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. How Do You Get to $2 Trillion?

    4H AGO

    How Do You Get to $2 Trillion?

    Hunt, Jason, and Mike break down the freshly filed SpaceX S-1 and ask the only question that matters: how do you justify a $2 trillion valuation on a company with almost no free cash flow? They work through the AI stack, the Starlink connectivity business, and the launch economics that quietly underwrite all of it. The Cashflow Memo Key Takeaways * SpaceX filed its S-1 targeting a ~$2T valuation against negligible free cash flow; Hunt frames it next to Tesla (~$1.5T on ~$6B FCF) as proof you can pile on valuation with no EBITDA or FCF, set against NVIDIA’s new record ~$163B FCF run-rate and Apple’s ~$120B. * The promotional $22T TAM rests mostly on the least-proven leg — AI (Macrohard agentic workloads, applications not yet invented), a Tesla/SpaceX JV pairing Tesla’s world model with xAI’s language models, plus a ~$55B Terafab chip-production plan. * The Anthropic lease puts a mark on the data centers: xAI leasing Colossus-1 to Anthropic at ~$1.5B/month against a * Launch is the real crown jewel, not Starlink: 122 SpaceX launches vs. 43 customer launches in 2025; re-pricing Starlink at market rate lifts space revenue from $4.1B to ~$11B straight to cash. Customer-launch gross margin is 65-75% and rising as cost/kg falls from Falcon’s ~$850 toward Starship’s ~$100 (NASA: ~$19,000); Starship R&D is $4B this year, up from $3B. * Space-based data centers are an extension of Starlink, not a monolith: each Starlink sat is ~25kW of servers, AI racks run ~125kW in sun-synchronous orbit, launched at daily cadence — a distributed inference network. The choke-point thesis: frontier labs (Anthropic/OpenAI/Gemini) may route inference through Starlink for performance, handing SpaceX negotiating leverage. Starlink itself did $11.4B revenue in 2025 at 39% operating / 63% EBITDA margin across 10.3M subs. Show Notes [00:02] Open & Disclaimer Welcome and the standard informational disclaimer. [00:30] Exhibits A, B & C: Energy and the Government’s Books Hunt on oil and gas pricing through the Iran disruption, weak Waha gas curtailing Permian supply, and a fiscal ’27 federal deficit that stays stuck near $1.5T. [05:51] Macro Grab Bag: Grid Curtailment, Taiwan, and Reshoring DOE clears PJM to curtail data-center power in a grid stress event; the hosts reject Chamath’s nobody cares about Taiwan in 18 months call; Gavin Baker’s point that the Iran war helps US reshoring by raising energy costs more abroad than at home. [09:55] NVIDIA & Apple: Free Cash Flow Records NVIDIA at a ~$163B FCF run-rate (new all-time record, eclipsing old Exxon peak), Apple at ~$120B, against $5.6T and ~$4.5T market caps. [11:34] The SpaceX Question: $2T With No Cash Flow Framing the S-1 alongside Tesla — huge valuations attached to businesses not yet generating EBITDA, income, or free cash flow. [12:51] The AI Stack: xAI, Colossus, the Anthropic Lease, Cursor & Macrohard The $22T TAM and its least-proven leg; xAI’s record build speed; Anthropic leasing Colossus-1 at ~$1.5B/month; the Cursor acqui-hire; Macrohard agentic workloads as a Tesla/SpaceX JV; the $55B Terafab plan. [19:05] Starlink: The Supposed Crown Jewel $11.4B 2025 revenue, 39% operating / 63% EBITDA margin, 10.3M subscribers — and why the hosts think the conventional crown jewel label is misplaced. [19:43] Launch Economics: The Real Crown Jewel 122 SpaceX vs. 43 customer launches; backing Starlink out at market rate to reveal true space economics; 65-75% and rising customer-launch margins; cost/kg from Falcon ~$850 toward Starship ~$100 vs. NASA’s ~$19,000. [22:52] Data Centers in Orbit Why a space data center is a distributed network of ~125kW AI racks in sun-synchronous orbit, not a monolith; the physics of power and heat; latency math vs. terrestrial fiber. [25:54] Q&A: Would You Switch? The Choke-Point Thesis, T-Mobile & Space Junk Whether you’d prefer Starlink inference in 24 months; routing frontier-model inference through Starlink as a negotiating choke point; Starlink V3 + T-Mobile direct-to-cell; Kessler-cascade space-junk risk. [32:18] Next Week Healthcare deep dive, then a future episode on Musk’s TSMC-replacement / Terafab vision and space junk. Subscribe and grab the Cashflow Memo at telltales.us. Cashtags $$SPCX $NVDA $AAPL $GOOGL $TSLA $XOM $MSFT $AMZN $TMUS $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

    34 min
  2. Weekend Update - W2621

    4D AGO

    Weekend Update - W2621

    The Cashflow Memo Capital Structure Week Nvidia confirmed the demand picture. Three companies restructured this week to monetize it. The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the 94 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 E2622. Chapter markers * 0:00 | Opening disclaimer * 0:15 | Cold open + the week’s throughline * 0:45 | Theme — Capital structure week (NextEra, Lantheus, FedEx) * 4:45 | Deep dive — Nvidia Q1 FY27 * 8:45 | Rapid-fire — CRM/SNOW pre-prints, Deere, BioNTech, Target/Walmart, Verizon/T-Mobile, Regeneron * 11:45 | Close + Consensus Watch + forward week * 12:30 | 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: the show is produced entirely with AI tools, and both voices you’re hearing are AI-generated. Send feedback through the Substack. And this is still a pilot — tell us what’s landing and what isn’t. Ava: Here’s the week. Nvidia confirmed the AI demand picture on Tuesday. And three companies in the memo restructured this week to monetize it. NextEra is buying a $67 billion utility for the power. Lantheus is in sale talks at roughly $7 billion for the radiopharma platform. And FedEx is spinning Freight on June 1. On Wednesday’s show, Hunt, Jason, and Mike walked through Google’s real AI risk — not ChatGPT, but agentic search and Gemini Spark[^ep-e2621]. Today’s show is what the rest of the universe did about it. Theme — Capital structure week Ava: Three restructurings in five days, three different time horizons, one read. Page 18, page 15, page 17 of the memo — all printing the same idea. The AI demand picture is now confirmed enough that companies are willing to redraw their balance sheets around it. Ava: NextEra. $67 billion all-stock bid for Dominion Energy[^news-nee-dominion-20260523]. Combined entity becomes the world’s largest regulated utility, and management is explicit about what it’s for — they’re contracted to build 30+ data center campuses, with 15 to 30 gigawatts of generation by 2035[^news-nee-data-center-campuses-20260523]. Meta already has a 2.5 gigawatt solar-and-storage deal signed[^news-nee-meta-partnership-20260523]. Dominion stock up 9% on the announcement. NextEra down 4%[^news-nee-dominion-20260523]. Ava: Lantheus. In talks to sell to Curium at roughly $7 billion — broke Thursday[^news-lnth-curium-20260523]. This is the radiopharma roll-up everyone in oncology imaging has been waiting for. Q1 beat, PYLARIFY TruVu cleared FDA in March with a 50% batch-size lift, and the LNTH-2501 PDUFA lands June 29[^news-lnth-q1-20260523][^news-lnth-pylarify-20260523][^news-lnth-pdufa-20260523]. Ava: FedEx. Freight spins June 1 as FDXF[^news-fdx-spinoff-20260523]. Dual-market trading starts Tuesday. FedEx retains a 19.9% stake; the rest goes to holders, tax-free for U.S. federal purposes[^news-fdx-dual-market-20260523][^news-fdx-tax-20260523]. CEO Raj Subramaniam separately dismissed the Amazon-logistics-threat narrative this week[^news-fdx-amazon-20260523]. Marcus — the cashflow take. Start with the one that’s actually changing right now. Marcus: NextEra is the one that matters this week. The memo had them at 28x trailing free cash flow at a 5% yield going in, Q1 10-Q confirmed[^memo-nee-evfcf-20260522]. That’s a clean number for a regulated utility. But the load-bearing line was already debt-to-FCF at 11x trailing[^memo-nee-debtfcf-20260522]. Now they’re eating Dominion’s leverage in an all-stock deal. The trade is: investors get the regulated-utility tail on AI infrastructure that hyperscaler multiples don’t price, and in exchange they take on a balance sheet that will look heavier before it earns through. What to watch on the next print is whether the contracted gigawatt backlog converts fast enough to absorb the debt the deal piles on. Ava: Translation: you bought the utility because the data centers needed the power, not the chips. Lantheus? Marcus: Lantheus is the cleanest balance sheet of the three. The memo had them at 35x trailing free cash flow at a roughly 3% yield, debt-to-FCF basically zero[^memo-lnth-evfcf-20260522][^memo-lnth-debtfcf-20260522]. $7 billion is a reasonable mark on a company with a Q1 beat, a fresh FDA approval, and a PDUFA five weeks out. The radiopharma platform is what Curium is buying — the imaging stack plus the therapeutic pipeline. Not financial engineering. Strategic consolidation in a category where the FDA pipeline is the asset. Ava: And FedEx is the third one — different structure entirely. Marcus: FedEx is the most interesting capital structure of the three. The memo had FDX at 22x trailing free cash flow at a roughly 6% yield, debt-to-FCF at 7.5x[^memo-fdx-evfcf-20260522][^memo-fdx-debtfcf-20260522]. The Freight spin lets the parent re-rate around the express business; the retained stake gives the holdco a forward monetization option. That’s not a tax dodge — that’s management taking the discount the market puts on the bundle and letting it trade separately. Ava: Three balance-sheet decisions, made the same week Nvidia gave you the demand picture they’re all pricing against. Mark that. Deep dive — Nvidia Ava: Nvidia’s Q1 fiscal 2027 print, after the close Tuesday. The bull case got everything it asked for. The bear case got nothing it asked for. Ava: Revenue $82 billion, up 85% year-over-year and 20% sequentially[^news-nvda-q1-rev-20260523]. Data Center alone was $75 billion — nearly double the prior-year quarter[^news-nvda-data-center-20260523]. Gross margin held at 75%, essentially flat to Q4[^news-nvda-gm-20260523]. Diluted GAAP EPS $1.87, up 140%[^news-nvda-eps-20260523]. Ava: Then they guided. Q2 revenue $91 billion, plus-or-minus 2%[^news-nvda-q2-guide-20260523]. Margin guide held at 75%[^news-nvda-margin-guide-20260523]. Blackwell 300 and the B200 line sold out through mid-2026 per management[^news-nvda-blackwell-demand-20260523]. Rubin platform confirmed for Q3 launch this year, Rubin Ultra in H2 2027[^news-nvda-rubin-20260523]. Ava: And then the capital return. They raised the dividend 25-fold — from $0.01 to $0.25 per share — and authorized an additional $80 billion of buybacks[^news-nvda-capital-allocation-20260523]. Marcus, the cashflow take. Marcus: Nvidia just gave you the next twelve months of justification in one forward number. The memo had them at 50x trailing free cash flow at about a 2% yield going in, Q4 FY26 10-K confirmed[^memo-nvda-evfcf-20260522]. We re-anchor when the Q1 10-Q files. The $91 billion Q2 guide is what changes the read[^news-nvda-q2-guide-20260523] — that’s a single quarter of revenue close to the company’s entire trailing-twelve free cash flow base[^memo-nvda-fcf-20260522]. The multiple was never the problem here. The problem was always whether the Q2 guide would hold the rate of change. It did. Ava: One sentence on why the dividend matters. Marcus: It signals that Jensen Huang now believes the cash generation is structural, not cyclical. You don’t 25x the dividend on a company you think is at the top. The $80 billion buyback authorization is the second signal — they’re going to be in the open market accumulating their own equity while the next product cycle ramps. The question for the next print isn’t whether the demand is real. The question is whether anything in the Blackwell-to-Rubin transition slips, because at this multiple, any timing miss is the entire risk. Ava: And the consensus narrative on the print? Marcus: Wall Street had a version of the law of large numbers eats Nvidia by 2027. The Q2 $91 billion guide just told you the law of large numbers gets eaten first. Bear modelers said this rate of change couldn’t continue at this base. They were wrong, and they’re going to be wrong again next quarter unless something physical breaks in the supply chain. Ava: So the bear case now has to argue physics, not math. Two prints from now, mark the calendar. Rapid-fire Ava: Five forward-week catalysts and one governance shock to close. Buckle up. Ava: Page 2 of the memo — Salesforce and Snowflake both report after the close Tuesday[^earn-crm][^earn-snow]. Consensus on Salesforce: $3.12 EPS, $11 billion revenue[^earn-crm]. Consensus on Snowflake: $0.32, $1.3 billion[^earn-snow]. The Salesforce setup has CEO Marc Benioff committing $300 million of Anthropic token spend for the year, with AI coding agents delivering 30% engineering productivity gains and no incremental engineering hires[^news-crm-benioff-anthropic-20260523]. And per Talnexis hiring data, Salesforce’s AI/ML postings spiked 5.6x in the last 7 days — 28 new roles versus 5 the week prior — heading straight into the print[^tlnx-crm-aiml-20260523]. Memo had Salesforce at 28x trailing free cash flow going in, Q4 10-K confirmed[^memo-crm-evfcf-20260522]. Re-anchor Wednesday morning. Ava: Snowflake is the harder one. Memo can’t anchor on a multiple — trailing free cash flow runs negative, capex still scaling against the AI workload ramp[^memo-snow-fcf-20260522]. What prices Snowflake right now is Cortex AI adoption — 9,100 customer accounts, 200%+ YoY AI-workload growth, NRR holding at 125%, RPO accelerating 42% year-over-year[^news-snow-cortex-20260523][^news-snow-nrr-20260523]. And the Talnexis hiring tracker shows Snowflake’s Partnerships postings up 5.5x in 7 days — 11 new partnerships roles versus 2 the week prior[^tlnx-snow-partnerships-20260523]. The pre-print read: the ecosystem-monetization push is hiring lik

    13 min
  3. Google's Real AI Risk Isn't Just ChatGPT

    MAY 20

    Google's Real AI Risk Isn't Just ChatGPT

    Hunt Lawrence, Mike Nicoletti, and Jason Wallace unpack why the Hormuz panic doesn’t hold, where Google’s real AI risk actually lives, and how the PBM business model is unwinding in real time. Get the Cash Flow Memo at telltales.us. The Cashflow Memo Key Takeaways * Hunt’s oil base case holds at $90 (Brent $108 / WTI $104 today) against consensus $150 calls: Saudi Aramco already posted higher March cash flow routing crude to the Red Sea, ADNOC is twinning the Oman→Fujairah line, Iraq/Kuwait are moving barrels by truck-and-pipe through Syria, and Iran loses leverage over time even without a nuclear deal. * Exhibit A is straining on interest expense (10Y at 4.5% vs. the 3.5% baseline assumption); Mike’s debt/GDP-stabilization-near-100% bet leans on Claude-class AI compressing Medicare/Medicaid spend into flat-to-declining, with defense and interest as the other binding lines. * Google ran from $162 to $400 in 52 weeks: AI Overviews defused the visible ChatGPT threat, but the real risk is agentic search rewiring monetization (Exa just raised $225M at $2B+ from a16z), and Jason posits a chunk of Google’s incremental search revenue is OpenAI paying for web-index grounding. * Gemini Spark (I/O) is Google playing innovator’s-dilemma offense, a 24/7 personal agent running across Gmail/Calendar/Drive that no entrant can replicate without Google’s existing data perimeter; the Google + Meta + Amazon ad-network moat remains durable enough that OpenAI is retreating to Anthropic-style subscription revenue. * PBM pricing power is unwinding in real time: Trump Rx relaunched with Cost Plus Drug + Amazon Fulfillment backends (drugs at ~25% of copay), UNH/OptumRx moving to a transparent flat-fee model and dropping prior auth on 30% of minor procedures, CVS adding biosimilars, and Lilly’s DTC channel proving out, all setting up a healthcare-investment deep dive in two weeks. Show Notes [00:00] Welcome to Telltales Mike opens the show and points listeners to this week’s Cash Flow Memo at telltales.us. [00:18] Disclaimer Standard disclosure. [00:31] Exhibits A, B, C — Oil, Hormuz, and the Federal Deficit Hunt walks through how Saudi Aramco, ADNOC, Iraq, and Kuwait are routing barrels around Hormuz via Red Sea ports, the Fujairah pipeline, and Syrian truck-and-pipe corridors. Why consensus $150 oil is wrong and Hunt’s $90 base case holds. Closes on interest expense and Medicare/Medicaid as the binding lines on Exhibit A. [06:52] More than Moats: Google Hunt frames Alphabet as the latest More than Moats target after Lilly, Nvidia, Goldman, and Microsoft. Mike and Jason work through the antitrust outcome (Chrome retained, web-index data opened to competitors), AI Overviews defending low-intent queries, and the real risk: agentic search and Exa’s $225M raise. [13:04] Google I/O and Gemini Spark Jason walks through I/O announcements including the new content-credentialing system and Gemini Spark, Google’s always-on personal agent running across Gmail, Calendar, and Drive. Why no entrant can replicate this without Google’s existing data perimeter. [15:34] The Advertising Moat Hunt frames Google + Meta + Amazon as a durable ad-network oligopoly. Why OpenAI’s billion-user advertising thesis is failing and the pivot back to Anthropic-style subscription revenue. [19:53] Healthcare: PBMs, Trump Rx, and Lilly DTC Jason walks through Bill Cassidy’s primary loss, Trump Rx’s relaunch on Cost Plus Drug and Amazon Fulfillment rails, CVS adding biosimilars, OptumRx moving to a transparent flat-fee PBM model, UnitedHealth dropping prior auth on 30% of minor procedures, and Eli Lilly’s working DTC channel. [23:50] Fixing the Premium Side Hunt asks how to bring the same rationalization to monthly health insurance premiums. Mike on diagnostic-monitoring opt-in plans with discounted premiums; Hunt on quarterly rebate structures that reward healthier behavior. Why emergency care is the structural hard problem. [29:33] Healthcare Investment Ideas — Two-Week Prep Hunt commits the team to identifying three or four entities running rational healthcare models that could be good investments. Surprise topic next Wednesday; healthcare deep dive in two weeks. [30:38] Sign-Off Stay healthy, back next Wednesday. Subscribe wherever you listen, and grab the Cash Flow Memo at telltales.us. Cashtags $GOOGL $AMZN $AAPL $MSFT $NVDA $META $CVS $LLY $UNH $GS 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

    32 min
  4. Weekend Update - W2620

    MAY 16

    Weekend Update - W2620

    The Cashflow Memo Earnings Week Is a Split Screen — Nvidia, the Big-Box Gauntlet, and the Retail Read-Across The AI trade and the tariff trade get tested on the same three days. By Friday, the market knows which side delivered. 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 E2621. Chapter markers * Time | Segment * 0:00 | Opening disclaimer * 0:15 | Cold open & this week’s split screen * 0:45 | Theme — The Big-Box Gauntlet (HD, LOW, TGT, WMT) * 4:45 | Deep dive — Nvidia going into Q1 FY27 * 8:45 | Rapid-fire — EQT, UnitedHealth, Microsoft * 11:45 | Close & Consensus Watch * 12:30 | 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: 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 the early run of the show — listener feedback is shaping what we do. Ava: This week is a split screen. Nvidia prints Wednesday after the close. The big-box retailers print Tuesday through Thursday — Home Depot, Lowe’s, Target, Walmart. The AI trade and the tariff trade get tested on the same three days. By Friday, the market knows which side delivered. Ava: On Wednesday’s main show, Hunt, Jason, and Mike walked through Meta as the most profitable AI application ever built — the answer to whether AI capex actually compounds back into the income statement[^ep-e2620]. This week the test moves to the picks-and-shovels side. Nvidia. And the read-across to whether the tariff regime is showing up at the cash register. Theme — The Big-Box Gauntlet Ava: On page 8 of the memo this week — Home Depot Tuesday morning[^earn-hd], Lowe’s and Target Wednesday before the open[^earn-low][^earn-tgt], Walmart Thursday[^earn-wmt]. Four big-box prints, three days, one customer. Ava: The customer is the same — middle America, mortgage-burdened, tariff-exposed. The wound isn’t. Ava: Home Depot is the only one of the four where the stock has already done the work. 25% off the 52-week high going into the print[^hd-performance-20260513]. Reports Tuesday at 9:00 AM ET, $3.42 consensus on $41.6B[^hd-earnings-20260505]. Truist cut its price target from $424 to $394 three days ago[^hd-truist-20260512]. And management already told everyone they will source no more than 10% of products from any single foreign country — that’s the tariff hedge, on the record, before the print[^hd-tariffs-20260512]. Ava: Lowe’s is the rare print where two top-tier analysts disagree on the same number. Reports Wednesday before the open. $2.96 EPS on $23B[^earn-low]. Citi upgraded to Buy on May 12, $285 target — they cited four straight quarters of positive comps[^low-citi-upgrade-20260512]. BofA downgraded to Neutral on May 5, $260 target — they cited housing turnover at multi-decade lows[^low-bofa-downgrade-20260505]. Same company. Same week. Two completely different setups. Ava: Target is the only one of the four printing into a customer base that left. Reports Wednesday before the open — same morning as Lowe’s. Consensus $1.41 on $24.5B[^tgt-earnings-20260515]. Foot traffic at Target stores is down year-over-year for 25 of the last 27 weeks since the January DEI announcement[^tgt-dei-20260515]. The boycott officially ended in March — with no new diversity commitments. And Ulta Beauty is walking out of the partnership in August after five years[^tgt-ulta-20260515]. Ava: Walmart is the only one of the four restructuring while expanding. New CEO John Furner cut 1,000 corporate roles this week[^wmt-restructuring-20260513]. Prints Thursday. $0.65 on $175B. The ad business is up 50% year-over-year and the U.S. e-commerce business posted its first profitable quarter globally[^wmt-ad-growth-20260323][^wmt-ecom-profit-20260215]. The memo can’t anchor Walmart — trailing-twelve free cash flow is negative because of capex[^memo-wmt-fcf-20260515]. Marcus stays off the name. Ava: Marcus, two of these are pricing differently than they look. The cashflow take. Marcus: Target is the wounded one in the gauntlet, and the memo isn’t pricing it as wounded yet. 24x trailing free cash flow on $3B of TTM FCF, 10-K confirmed[^memo-tgt-evfcf-20260515][^memo-tgt-fcf-20260515]. The market is pricing Target like the foot-traffic hole closes on its own — 25 of the last 27 weeks say it doesn’t[^tgt-dei-20260515]. The gross-margin guide Wednesday morning is what tells you which side is closer to right. Marcus: Home Depot is the cleanest test of the four. 24x trailing free cash flow on $16B of TTM FCF, 10-K confirmed[^memo-hd-evfcf-20260515][^memo-hd-fcf-20260515]. That’s not punitive for the share-leader of home improvement. The stock is 25% off the high — most of that move is housing turnover, not Home Depot losing share[^hd-performance-20260513]. The test Tuesday morning is whether the pro-contractor segment is actually offsetting DIY weakness[^hd-contractor-20260224], or whether management has been packaging hope as a thesis. Ava: Two prints, two questions. Whether the foot traffic comes back at Target. Whether the pro contractor is real at Home Depot. Lowe’s settles a disagreement between two analysts. And Walmart has to convince anyone watching that cutting jobs is part of the growth story, not in spite of it. Deep dive — Nvidia Ava: Nvidia. Wednesday after the close. This is the most consequential print of the year. Ava: Consensus is $1.74 EPS on $78B in revenue, plus or minus 2%[^nvda-fy27-guidance-202605][^nvda-earnings-consensus-202605][^earn-nvda]. Blackwell B200 and GB200 are sold out through mid-2026 on a backlog described as, quote, insane — 3.6 million units[^nvda-blackwell-backlog-202605]. Hyperscaler capex for 2026 is guided at $725B, up 77% year over year[^nvda-hyperscaler-capex-202605]. Nvidia takes roughly 90% of the AI accelerator dollar inside that. Ava: Now the China complication. On March 5, Nvidia halted all H200 production for China — about 400,000 units of orders that don’t get filled, roughly $30B of walked-away revenue[^nvda-h200-halt-202605]. Then this week, Jensen Huang rode Air Force One to Beijing. For context — Trump brought 17 CEOs to China; only two got Air Force One seats. Musk and Huang[^nvda-huang-trump-202605]. Huang secured U.S. export approval for H200 sales to 10 Chinese firms — Alibaba, Tencent, ByteDance, JD.com — opening an estimated $50B annual market[^nvda-china-h200-202605]. And then Beijing told its tech companies to pause orders while the government decides on import approval[^nvda-china-h200-pause-202605]. Ava: And one more. The Rubin platform was announced at CES — 5x Blackwell on inference, 3.5x Blackwell on training, 10x reduction in inference token cost[^nvda-rubin-202605][^nvda-rubin-economics-202605]. Production ramps the back half of this year. AWS, Google Cloud, Microsoft, and Oracle are first in line. Ava: Marcus, the cashflow take. Marcus: Nvidia going into Wednesday night is the only mega-cap in the AI stack where the multiple looks reasonable against the cash. 50x trailing free cash flow on $103B of TTM FCF, fiscal year 2026 10-K confirmed[^memo-nvda-evfcf-20260515][^memo-nvda-fcf-20260515]. Free cash flow grew about 80% year over year[^memo-nvda-fcfgrowth-20260515]. That’s the reasonable end of expensive in the AI stack — and reasonable means the math has to keep compounding. The test Wednesday isn’t the print. It’s whether the Q2 guide carries Rubin pricing. Ava: The China story — net positive or net negative for the next twelve months? Marcus: Net positive. And that’s the contrarian read. The H200 halt walked away from roughly $30B in China revenue, which everyone scored as a loss. But Nvidia carries a $95B supply commitment with TSMC[^nvda-tsmc-supply-202605]. That capacity doesn’t sit idle — it reallocates to Vera Rubin. So the trade is: walk away from H200 China at H200 margins, redirect TSMC capacity to the highest-priced product in the lineup. That’s the better margin trade. The export approval and the China pause net to noise — Nvidia keeps the option, the 10 Chinese firms stay in the queue. The downside case is the policy whiplash recurs and the option goes to zero. Probability-weighted, I take the trade. Ava: And what changes the read after Wednesday? Marcus: The demand side is set. Hyperscaler capex guided up 77% this year[^nvda-hyperscaler-capex-202605], Nvidia at the center of the dollar. The variable is Rubin pricing on the Q2 call. The new platform is 5x Blackwell on inference[^nvda-rubin-202605]. If management talks Rubin pricing on Wednesday, the multiple has room. If they don’t, this is as good as the cycle gets — and the next derate comes in the back half. The print is consensus minus surprise. The guide is the trade. Ava: So the question Wednesday night isn’t whether Nvidia beats. It’s whether Rubin shows up in the language. Rapid-fire Ava: Three quick ones, a forward week sweep, and we’re out. Ava: EQT just printed the best quarter in its history. Q1 free cash flow of $1.8B exceeded the company’s full-year 2022 free cash flow[^eqt-fcf-20260513]. Net debt fell below $5.7B[^eqt-debt-20260513]. Fitch upgraded EQT to investment grade BBB and Citi upgraded to Buy on May 13[^eqt-earnings-20260513][^eqt-citi-20260513]. The company is openly marketing itself as the preferred power partner for Appalachian AI data centers through the 2030s[^eqt-datacenters-202604]. The

    12 min
  5. "Senator, We Run Ads": Inside Meta's AI Cash Machine

    MAY 13

    "Senator, We Run Ads": Inside Meta's AI Cash Machine

    This week on Telltales: Hunt, Jason, and Mike pressure-test Meta’s position as an AI cash machine, walk through the Iran-driven oil setup, and unpack the FDA shake-up and a second consecutive Harrow miss. The Cashflow Memo Key Takeaways * Meta is the most profitable AI application ever built: revenue doubled from ~$110B in 2021 to a $220B run rate today, with current growth still ~30% YoY on a $200B base — AI rebuilt the attribution layer Apple’s ATT broke in 2021. * Meta’s capex bet is uniquely uncomfortable: on the last call management told analysts they have no idea what ROIC will be on AI infrastructure spend because, unlike Amazon, Microsoft, Google, and Oracle, Meta isn’t renting the servers — it’s defensive spend against what Zuckerberg called an existential paradigm shift. * Oil setup remains manageable despite the Iran standoff and Strait of Hormuz closure: near-month crude is in the $90s, full-year 2026 futures price in at ~$79, 2027 at ~$73, against a ~2M bbl/day inventory draw absorbable by ~1B bbls of US crude inventory plus China’s reserve. * Trump administration’s $1.5T defense ask for fiscal 2027 won’t fully land, but the equipment refresh is happening — bigger concern is US public debt >100% of GNP with no clear capital-markets trigger date, only inevitability. * Harrow/Vevye second consecutive miss: CVS formulary win pulled high-deductible Q1 patients into the $59 Access-for-All program, forcing rebate outflows of $200–$300 per patient back to the PBM — thesis intact but a 5-year payout, not 2–3 years. Show Notes [00:18] Iran Standoff & Crude Inventory Math Hunt frames the Strait of Hormuz closure and the inventory math: a 2M bbl/day draw is absorbable against ~1B bbls of US crude inventory plus China’s reserve. [04:00] Oil Futures Curve & Gasoline Politics Near-month crude in the $90s, 2026 futures price in at ~$79, 2027 at ~$73 — manageable prices, with a path to ~$3.50 gasoline rather than $4.50. [05:36] Defense Budget & US Deficit Trump’s $1.5T fiscal 2027 ask won’t fully land, but the equipment refresh is happening. The real concern is public debt >100% of GNP and an uncertain capital-markets trigger. [07:12] Meta Deep Dive: The Attention Machine Why Meta is unlike other Mag-7 stories — Facebook and Instagram as discovery-to-purchase advertising machines, not search-intent platforms. [13:04] Meta as the Most Profitable AI Application Ever Built ML in ad ranking since 2007, PyTorch’s origin story, and why Meta has been quietly compounding AI dollars longer than any of the LLM darlings. [16:56] Apple ATT, Revenue Doubling, and the CapEx Question The 2021–2022 flat-revenue year that forced Meta to rebuild attribution with AI — and the open question of ROIC on today’s infrastructure spend. [23:29] News: Amazon Supply Chain Services, OpenAI Ads, Nvidia, Tesla, TSMC Amazon stands up a UPS/FedEx competitor, OpenAI launches a self-service ad platform aimed at Google, Nvidia’s earnings cadence, and Musk’s Terra Fab ambitions. [25:49] FDA Shake-Up: Makary Out, Real-Time Reviews In Marty Makary resigns under pressure after the Replimune denial — but the real story is two new real-time review studies with Amgen and AstraZeneca that could compress drug-approval timelines. [30:26] Harrow Earnings: The PBM Rebate Trap on CVS Formulary Second consecutive miss. Winning CVS formulary placement pulled high-deductible patients into the $59 Access-for-All program, forcing rebate outflows that wiped out Q1 economics. Thesis intact, payout extended. Subscribe and get the Cash Flow Memo at telltales.us — financials on ~80 companies plus the oil, natural gas, and US deficit exhibits referenced in every episode. Cashtags $META $AAPL $AMZN $GOOGL $MSFT $ORCL $NVDA $TSLA $TSM $CVS $LLY $REPL $AMGN $AZN 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

    35 min

  6. MAY 11

    Weekend Update - W2619

    The Cashflow Memo AI Is Finally Paying. The Question This Week Was How The Suppliers Got Paid. Micron’s 2026 memory is sold out, with customers writing prepayment checks for chips that don’t exist. AMD’s Q1 data center revenue was up 57% — but the strength was EPYC CPUs, not the OpenAI GPU deal, which contributed zero dollars to the quarter. Palantir printed an 85% revenue quarter at a Rule of 40 score of 145. And Disney’s first full quarter under D’Amaro printed streaming income up 88%. 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 2620. Chapter markers * Time | Segment * 0:00 | Opening disclaimer * 0:15 | Cold open — throughline + E2619 callback + Mother’s Day * 1:30 | Theme — How the suppliers got paid (MU vs AMD compare/contrast) * 5:30 | Deep dive — Palantir * 9:30 | Rapid-fire — CRCL / DIS / UBER / ABNB + forward-week earnings * 13:00 | Close — Consensus Watch + Wednesday tease * 13:30 | 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: the show is produced entirely with AI tools, and both voices you’re hearing are AI-generated. Send feedback through the Substack. Ava: And before we get into it, happy Mother’s Day to all the moms in the audience. Thank you for spending part of today with us. Ava: Here’s the throughline for this week. AI is finally paying — but the unit economics of how suppliers capture that demand vary wildly. On Wednesday’s show, Hunt, Jason, and Mike walked the hyperscaler stack from the bottom up — Hunt made the case that Amazon’s networking edge may matter more than Nvidia’s chips[^ep-e2619]. That was the architecture conversation. Today is the unit-economics conversation, supplier side. Micron printed the cleanest version. Customers writing prepayment checks for memory that doesn’t exist yet[^news-mu-hbm-booked-20260507] — pricing power so strong it’s hitting cash this quarter. AMD printed the messier version. Data center revenue up 57%[^news-amd-datacenter-20260505], but the strength was in EPYC server CPUs and existing Instinct GPU shipments[^news-amd-10q-q1-20260505], not the OpenAI deal everyone is anchoring on. That deal — signed last October[^news-amd-openai-20251006], OpenAI got warrants for up to 10% of AMD[^news-amd-openai-warrant-20251006] — contributed zero dollars to Q1[^news-amd-10q-q1-20260505]. Shipments don’t start until the second half of this year[^news-amd-openai-20251006]. The CPU side of AMD’s print connects back to Intel two weeks ago — same demand signal, different name on the box[^news-intc-q1-20260423][^news-intc-cpu-ai-20260423]. That’s the show. Theme — How the suppliers got paid Ava: Two supplier prints this week, both validating AI demand, very different evidence about pricing power. Micron is charging customers in advance. AMD’s strength came from CPUs, not GPUs — same story Intel told the market two weeks ago[^news-intc-q1-20260423]. Ava: Page 8 — Micron. Micron’s 2026 high-bandwidth memory is fully booked, with customers signing multi-year prepayment agreements for supply that doesn’t exist yet[^news-mu-hbm-booked-20260507]. CEO Sanjay Mehrotra said Micron can only meet between 50 and 67% of demand from key customers[^news-mu-demand-constrained-20260507]. The four largest spenders on AI — Meta, Microsoft, Amazon, and Apple — all publicly cited memory cost and availability constraints. Stock hit an all-time high of $683 on May 7[^news-mu-ath-20260507]. Mizuho took the price target from $545 to $740[^news-mu-mizuho-20260509]. And Micron raised fiscal 2026 capex to $25B from $20B, a 25% step up[^news-mu-capex-raise-20260507]. Ava: And AMD. After the close on May 5 — Q1 revenue $10.3B, data center revenue up 57% to $5.8B[^news-amd-q1-20260505][^news-amd-datacenter-20260505], Q2 guidance $11.2B[^news-amd-q2-guidance-20260505], stock up 16% the next day[^news-amd-stock-20260505]. Important read on that data center number. The growth came from EPYC server CPUs and the existing Instinct GPU ramp[^news-amd-10q-q1-20260505], not from the OpenAI deal. Per the 10-Q, the OpenAI warrants haven’t vested[^news-amd-10q-q1-20260505]. Shipments don’t start until the second half of this year[^news-amd-openai-20251006]. The deal contributed zero dollars to Q1[^news-amd-10q-q1-20260505]. Which connects directly to Intel two weeks ago. Intel printed Q1 revenue up 7% YoY to $13.6B[^news-intc-q1-20260423], with management telling the Street that the CPU-to-GPU ratio in AI deployments is shifting from 1:8 toward 1:4, and trending toward parity as multi-agent workloads scale[^news-intc-cpu-ai-20260423]. That’s the read across both prints. AI demand isn’t just GPUs. Hyperscalers are buying head-node CPUs alongside their accelerator clusters. CPU demand is having a renaissance after two flat years, and AMD just confirmed what Intel told the market. Marcus, the cashflow take. Marcus: The Q1 print is good news about the business AMD already has — EPYC share, current Instinct shipments — not about the business AMD bought with equity last October. The OpenAI warrant is a future customer-acquisition cost that doesn’t hit the P&L until the second half of this year[^news-amd-openai-20251006]. When it does, it shows up in two places. As data center revenue ramping into 2027. And as up to 160 million shares of dilution if MI450 hits its deployment milestones[^news-amd-openai-warrant-20251006]. The memo had AMD at 167× trailing free cash flow on $4.4B of TTM FCF[^memo-amd-evfcf-20260328][^memo-amd-fcf-20260328] — that multiple is being defended by the forward book, not by what just printed. Q1 didn’t move the actual question. The actual question is whether the OpenAI relationship — paid for in October at a penny a share — produces enough revenue to justify the dilution. We don’t get a read on that until late 2026 at the earliest. Marcus: The compare to Micron sharpens the point. Both names trade at multiples that need a story. Micron at 209× free cash flow has the leverage of a structural shortage[^memo-mu-evfcf-20260226][^memo-mu-fcf-20260226] — they’re charging in advance because they can, and the cash is hitting the bank this quarter. AMD at 167× paid customers in equity last October to lock in revenue that hasn’t started yet, while the current quarter prints fine on CPU demand and legacy GPU shipments. Both stocks are pricing forward stories. Only one of those stories has cash hitting the bank right now. Ava: Two suppliers, two prints. Same end-market. Pick your fighter. Deep dive — Palantir Ava: Page 17 — Palantir. The cleanest AI is actually paying print of the cycle so far. After the close on May 4. Q1 revenue $1.63B, up 85% year-over-year — the fastest growth rate since the company went public in 2020[^news-pltr-q1-20260504]. Beat consensus by $90M. U.S. revenue alone hit $1.28B, up 104% — the first time U.S. growth has crossed 100%[^news-pltr-us-growth-20260504]. U.S. commercial revenue up 133%. Net income quadrupled to $870M, EPS $0.34[^news-pltr-netincome-20260504]. Adjusted free cash flow $925M on a 57% margin[^news-pltr-fcf-20260504]. Rule of 40 score: 145[^news-pltr-rule40-20260504]. Quick aside for readers who don’t follow software — Rule of 40 is the standard efficiency benchmark in SaaS. You add a company’s revenue growth and its profit margin. 40 is the bar for a healthy software company. Palantir is at 145. Roughly four times the bar. Back to the print — full-year guidance raised from $7.2B to $7.65B[^news-pltr-fy-guidance-20260504]. Marcus, the cashflow take. Marcus: Palantir is the proof point that the rest of enterprise software has been promising and not delivering. The memo has Palantir at 160× trailing free cash flow on $1.9B of TTM FCF[^memo-pltr-evfcf-20260331][^memo-pltr-fcf-20260331]. We re-anchor when the Q1 10-Q files. The number that actually matters from this print is the U.S. commercial line — that’s a customer set that didn’t exist 18 months ago paying real money for AI deployment, not pilots. Ava: 160× free cash flow. Rule of 40 at 145. Marcus: 160× is not defensible on a static comp. It’s defensible only if you believe the U.S. commercial growth rate doesn’t normalize for another four to six quarters. So weight the outcomes. Bull case at 30% — government and commercial both compound from here. Base case at 50% — commercial decelerates and the multiple compresses by half on a stock that still works. Bear case at 20% — the multiple does what 160× multiples always do. Ava: So the question is whether the customer set that just appeared keeps appearing. Marcus: That’s the entire question. Argus upgraded to Buy on May 7, Citi took the target to $225[^news-pltr-argus-20260507][^news-pltr-citi-20260507]. The Street is pricing the bull. Watch the Q2 print for whether U.S. commercial holds north of 100% year-over-year, or steps down to 70. 70 is still a great number. It is also a different multiple. Ava: 145 Rule of 40. That number alone is the whole story. Rapid-fire Ava: Five to close out — one pre-print catalyst, three prints already on the tape, and the forward week. Ava: First — Circle. Reports Monday morning[^earn-crcl]. Wall Street consensus $717M revenue, $0.18 EPS[^news-crcl-q1-20260510]. The setup matters more than the number. The CLARITY Act compromise cleared Congress, banning passive deposit-style yield on stablecoins while explicit

    15 min
  7. Hyperscaler Stack Decoded (e2619)

    MAY 6

    Hyperscaler Stack Decoded (e2619)

    This week we walk through the post-Hormuz oil setup, the Intel stake math, and a layer-by-layer review of the hyperscalers — Amazon, Microsoft, Google, Oracle, and Meta — as they navigate the AI buildout. The Cashflow Memo Key Takeaways * Iran ceasefire path holds — Project Freedom (defended Hormuz corridor) is suspended in favor of a one-page 14-point memo brokered via Pakistan; Hunt models ’26 oil at ~$80 and ’27 at ~$72-73, with the $25B war cost absorbable inside the existing $900B defense budget. * Intel stake now worth ~$56B against a $9B cost basis — Mike argues the administration should sell to fund war spending; UAE exiting OPEC+ adds another structural shift but limited ’27 price impact. * Hyperscaler stack defined as five layers (infrastructure → platform → model → harness → application); Amazon leads on platform breadth and faces fewer internal-vs-customer capacity conflicts than Microsoft or Google because AWS scale dwarfs internal compute needs. * Networking has become the strategic differentiator — post-Mellanox, Nvidia prioritizes customers buying bundled compute and networking, which de-prioritized Amazon and forced Trainium; Anthropic running on Trainium signals inference workloads splitting from Nvidia-dominated training. * Meta got beat up unfairly on CapEx ROIC questions — the same AI infrastructure that rebuilt ad attribution post-Apple ATT is what analysts criticize; Zuckerberg pushed back that ROIC isn’t the right lens for application-layer businesses where you build user experience first and monetize later. Show Notes [00:00] Cold Open & Disclaimer Mike opens the show. Standard disclaimer. [00:26] Iran, Hormuz, and Project Freedom Hunt’s six minutes on Iran. Project Freedom (defended-corridor through the strait) is suspended for a one-page memo brokered via Pakistan. Modeling ’26 oil at ~$80 and ’27 at ~$72-73. [06:56] The $56 Billion Intel Stake The Trump administration’s 10% Intel position bought for $9B is now worth ~$56B. Mike’s view: sell it. Plus JP Morgan’s bearish 50s oil call and UAE leaving OPEC+. [09:27] Hyperscaling: Who’s In and How They Got Here Setting up the hyperscaler review. Hunt on Amazon’s accidental cloud head start and the AWS vs. Azure share question. [11:45] The Five Layers of the Stack Mike defines the stack: infrastructure → platform → model → harness → application. Where CoreWeave fits, where Bedrock fits, why platform-as-a-service is the moat. [16:06] Amazon, Trainium, and the Nvidia Networking Story Why Amazon was slow on Nvidia, what changed after Mellanox, and how Anthropic running on Trainium signals the training-vs-inference split. [18:26] Google and Microsoft’s Capacity Conflict Both face the same problem: allocate compute to customers or to internal AI products. Google opening up TPUs. Microsoft’s Copilot and GitHub agent-driven usage explosion. [21:52] Meta’s Already-Built AI Infrastructure Meta rebuilt ad attribution after Apple’s ATT changes by pouring CapEx into AI infra — the spending analysts now punish them for is what powers the ad-business growth. [23:10] Earnings Calls and the ROIC Debate Every analyst question this earnings season was about return on invested capital. Zuckerberg’s pushback: that’s not how application-layer businesses are built. [24:45] The Five-Year CapEx Amortization Question Hunt presses on whether $700M-per-build economics actually pencil with five-year chip obsolescence. Jason on AI displacing labor, Mike on Rentahuman.ai. [27:28] Healthcare Cost Disruption Off-patent generics, GP-as-AI-agent, Amazon One Medical, and the Obamacare 15% profit cap that incentivizes insurers to grow costs. [32:09] Closing & Next Week Next episode: Meta and its digital advertising competitors, plus more on flatlining federal healthcare spending. Subscribe and download this week’s Cash Flow Memo at telltales.us. Cashtags $AMZN $MSFT $GOOG $GOOGL $META $NVDA $ORCL $CRWV $INTC $AAPL $JPM 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

    34 min
  8. Weekend Update — W2618

    MAY 2

    Weekend Update — W2618

    The Cashflow Memo Apple, Meta, Microsoft, Alphabet, and Amazon all printed in 48 hours. Same capex cycle, three different cloud margin stories — and Apple and Meta both quietly repositioned the balance sheet. 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. Mike, Jason, and Hunt are back Wednesday on episode 2619. Chapter markers * Time | Segment * 0:00 | Opening disclaimer * 0:15 | Cold open * 0:45 | Theme — capital cycle resets (AAPL, META) * 4:45 | Deep dive — hyperscaler cloud face-off (AMZN, MSFT, GOOGL) * 8:45 | Rapid-fire (OXY, DIS, PLTR, CVS, MCD) * 11:45 | Close * 12:30 | 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 analyst on the show. Every beat you hear from me is anchored in a number from the Cash Flow Memo. If the multiple doesn’t make sense against the cash the company actually generates, I’ll say so. Ava: Quick note: the show is produced entirely with AI tools, and both voices you’re hearing are AI-generated. We’re still in the pilot window, so feedback is welcome through the Substack at telltales.us. Ava: This was the heaviest hyperscaler print week of the year. Microsoft, Alphabet, and Amazon all reported within a 36-hour window. Apple printed Thursday after the close. Meta printed Wednesday. Roughly $9 trillion of market cap reset its capex framing in the same 48 hours. On Wednesday’s episode 2618, Mike, Jason, and Hunt walked through the big-retail valuation convergence and Hunt’s revised oil thesis, and they teed up a dedicated hyperscaler episode for this week.[^ep-e2618] We did the work for them. The theme first. Theme — capital cycle resets Ava: Two prints this week did something the headline numbers don’t capture. Apple and Meta sit on opposite ends of the universe — different businesses, different buyers, different growth profiles. Both used this print to reposition the balance sheet for an AI capex cycle bigger than what’s been committed to publicly. That’s the theme. Ava: Apple, page 1 of the memo, after the close Thursday. Q2 revenue ~$111B, up 17% YoY, beating consensus.[^news-aapl-q2-print-20260430] Gross margin 49.3% — an all-time record.[^news-aapl-margin-record-20260430] iPhone up 22%. Greater China up 28% on top of December’s 38%.[^news-aapl-china-beat-20260430] Apple authorized an additional $100B in buybacks and raised the dividend 4%.[^news-aapl-buyback-dividend-20260501] Incoming CEO John Ternus made his first public appearance on the call, taking the chair September 1.[^news-aapl-ternus-debut-20260430] But three structural changes hit at once. Apple guided June gross margin to 47.5–48.5% — the first explicit step-down in 8 quarters.[^news-aapl-guidance-strong-20260501] CFO Parekh said Apple is no longer providing net cash neutral as a formal target.[^news-aapl-cash-framework-20260430] And the buyback was halved during the transition window — 93M shares in Q1 to 42M in Q2.[^news-aapl-buyback-halved-20260430] Marcus, the cashflow read. Marcus: Going into this print, the memo had Apple at 37x trailing free cash flow at a 2.6% yield, on ~$106B of trailing free cash flow.[^memo-aapl-priorqtr-20260424] Buyback over that window, ~$92B.[^memo-aapl-buyback-priorqtr-20260424] Those are Q1 10-Q confirmed; we re-anchor when the Q2 10-Q files later this month. The operating print is the best in Apple’s history — that’s not the question. The capital structure read is what matters here. Net cash went from $54B to $62B despite $15B of returns this quarter. The most plausible read on retiring the net-cash-neutral framework is Apple is creating room to issue debt against an AI capex step-up without an offsetting drawdown obligation. Halving the buyback during the transition window gives Ternus operating room to redirect capital differently than Cook. The 37x multiple is being asked to hold at peak gross margin right before a confirmed step-down. The memory cost language has now escalated three calls running — minimal, then a bit more, now significantly higher in June with an increasing impact beyond June. Action: hold; do not add at this price. Ava: Meta, page 1. Q1 revenue ~$56B, up 33% YoY, beating consensus.[^news-meta-q1-earnings-20260429] Meta raised 2026 capex guidance to $125–145B, $10B higher at both ends of the range.[^news-meta-capex-raise-20260429] JPMorgan downgraded the stock to neutral from overweight; the stock fell about 9% after-hours.[^news-meta-jpmorgan-downgrade-20260430] Mark Zuckerberg announced 8,000 layoffs effective immediately, citing AI infrastructure costs.[^news-meta-layoffs-20260501] Marcus. Marcus: Meta is mid-AI-capex cycle. Capex was running ~$70B trailing twelve going into this print, and the company just guided this year up to $125–145B.[^memo-meta-capex-priorqtr-20260424] When you spend that much faster than you generate operating cash, free cash flow goes negative — that’s the cost of the build, not a flag. So the multiple isn’t the right frame on this name. What actually prices Meta right now is the revenue acceleration. Price-per-ad doubled growth rate Q-over-Q from +6% to +12% — single most important number on the print.[^news-meta-price-per-ad-20260429] The bigger number behind the capex raise is what Susan flagged separately on the call — Meta booked $107B of new contractual commitments in this quarter alone. A single-quarter increment larger than Meta’s entire 2026 capex range.[^news-meta-contractual-commitments-20260429] Off-balance-sheet commitment is now the figure to track. Same pattern as Apple — the headline is operating, the actionable is capital structure. Deep dive — hyperscaler cloud face-off Ava: All three hyperscalers reported within 36 hours. All three guided capex up. All three said AI demand exceeds supply. The margin print is wildly different — and the divergence tells you where each cloud sits on the AI maturity curve. Ava: Amazon, page 1. AWS revenue ~$38B, up 28% YoY — the fastest AWS growth in 15 quarters.[^news-amzn-aws-q1-20260429] Total Q1 revenue beat at ~$182B against consensus of ~$177B.[^news-amzn-q1-earnings-20260429] AWS operating margin stepped from 35.0% in Q4 to 37.8% in Q1 — +280bp sequentially, on $14B of segment operating income.[^news-amzn-aws-margin-20260429] Custom silicon run rate $20B; Trainium-specific revenue commitments now $225B, the first ever disclosure.[^news-amzn-trainium-20260429] Andy Jassy said selling Trainium racks externally is a good chance over the next couple of years.[^news-amzn-trainium-20260429] Q2 revenue guide $194–199B, well above estimates.[^news-amzn-q2-guidance-20260429] Marcus. Marcus: AWS is the cleanest single-number story of the week. The +280bp sequential step-up in segment margin lands directly inside Andy Jassy’s several hundred basis points claim from the shareholder letter. Management did not attribute the lift explicitly to Trainium — leaving headroom on later calls. Going into the print, the memo had Amazon at ~63x trailing free cash flow at a ~1.5% yield, on ~$43B of trailing free cash flow. That’s Q4 10-K confirmed.[^memo-amzn-priorqtr-20260424] The Q1 print just absorbed even more operating cash into capex — Amazon’s free cash flow collapsed ~95% on a TTM basis.[^news-amzn-fcf-20260429] So the consolidated multiple is even less informative; we re-anchor when the Q1 10-Q files. The anchor is the segment margin, not the consolidated multiple. Action: the obvious sub-$200 entry window has closed; hold or add modestly into the Q2 print, where the test is whether 37.8% holds or compounds. Ava: Microsoft, page 1. Q3 revenue ~$83B, up 18% YoY. Azure growth +40% cc.[^news-msft-q3-earnings-20260429] Microsoft guided 2026 capital spending to $190B, well above Wall Street estimates.[^news-msft-capex-190b-20260429] Microsoft 365 Copilot crossed 20M paid enterprise users — up 250% YoY, with Accenture taking 740K seats in the largest-ever single deal.[^news-msft-copilot-20m-users-20260430] And the OpenAI partnership was restructured. The exclusive Azure arrangement ended; OpenAI can now deploy on competing clouds. Microsoft retains royalty-free IP through 2032 and a capped revenue share through 2030.[^news-msft-openai-nonexclusive-20260427][^news-msft-openai-revised-terms-20260427] Stock fell ~6% post-earnings.[^news-msft-stock-decline-earnings-20260430] Marcus. Marcus: Going into this print, the memo had Microsoft at ~97x trailing free cash flow at a 1% yield, on ~$32B of trailing-twelve free cash flow. That’s Q2 10-Q confirmed.[^memo-msft-priorqtr-20260424] Q3 pushes the multiple higher and the yield lower — peak multiple on peak narrative. We re-anchor when the Q3 10-Q files. The Copilot inflection at 20M seats with Outlook-level usage intensity is the single most important data point on the print. The buried quality concern — all-in commercial bookings −6% cc, +7% ex-OpenAI.[^news-msft-bookings-20260429] Mostly comp lap and OpenAI restructure timing, not core enterprise weakness, but the kind of number that would headline a different stock. The OpenAI restructure is the swap of exclusivity premium for predictability — Microsoft now collects a capped annuity through 2030 off a counterparty that just recapped. Lower upside, higher floor. Action: hold; do not add unless Maia and Cobalt margin compounds faster than guided. Ava: Alphabet, page 1. Q1 revenue ~$110B, up 22% YoY.[^news-googl-q1-earnings-20260429] Google Cloud revenue $20B, up 63% — the first time crossing the $20B threshold and the stron

    18 min
See All (273)

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
    273
  • Rating
    Clean
  • Copyright
    © Mike Nicoletti
  • Show Website
    Telltales

You Might Also Like