nothing this week. The companies that beat got sold. The one that missed got an activist who wants the CEO’s job. And the biggest checks anyone wrote were for assets that don’t pay off until the 2030s — $18 billion of biotech changed hands inside 72 hours[^news-vrtx-crinetics-20260804][^news-lnth-curium-20260803]. Nobody was paying for the quarter. They were paying for the next five years. So today: two beats that got punished, one miss that turned into a proxy fight, and three companies on the same page of the memo making three incompatible bets on the same decade. Ava: Telltales Weekend Update. I’m Ava Cabot, with Marcus Graham at the cashflow desk. Theme — the quarter is a receipt Ava: AMD delivered about as clean a print as semis produced this quarter, and the market took 7% out of the stock for it. Revenue $11.5 billion, up 52% year-over-year. Gross margin 54%. Net income $2.3 billion, EPS $1.30[^news-amd-q2-20260805]. Then the Q3 guide landed at $13.0 billion at the midpoint, above consensus[^news-amd-q3guide-20260805]. Beat, beat, and beat. Down 7%[^news-amd-stockdrop-20260805]. Marcus — what did they actually get punished for? Marcus: Not the quarter. The price of admission. The memo had AMD at 88x trailing free cash flow going into this print, Q1 10-Q confirmed[^memo-amd-evfcf-20260807], on $8.7 billion of trailing free cash flow[^memo-amd-fcf-20260807]. We re-anchor when the Q2 10-Q files. At 88x you are not buying a beat, you are buying years of uninterrupted acceleration — and Jean Hu just told you the data center step-up is second-half weighted[^news-amd-datacenter-20260805]. Second-half weighted means the proof shows up after the multiple already has to hold. That’s the part that got sold. Ava: And in the same week they went shopping. AMD agreed to buy Taalas, a Canadian startup, to add another category of AI silicon for the data center[^news-amd-taalas-20260806]. Marcus, is that a company that thinks it’s ahead? Marcus: It’s a company hedging its own roadmap, which is the correct thing to do and an uncomfortable thing to watch. Going into this print the memo had AMD running $1.2 billion of capex trailing twelve[^memo-amd-capex-20260807] against under $1 billion of buybacks[^memo-amd-buyback-20260807], Q1 10-Q confirmed — and that was before Taalas. Lisa Su spent the week praising Elon Musk after SpaceX committed to Nvidia exclusively[^news-amd-musk-20260805]. Buying a second accelerator architecture says management does not belie

  7. Aug 5

    900 Gigawatts, 90 Gigawatts of Grid

    Hunt, Mike, and Jason walk the Cash Flow Memo through an oil tape driven by Hormuz headlines, a power grid that has started saying no to data centers, and the valuation question underneath the Tesla-SpaceX merger. The Cashflow Memo Key Takeaways * Oil is trading the Hormuz headline, not the fundamentals: Saudi Aramco earned $33B in Q2 because price more than offset shipped volume, and its CEO says ~2 mmbbl/d of Saudi supply covers customers with or without the Strait open. Hunt’s pattern holds (roughly $90 on missile strikes, high-$70s to $80 when calm, ~$20 of backwardation to the 12-month strip at ~$70-72), and he reads the market as pricing this better than the commentators do. * Natural gas is holding $3.50 on LNG alone (13 Bcf/d in 2024, 16 in 2025, 18+ this year, 20 next) with gas-for-power flat since 2025; the swing factor is data centers forcing on-site turbines, which would restore ~1.5 Bcf/d/yr of power demand and firm gas toward $4. Supply growth is 10 of the last 15 Bcf/d from Permian associated gas, so incremental supply keys off the oil price, not the gas price. * Siting, not chips, is now the binding constraint on the buildout: New York has a one-year hold, Virginia is saturated (Google will not propose more), and the Texas governor just ordered audits (effectively ~12 months) of 900 GW of proposals against ~90 GW of installed state capacity. Combined cycle cannot be built fast enough and turbines are sold out, so on-site generation is the only path. * SpaceX is the episode’s central valuation debate: hosts expect Tesla merged into SpaceX on a trailing-45-day price basis after the China operations are spun out, and treat the equity as a data center business that buys land-free siting. Jason underwrites just south of $90 (~8 GW next year at ~$35B/GW, 20 GW target, 15% IRR), Hunt anchors $50 at roughly half the current price, and Mike flags lockup expiries and launch or regulatory stumbles as the cheaper entry. Launch economics gate the space leg: $1,000/kg makes a gigawatt cost $31B to loft, $150/kg makes it $4.7B, and Starship has flown four times. * Healthcare AI expands capability rather than cutting cost, with one exception. Lilly’s Isomorphic Labs partnership is about a year old with nothing to show yet, and scientists at Lilly, Regeneron, and Pfizer will assault IT budgets for tokens rather than save money. UnitedHealth is the real cost-out: $1.5B of IT spend, one-third to make Optum Insight AI-first and two-thirds to insurance systems, with a pre-auth pilot cutting missing-information denials 68% and appeals nearly 90%. Energy IT departments (Exxon, midstream, EOG) get genuine savings, and the token spend routes through Amazon, Microsoft, and Google because no one gets fired for running a Chinese open-weight model on a hyperscaler. Show Notes [00:00] Intro & Cash Flow Memo Download the memo at telltales.us; 30 minutes on energy, technology, and healthcare cash flows. [00:27] Iran, Hormuz, and $90 Oil Saudi Aramco earned $33B in Q2 as price offset lost volume, and management says ~2 mmbbl/d covers customers either way. Oil runs to roughly $90 on strikes and back to the high-$70s when things calm, with ~$20 of backwardation to the 12-month strip. [03:19] Exhibit B: Gas, LNG, and Permian Supply Gas holds $3.50 on LNG growth from 13 Bcf/d in 2024 to 20 next year, while gas-for-power has been flat since 2025. Ten of the last 15 Bcf/d of supply growth is Permian associated gas, so supply follows the oil price. [05:01] The Grid Says No: New York, Texas, Virginia New York’s one-year hold, Virginia’s saturation, and the Texas governor’s audit letter against 900 GW of proposals versus ~90 GW of state capacity. Combined cycle is too slow and turbines are sold out, so on-site generation wins. [10:40] Tesla into SpaceX: What Is It Worth The hosts expect a trailing-45-day merger after a China spin-out, then split on price: Jason just south of $90 on 8 GW next year at ~$35B/GW and a 15% IRR, Hunt at $50, Mike waiting on lockups and launch risk. Launch cost decides the space leg at $31B versus $4.7B per gigawatt. [18:18] AI in Healthcare: Harrow, Lilly, UnitedHealth Harrow is a commercialization business where AI is not decisive. Lilly’s Isomorphic Labs tie-up is a year old with no results yet. UnitedHealth is spending $1.5B, with a pre-auth pilot cutting missing-information denials 68% and appeals nearly 90%. [22:53] Token Budgets at Pfizer and Regeneron Scientists will consume the IT budget rather than shrink it. Expect more candidates and fewer late failures, not cost savings. [24:20] Energy IT: Exxon, Midstream, EOG Upstream and midstream have run machine learning for years and get real cost savings here. EOG’s decade-old well-file system is the template. [26:12] Open Weights and Why the Hyperscalers Win Open-weight models cut token cost, but IT departments would rather buy from Amazon, Microsoft, and Google than defend a Chinese model to their board. That routes the savings through the hyperscalers. [28:06] Apple’s Siri Problem Apple still has no AI-enabled phone and a software team behind the eight ball. Ecosystem lock-in buys time; it does not buy stagnation. Subscribe for the weekly Cash Flow Memo walkthrough, download the memo at telltales.us, and join us next Wednesday. Cashtags $AAPL $AMZN $EOG $GOOGL $HROW $ILMN $KMI $LLY $MSFT $PFE $REGN $SPCX $TSLA $UNH $XOM This post and the information herein are intended for informational purposes only. The views expressed herein are the author’s 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 author nor any of his 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 author 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. 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

    900 Gigawatts, 90 Gigawatts of Grid

  8. Aug 2

    Weekend Update - W2631

    ▶ Explore this week’s Tape — live, sortable, drill-down → Microsoft Bought Twenty Years of Power. The Market Graded the Buyer. The AI build spent this week being graded on the buyer’s income statement. The most consequential thing that happened to it was signed by a seller. Chevron agreed to supply Microsoft with two-point-six-seven gigawatts of behind-the-meter power at a West Texas data center, for twenty years, disclosed inside Chevron’s own second-quarter release.¹ Twenty years. In a week the market spent deciding whether Microsoft’s spending was disciplined and Meta’s was reckless, one of them committed to a two-decade fixed obligation that lands on neither company’s capex line. Start with what a behind-the-meter contract actually is, because the phrase is doing real work. The power never touches the public grid. It is generated on site and delivered straight into the data center, which means Microsoft is not waiting in an interconnection queue and is not buying at a utility tariff that moves. It is buying a fixed claim on generation through 2046. That is not capital expenditure. It does not show up in the capex line the entire market spent the week staring at, it does not depress free cash flow in the quarter it is signed, and it does not appear on any screen sorting hyperscalers by how fast the revenue is catching the build. Which is the same maneuver, in a different costume, as the fourteen-billion-dollar data center venture Meta struck with BlackRock three days earlier.² Both companies are moving the build off their own cash flow statement. One did it with a joint venture and one did it with an offtake contract, and both were signed by companies whose stocks the market was busy re-rating on precisely the cash flow statements the deals are designed to bypass. The show asked who got paid this week. The answer neither company put in a headline is that the counterparties did. Now look at where those counterparties sit on this week’s Tape. Energy carries the lowest expected forward revenue growth of the ten sectors on the board, barely above zero, and the second-worst median composite score in the universe. The screens price the sector as structurally ex-growth. The screens are reading a strip. The strip does not know that the marginal buyer of new American electricity is now a company with a compute deadline and a balance sheet that makes twenty-year commitments look cheap. Be precise about the size, because the size is not the argument. Chevron earned twelve-point-one billion dollars in the quarter, six dollars and eleven cents a share, up nearly four hundred percent year over year.³ Against that, one power contract is a rounding error, and it will be one for years. What is not a rounding error is the precedent: a supermajor just booked two decades of investment-grade contracted demand for a product the market values at spot. The rest of Chevron stays exactly what it was, priced off the same cyclical molecules, and CEO Eimear Bonner told Bloomberg she expects fuel-making margins to stay high for as long as energy markets remain, quote, under stress.⁴ Under stress. That is a CEO describing the best margin environment she has, in the vocabulary of a hostage. The cashflow read is in Marcus’s column below; short version, the highest-ranked name on this week’s board is an oil and gas producer, and the Cash Flow Memo ranked it before the deal that changes it. What changes the read is whether this contract is a species or a specimen. The forward calendar tests it immediately. Occidental reports Wednesday, ConocoPhillips and Cheniere on Thursday.⁵ Cheniere is the cleanest tell, because long-dated contracted offtake against a spot-priced sector is the entire LNG business model and the market has never paid it for the duration. The test on those calls is whether any management team names a data center or hyperscaler counterparty, or whether power demand stays in the abstract-tailwind register it has occupied for two years. Thesis breaks if the Chevron deal stays a one-off through the fourth quarter. One contract is an anecdote. Four is a repricing. Wall Street’s consensus on energy: a terminal-decline sector, correctly priced for no growth. The largest incremental buyer of American electricity just signed up through 2046. The Tape — W2631 Universe of 94 cashflow-memo names, snap dates 2026-07-26 → 2026-07-31. 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 top of this week’s board is ranking a balance sheet that no longer applies. Magnolia takes the number one composite at 13.0x EV/FCF and a 7.7% FCF yield, both computed off the Q1 10-Q, before the company agreed to buy WildFire Energy for $4.06B and priced a stock offering to help fund it. So the row is accurate and it is stale — a producer that pays for an acquisition partly in equity moves its share count, its debt, and its per-share cash flow in the same week the screen ranked it. Consensus is treating the offering as the news; the offering is the financing. The test is the August 5 print: whether management sizes the combined capex program for 2027, or defers it to the close. 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 89 of 92 ranked-eligible names ranked. 3 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable). Source: cashflow-memo master_2026-07-31.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation. The Issue — This Week's Brief The Cashflow Memo Who’s Getting Paid Microsoft got paid for spending. Meta got billed for it. Same week, same build. The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the 86 companies in the Cash Flow Memo. About 14 minutes. No filler. Download the memo at telltales.us. Mike, Jason, and Hunt are back Wednesday on episode E2632. Chapter markers * Time | Segment * 0:00 | Opening disclaimer * 0:15 | Cold open — cash flow did the grading * 0:45 | Theme — who’s getting paid: Microsoft, Meta, Apple * 4:45 | Deep dive — page fifteen: Vertex and Lantheus * 8:45 | Rapid fire — ASML, Intel, Chevron, and the forward week * 11:45 | Close — Consensus Watch * 12:45 | Closing disclaimer Full transcript Opening 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 before we start: the show is produced entirely with AI tools, and both voices you’re hearing are AI-generated. Send feedback through the Substack. Ava: This was the week the AI build stopped being a guide and started being a cash flow statement. Three of the largest companies on earth reported inside 72 hours, and the market handed out three completely different grades for what looks, from the outside, like the same behavior. On Wednesday’s show, episode 2631, Hunt, Jason, and Mike spent their time on open weights versus closed labs, and where the value goes once model economics commoditize[^ep-e2631]. This week the market answered a much narrower version of that question, with money. Cash flow did the grading. Theme — Who’s getting paid Ava: Microsoft just got a standing ovation for spending money. Azure crossed $100 billion of revenue in a single fiscal year for the first time[^news-msft-azure-20260730]. Fourth-quarter earnings, $4.74 adjusted, up 23%[^news-msft-q4eps-20260729]. The stock jumped 15% on it[^news-msft-stock-20260730]. And buried underneath the applause: Microsoft Cloud gross margin fell to 68%, and the company’s own explanation was the cost of scaling AI infrastructure and the growing usage of AI features[^news-msft-cloud-margin-20260724]. So the build is already in the margin line. On page 1 of the memo, Apple and Microsoft printed inside two days of each other, and only one of them got that reception. Marcus — who actually got paid this week? Marcus: Microsoft got paid on the income statement and billed on the cash flow statement, and the market only graded the first one. Going into this print the memo had them at about 38x trailing free cash flow[^memo-msft-evfcf-20260331], on roughly $76 billion of trailing free cash flow that was down about 21% year over year[^memo-msft-fcf-20260331]. That’s Q3 10-Q confirmed; we re-anchor when the 10-K files. Azure crossing that line is real money. So is the cash disappearing into the build. What I’d watch on the next print is whether the revenue line starts catching the capex line, or whether we’re still calling this a growth story two years from now. Ava: Meta spent the same kind of money and got the opposite grade. Revenue over $60 billion, up 28%[^news-meta-rev-20260730]. Second-quarter free cash flow down 91%[^news-meta-fcf-20260731]. The stock fell 8%, extending a record losing streak[^news-meta-stock-20260731]. And on the call, Mark Zuckerberg said Meta is, quote, getting a lot of offers for compute at a significant premium over what the company paid[^news-meta-compute-2026

See All (300)

Ratings & Reviews

5
out of 5
6 Ratings

About

An investing podcast + substack for people who want to compound their wealth over the long run and don't mind sailing analogies telltales.substack.com

Information

  • Creator
    by Top Mark Capital
  • Years Active
    2021 - 2026
  • Episodes
    300
  • Rating
    Clean
  • Copyright
    © Mike Nicoletti
  • Show Website
    Telltales

You Might Also Like