Telltales

by Top Mark Capital

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. 2d ago

    Weekend Update - W2626

    ▶ Explore this week’s Tape — live, sortable, drill-down → The Week Memory Stopped Being a Commodity For forty years, memory was the worst business in technology. Brutally cyclical, structurally commoditized, a graveyard of balance sheets that overbuilt into every upcycle and got buried in the glut that followed. You did not own memory. You rented it, for one cycle, and you got out before the supply caught up. This week Micron told you that business is over — and the people still pricing it as a cycle are fighting the last war. Start with the number that actually matters, and it is not the print. Yes, Micron guided next quarter to fifty billion dollars against a Street modeling forty-three, and yes, the market cap crossed a trillion.¹ Spikes like that are exactly what the cyclical bears are built to fade. The thing that should stop you is buried below the headline: sixteen legally binding take-or-pay contracts, locking in roughly a fifth of Micron’s DRAM capacity through the end of the decade.² Take-or-pay is the language of pipelines and LNG terminals, not chips. It is what a supplier signs when the buyer is more afraid of not having the product than of overpaying for it. That sentence has never been true about memory before. This is a capital-cycle story, and the cleanest analog is crude. The oil majors spent decades destroying their own returns by spending every dollar of cash flow drilling into the next price spike. Then, somewhere after 2015, the survivors consolidated and discovered discipline — capex restraint, returns over volume, supply that no longer rushed to kill every upcycle. The multiples did not re-rate because demand exploded. They re-rated because the industry stopped overbuilding. Memory now has three players who matter, AI demand that arrives on multi-year contracts instead of a consumer whim, and customers signing away their option to walk. That is not a cycle turning. That is a commodity becoming a toll road. And you can read the toll on everyone downstream. Apple raised prices on fourteen products this week and took its worst single day in over a year, with Tim Cook calling the memory spike a hundred-year flood unlike anything in his forty years.³ A company that prints a hundred-thirty billion in trailing free cash flow does not pass costs to the customer over a blip — it eats them. It passes them when the input is structural, and the same week it is quietly lobbying Washington to buy from a blacklisted Chinese supplier just to get the parts.⁴ Oracle is funding its AI cloud with a twenty-four-billion-dollar free-cash-flow deficit and forty billion more in planned debt, the whole build premised on memory it has to secure years out.⁵ Broadcom and OpenAI unveiled a chip designed to cut inference cost in half — which is what you do when the underlying components got expensive enough to engineer around. Every one of those moves is a payment, in a different currency, to the same toll booth. The cashflow read is in Marcus’s column below — short version: the Cash Flow Memo had Micron going into the print at a hundred-plus times trailing free cash flow, and it was the wrong frame, because the contracts changed what the denominator will look like. What changes the read is the supply side, and the calendar is short. The test is whether Samsung and SK Hynix hold the same discipline or break ranks and flood capacity into these prices — the move that has ended every prior memory upcycle. Watch the fiscal-2027 capex commitments from all three, and watch whether more take-or-pay contracts get signed or this stays a sixteen-deal anomaly. The thesis breaks the moment one of the three decides that share matters more than price. It always has before. Wall Street’s consensus on memory: enjoy the spike, the cycle always rolls over, it always has. Sixteen take-or-pay contracts running to 2030 say the cartel finally learned what the oil majors learned — that the most valuable thing you can do with a commodity is refuse to make too much of it. The Tape — W2626 Universe of 94 cashflow-memo names, snap dates 2026-06-19 → 2026-06-27. 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 isn’t in this table, and that’s the point — the print is too fresh and the fiscal-year filing hasn’t landed, so the memo number is already stale. Going into the quarter the memo had it at 102x trailing FCF, which was the wrong frame even then. The reframe is the contract book: 16 take-or-pay deals locking roughly a fifth of DRAM capacity through 2030. Take-or-pay converts a commodity denominator into something closer to contracted revenue, and the screens won’t reprice that until the filings show it. The test is the next fiscal-year filing — and whether Samsung and SK Hynix sign the same paper or break ranks on price. 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 90 ranked-eligible names ranked. 0 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable). Source: cashflow-memo master_2026-06-27.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation. The Issue — This Week's Brief The Cashflow Memo Memory Hits Escape Velocity The week the memory shortage stopped looking like a cycle, and everything downstream paid the bill. The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the 90-plus 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 E2627. Chapter markers * Time | Segment * 0:15 | Cold open — memory’s escape velocity * 0:45 | Theme — memory’s downstream: Apple and Oracle * 4:45 | Deep dive — Micron and Broadcom * 9:00 | Rapid-fire — Nike, pharma, Meta, Tesla * 12:15 | Close — Consensus Watch + forward week Full transcript Cold open Ava: You’re listening to the Telltales Weekend Update. I’m Ava Cabot. Marcus: And I’m Marcus Graham — the cashflow desk. Ava: Quick note: the show is produced entirely with AI tools, and both voices you’re hearing are AI-generated. Send feedback through the Substack. We’re still in pilot, so tell us what’s working and what isn’t. Ava: Here’s the one thing to take from this week. Memory hit escape velocity. And everything downstream — Apple’s price tags, Oracle’s balance sheet, the chips hyperscalers are now designing just to get out from under the cost — is a consequence of that one fact. On Wednesday’s show, episode 2626, Hunt, Jason, and Mike flagged the memory squeeze forcing Apple to raise prices.[^ep-e2626] Then Micron reported. And the squeeze stopped looking like a cycle and started looking like a supercycle. Theme — memory’s downstream Ava: Start where it hits you at the checkout. Apple just told you the memory shortage has reached the price tag. The company raised prices on 14 products this week — MacBooks up $200, the iPad up to $449, Vision Pro now $3,699.[^aapl-price-hikes-20260625] The stock had its worst day in over a year, down about 6%, roughly $265 billion of market value gone in a session.[^aapl-stock-decline-20260625] And Tim Cook didn’t hedge it. He told the Wall Street Journal memory and storage prices have quadrupled in three quarters, and called it, quote, a hundred-year flood, unlike anything he’s seen in over 40 years.[^aapl-cook-quote-20260625] Ava: And then the quiet part. The same week Apple raised your prices, it was lobbying the Trump administration for permission to buy memory from ChangXin — a Chinese maker that sits on the Pentagon’s blacklist.[^aapl-cxmt-lobby-20260627] That’s how short memory is. And the talent is chasing the same scarcity — Apple’s head of Vision Pro and smart-glasses engineering left this week for OpenAI’s hardware unit, after seven years on the project.[^aapl-meade-openai-20260626] Marcus — Apple can afford to eat this. Why is it passing it on? Marcus: Because the flood is real and Apple is telling you it can’t source its way around it. This is a company that prints cash — the memo had Apple at 35 times trailing free cash flow at a roughly 3% yield going in, on about $130 billion of trailing free cash flow.[^memo-aapl-evfcf-20260627][^memo-aapl-fcf-20260627] A company with that cash machine raises prices on the customer only when the input cost is structural, not a blip. The price hikes aren’t the story. The lobbying to buy from a blacklisted supplier is the story. That’s a hardware company admitting the supply chain it actually wants is the one it’s not allowed to use. Ava: From the checkout to the data center. Oracle is building the AI cloud with borrowed money and a shrinking payroll. The company cut 21,000 jobs over the past year, about 13% of the workforce, while it pours money into infrastructure.[^orcl-layoffs-20260627] Free cash flow for fiscal 2026 swung to a deficit of nearly $24 billion, driven by $55.7 billion in capital spending.[^orcl-fcf-20260627] And it’s guiding next year to $90 billion in revenue while planning to raise another $40 billion in debt.[^orcl-guidance-20260627][^orcl-financing-20260627] Marcus, what

    12 min
  2. Who Owns Your Data Now

    5d ago

    Who Owns Your Data Now

    Mike Nicoletti, Hunt Lawrence, and Jason Wallace dig into the looming Treasury market stress, proprietary-data risk in the age of frontier AI models, and a Medicare for All blueprint that could reshape healthcare investing. A 30-minute tour across energy, technology, and healthcare, anchored by this week’s Cash Flow Memo. The Cashflow Memo Key Takeaways * Hunt puts better-than-even odds on a 2008-style Treasury market dislocation within 12-24 months as ~$2.5T of new government paper collides with the incoming Fed chair’s plan to run the balance sheet from ~$7T down to ~$1.5T, plus the financing wave funding AI data centers (SpaceX alone just did $25B of public debt after its equity raise). * The only fiscal lever large enough to bend the deficit is healthcare, and Hunt floats Medicare for All run by an independent, Fed-style administrator (he names Dr. Oz) that ends the ~$600B/year Medicaid transfer to states; investment read is to avoid new healthcare positions broadly but identify the specific winners of a restructured system. * Apple is running Apple Intelligence inference on Nvidia Blackwell inside Google Cloud (not TPUs) specifically because Nvidia uniquely offers encrypted-data-in-memory, making proprietary-data protection the new competitive axis for frontier-model compute. * On a forced 5-year choice between Nvidia and Alphabet, the hosts lean Nvidia on relative value: ~$160B free cash flow trending toward ~$200B with minimal CapEx (vs. Apple’s $120B and Alphabet’s heavy capital program compressing its FCF), framed as the picks-and-shovels seller in the AI buildout. * The memory-supply shortage is forcing Apple to raise device prices (an estimated +$275 on the iPhone Pro just to hold gross margin); Micron reports tonight, and the squeeze is driving a stack-wide efficiency push (Nvidia cites ~2.5x inference efficiency gains since OpenAI’s 5.5 models). Show Notes [00:26] Oil Holds in the 70s, and a Treasury Warning Hunt sees no change in the oil supply/demand picture, WTI holding around $71, and lower oil prices paradoxically supporting natural gas via reduced Permian growth. He lays out a better-than-50% case for a 2008-style Treasury dislocation in the next 12-24 months, with healthcare as the only spending lever big enough to matter. [05:23] Medicare for All and Fixing Sick Care The hosts debate whether the U.S. can shift from triaging illness to keeping people healthy, why insurers’ one-year underwriting horizon works against long-term health, and how government’s interest in Medicare aligns with prevention. [08:07] Proprietary Data in the Age of Frontier Models How do Lilly, Citadel, or a startup protect proprietary information when agents can analyze anything on the open internet? Apple’s choice to run inference on Nvidia Blackwell for encrypted-in-memory data, cloud providers’ security track record, and the new attack surface agents introduce. [17:42] Nvidia vs. Alphabet for Five Years A forced one-stock choice: Alphabet as the chicken bet spanning search, cloud, Gemini, and DeepMind, versus Nvidia’s ~$160B-trending-$200B free cash flow on minimal CapEx. Plus the memory shortage forcing a ~$275 iPhone price hike and a stack-wide efficiency race. [23:44] Healthcare Science: CRISPR, the FDA, and China A generalizable CRISPR approach that shreds cancerous tumor DNA, an FDA program to cut 6-12 months off Phase 1 trials by partnering biotechs with academic centers, and the competitive threat from faster-moving Chinese biotech. [29:49] Closing: Medicare for All, the Fed, and the Financing Wave Dr. Oz as a Fed-style independent Medicare administrator, identifying the healthcare winners of a restructured system, and the collision of Treasury issuance with massive data-center financing (SpaceX’s $25B debt raise). Download this week’s Cash Flow Memo at telltales.us, and subscribe for energy, tech, and healthcare insights every Wednesday. Cashtags $AAPL $AMZN $CHTR $CMCSA $GOOGL $LLY $MSFT $MU $NVDA $TGT $UPS $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

    34 min
  3. Jun 21

    Weekend Update - W2625

    ▶ Explore this week’s Tape — live, sortable, drill-down → Three Bottlenecks, One Multiple, and Only One Monopoly The AI argument flipped sides this week. For two years the only question that mattered was whether anyone would actually buy all this compute, and Wednesday’s show put that one to bed. The question left standing is whether the supply chain can physically build it — the lithography, the power, the memory. And in the rush to own the chokepoints, the market did something lazy. It paid all three the same monopoly multiple. Only one of them is a monopoly. Start with the one that is. ASML makes the machine that etches the most advanced chips on earth, and it makes it alone — no second source, no roadmap to one, no credible challenger inside a decade. The Cash Flow Memo has it trading around sixty times free cash flow, on roughly eleven billion dollars of trailing free cash flow translated from euros.¹² That is a monopoly multiple, and for once it is bolted to an actual monopoly. The thing that can dent it is not a competitor — it is a government. Commerce Secretary Howard Lutnick told ASML, per Bloomberg, that he is concerned a top-tier extreme-ultraviolet machine may have reached China in violation of export controls; the company denied it has ever shipped one there.³ Notice what the risk actually is. Not the machine. The map of who is allowed to buy it. ASML’s customer list is the only variable that has ever moved this stock, and it is the one variable ASML does not control. Now the two the market is paying as if they were ASML. GE Vernova booked more data-center power orders last quarter than it did in all of the prior year, and its turbine backlog now stretches past a hundred-ten gigawatts into 2029.⁴⁵ Free cash flow grew almost four hundred percent.⁶ Real demand, real scarcity — for now. But a gas turbine is not an EUV machine. Siemens Energy and Mitsubishi build them too, and slot scarcity is the kind of bottleneck competition fills in three or four years, not three or four decades. GE Vernova is renting its scarcity. The order book is genuine; the durability is the open question, and a thirty-times multiple is pricing the scarcity as if it were permanent. Then memory — the bottleneck with the longest rap sheet. Micron reports Wednesday with its entire 2026 high-bandwidth-memory output already sold under contract and DRAM pricing set to jump fifty percent this quarter.⁷⁸ The cashflow read is in Marcus’s column below — short version, the trailing multiple is the cheap number, not the scary one. But here is the part the show had no room for: memory is the one bottleneck on this list that has detonated itself before. Every prior cycle ended the same way. All three makers see the shortage, all three turn the capex spigot at once, supply overshoots demand, and the pricing that looked structural turns out to have been a moment. Micron is committing roughly two hundred billion dollars to new capacity, and its two competitors are spending into the same shortage.⁹ That is either supply discipline or the seed of the next glut, and you cannot tell which from inside the shortage. You never can. That is what makes memory memory. So three bottlenecks, three very different half-lives, and a tape paying all three the monopoly rate. The durable one — the cornered machine — arguably earns it. The other two are borrowing the multiple from the durable one, and the loan comes due the day a competitor adds a turbine line or a third memory maker blinks first on capex. What changes the read is Wednesday, and the tell is Micron’s gross-margin guide. Guide above where consensus already sits and the cycle is still accelerating into the glut question, not away from it.¹⁰ Merely meet it, and a sold-out company that only meets gets sold on the news. The trade across the whole supply chain breaks the moment any one of the three memory makers signals it is racing the other two to fill capacity rather than pacing itself. Wall Street’s consensus on the supply chain: own the bottleneck, any bottleneck, at any multiple. The math says own the one nobody else can build — and rent the other two only as long as the scarcity lasts. The Tape — W2625 Universe of 94 cashflow-memo names, snap dates 2026-06-15 → 2026-06-19. 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 reports Wednesday, and the reporters table holds the whole argument in two cells that read like a contradiction: a 1.0% trailing FCF yield sitting right beside 64% forward revenue growth. Trailing, Micron screens at 102x free cash flow — and that is a cyclical at the bottom of its cash cycle, where the trailing multiple measures the past and tells you nothing about the print. The 1.0% yield is the trough. The 64% is the market pricing what sold-out capacity earns once contract prices reset higher. The cheap-looking cell is the lie; the expensive-looking one is the tell. The test Wednesday is the gross-margin guide — clear where consensus already sits with the cycle still accelerating, or a sold-out name that merely meets gets sold on the news. 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 90 ranked-eligible names ranked. 0 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable). Source: cashflow-memo master_2026-06-19.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation. The Issue — This Week's Brief The Cashflow Memo The Week AI Became a Supply Story The bottleneck moved from demand to supply this week. Micron’s Wednesday print is the first real test. 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 E2626. Chapter markers * Time | Segment * 0:15 | Cold open — the supply side of AI * 0:45 | Theme — the two hardest things to build (ASML, GE Vernova) * 4:45 | Deep dive — Micron, the print that sets the cycle * 8:45 | Rapid-fire — Moderna, FedEx, Nike, Cheniere, Intel * 11:45 | Close — Consensus Watch + the week ahead 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. This is still a pilot, so tell us what’s working and what isn’t. Ava: And one more thing before we start — this one lands on Father’s Day. So happy Father’s Day to all the dads listening. We’re glad you’re spending part of the morning with us. Ava: Here’s the week in one sentence. The argument about whether anyone will actually buy all this AI compute? That got settled on Wednesday’s show. This week the question flipped to the other side of the ledger — can we build the supply. The machines that print the chips, the power that runs them, and the memory that feeds them. All three flashed at once. Three different stories, one spine — the build is now gated by what can physically be manufactured, not by whether the demand shows up. On this week’s Telltales, episode 2625, Hunt, Jason, and Mike spent the hour on the demand side — NVIDIA at $5 trillion, the end of the subsidized-token era, who actually pays for the compute.[^ep-e2625b] We’re taking the other half. The supply chain that has to deliver before any of that demand means a thing. Theme — The two hardest things to build Ava: Two things in this build are genuinely hard to make. One is the machine that etches the chip. The other is the electricity that runs it. Both made news this week, and both made it for the same reason — the demand has gotten ahead of the supply. Start with the machine almost nobody owns. This week the US government told the company that makes it that one of those machines may have ended up somewhere it isn’t allowed to go. US Commerce Secretary Howard Lutnick told ASML’s leadership he’s concerned one of its top-tier extreme-ultraviolet lithography machines may have reached China, in violation of export controls.[^asml-us-export-concern-20260619] ASML denied it flatly — says it has never shipped an EUV machine to China.[^asml-us-export-concern-20260619] This on the same company that just raised its full-year guide to €36–40 billion on AI demand.[^asml-q1-earnings-20260619] Marcus — what does a monopoly on the most important machine in the world actually cost? Marcus: ASML is the only company on earth that can make this machine, and the market has always paid it like one. The memo has ASML at about 62 times free cash flow,[^memo-asml-evfcf-20260619] on roughly $11 billion of trailing free cash flow, translated from euros.[^memo-asml-fcf-20260619] That’s a monopoly multiple for a monopoly. The export-control story doesn’t touch the c

    13 min
  4. The End of the Subsidized-Token Era

    Jun 20

    The End of the Subsidized-Token Era

    All three hosts in one room: a macro tour through oil, gas and a new Fed chair, then a deep dive on NVIDIA at $5 trillion, the software stack’s capex problem, and the Anthropic export-control fight that turns proprietary data into the next moat. The Cashflow Memo Key Takeaways * Macro (Hunt’s exhibits): Oil sits ~$80 heading toward $70 (vs $60 on its way to $50 when the Iran event started), with backwardation compressing to under $10; natural gas averages ~$3.50 across both ’26 and ’27. New Fed chair Kevin Warsh signaled aggressive balance-sheet runoff (~$750B/yr toward a target near $1.5T) and a possible bias to hike — a lot of paper for the market to absorb against a ~$1.5T deficit. * NVIDIA at ~$5T is turning into a value stock: free cash flow on a $200-250B run-rate by year-end (vs a record ~$160B). The bull case has shifted from the chip cycle to TAM expansion — server → rack → row → full reference data-center design, an x86-killer CPU, and direct buildouts for cash-rich non-hyperscalers like Eli Lilly (~$20B FCF), Exxon/Chevron, and Citadel — though AI capex at ~3% of GDP raises a law-of-large-numbers ceiling on incremental budget growth. * Software dispersion is about capex risk: Salesforce screens cheap at ~18x FCF because it owns its own data centers and will likely have to deploy GPUs (capex rising from ~zero), while ServiceNow trades >50x renting AWS; Snowflake stays cash-light on heavy SBC but saw NRR re-accelerate on ~30% revenue growth as its AI product (chat over enterprise data) ramps. * The Anthropic throughline: a cyber-capable model released to ~40 entities (JPMorgan et al.), Amazon/Jassy lobbying Washington for export controls, the guardrailed Fable version jailbroken within two days, and distillation risk from Tencent/Alibaba (~1 year behind). The hosts read Jassy’s move as AWS self-protection, not public spirit — AWS was almost certainly one of the 40. * AI economics as a J-curve, proprietary data as the new moat: OpenAI/Anthropic head public with no profits against massive compute rent; the tell to watch is the end of the subsidized-token era (Uber/Lyft tripled fares post-IPO once VC subsidies ended). Microsoft blocked Anthropic’s models internally after a ToS change let Anthropic capture and train on user prompts (i.e., customer code) — a breach of trust that makes walled-off hosting (Citadel on Google) the real battleground. Show Notes [00:00] Intro & this week’s Cash Flow Memo Mike sets up the episode and points listeners to the memo at telltales.us. [00:26] Exhibits A/B/C — Oil, Gas & the Fed Hunt’s five minutes: oil $80 toward $70 with backwardation under $10, nat gas ~$3.50 across ’26-’27, and new Fed chair Kevin Warsh signaling balance-sheet runoff and a possible hike bias. [04:37] Apple & Snap — foldables, camera AirPods, AR glasses Apple’s underwhelming conference (a 2028 foldable, AirPods with cameras) versus Snap’s see-through AR glasses and why Meta’s audio-only glasses are the best product today. [07:30] Why Amazon called Washington on Anthropic Anthropic’s cyber-capable model, the limited release to ~40 entities, and the hosts’ read that Jassy’s export-control push is AWS self-protection. [10:53] Jailbreaking the guardrails & the China distillation risk The Fable guardrailed version cracked in two days, and why Tencent/Alibaba stay roughly a year behind. [13:46] What is SpaceX worth? — AI’s J-curve to profits SpaceX public and +30%, OpenAI/Anthropic going public with no profits, and how to think about revenue growth that costs enormous compute. [15:44] The end of the subsidized-token era Jason’s Uber/Lyft analogy: cheap VC-funded tokens today, tripled rates after the IPO. [16:11] Microsoft’s per-seat agent bet Why fixed price-per-seat plus usage upside is the right model for a slow-moving enterprise base, and the job-displacement question. [18:54] Software dispersion — Salesforce, ServiceNow & Snowflake Salesforce at 18x FCF and a coming GPU capex bill, ServiceNow renting AWS, and Snowflake’s NRR re-acceleration. [21:34] Oracle, Broadcom & NVIDIA at $5 trillion Oracle and Broadcom’s pivots, then NVIDIA as a record-FCF value stock. [24:45] NVIDIA’s next TAM — racks, rows, x86 & selling direct From server to full reference data-center design, going after the x86 CPU market, and building private clouds for enterprises. [25:53] Healthcare & the cash-flow hunt — Lilly, Exxon/Chevron, Citadel NVIDIA chasing free-cash-flow-rich customers: Lilly’s walled-off AI buildout, oil majors, and Citadel’s Google deal. [29:48] Proprietary data as the new moat Microsoft blocks Anthropic’s models after a ToS change to capture and train on customer prompts — a breach of trust. [32:31] Next week & sign-off A teed-up deep dive on protecting proprietary data: defense versus offense, and whether anything can truly be walled off. If these conversations have earned a place in your week, send the show to one person who’d genuinely enjoy it. Download the Cash Flow Memo at telltales.us. Cashtags $AAPL $AMZN $AVGO $BABA $CRM $CVX $GOOGL $JPM $LLY $LYFT $META $MSFT $NOW $NVDA $ORCL $SNAP $SNOW $TSM $UBER $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

    36 min
  5. The One Week We Were All in the Same Room

    Jun 19

    The One Week We Were All in the Same Room

    For the first time in four years, there’s no new Telltales episode this week. The one week all three of us were finally in the same city — Hunt, Jason, and Mike, in one room in New York instead of scattered across two coasts — the recording gremlins picked their moment. The primary recording failed partway through. There’s a backup, and we’re actively working to recover it. So here’s the plan: if we recover the audio, this week’s episode lands in your feed the moment it’s clean. If we don’t, no harm done — Hunt, Jason, and Mike are back in the chair next Wednesday, and Ava is back with the Weekend Update on Saturday. While you wait, this week’s Cash Flow Memo is below. And if these conversations have earned a place in your week, send the show to one person who’d genuinely enjoy it. Almost all of our growth has come from listeners doing exactly that — and we don’t take a single recommendation for granted. 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. The Cashflow Memo 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

    1 min
  6. Jun 15

    Weekend Update - W2624

    ▶ Explore this week’s Tape — live, sortable, drill-down → Good News, Sold: The AI Buildout’s First Bill Came Due Every capital cycle has the same tell. It is not the day the spending stops. It is the day the market stops clapping for it — when a company does exactly what it promised and the stock falls anyway. This week the most expensive trade on earth hit that day three times. Oracle delivered the backlog it told everyone it would deliver. A six-hundred-thirty-eight-billion-dollar pile of contracted future revenue, up more than three-and-a-half times in a year.¹ Cloud infrastructure revenue up ninety-three percent.² It did the thing. The stock fell ten percent.³ For two years the AI trade was a referendum on demand: is it real, how big, how fast. The bull won that argument. So the market moved the goalposts, the way it always does at this point in a buildout — from is the demand there to who pays to meet it, and what does the bill do to the balance sheet carrying it. Bill Maris said the quiet part on All-In this week: a trillion dollars of spend commitments sitting on sixty billion of revenue, and now you go to the public markets and hope retail picks up the difference.⁴ That is the bear case in one sentence. Three names walked straight into it this week, and the interesting part is that each of them is paying the bill a different way. Oracle is borrowing it. Trailing free cash flow is already negative — they spent the better part of fifty billion dollars on capex to build the capacity behind that backlog, and the money to fund the next leg is coming out of the debt market into a tape that has stopped rewarding capital plans.⁵ NextEra is diluting for it. The biggest bet in the history of regulated utilities — a sixty-seven-billion-dollar, all-stock takeover of Dominion, built explicitly to wire thirty gigawatts of data-center power by twenty-thirty-five, Google and Meta already signed.⁶ All-stock is the tell. A utility already carrying sixteen times debt to free cash flow cannot borrow its way to the biggest deal in its sector’s history, so it is paying with equity and handing the cost to the shareholder it already has.⁷ The stock fell nine percent.⁸ Micron is the one name pre-selling it: its entire next fiscal year of high-bandwidth memory is already spoken for, certified into Nvidia’s roadmap, the demand contracted before the capacity ships.⁹ Three funding strategies — debt, equity, pre-sold demand — and one underlying wager: that AI demand is durable enough to convert all of it to cash before the interest comes due. The cashflow read is in Marcus’s column below; short version, two of these three have no multiple to quote because the denominator is still negative. Page one of the Cash Flow Memo this week is a capex ledger, not an earnings sheet. This is what the back half of a capital cycle looks like, and it rhymes. The railroads, the fiber glut, the shale decade — the capital cycle never turns when the building stops. It turns when the market stops paying for the announcement and starts pricing the lag between the spend and the cash. The demand did not get worse this week. The accounting for it did. The next data point is dated. Micron reports Wednesday the twenty-fourth.¹⁰ On the screen it trades at a hundred times trailing free cash flow, which sounds insane and tells you nothing — that is trough cash flow at the bottom of a memory cycle. Price it forward and consensus has it at a single-digit multiple of next year’s earnings,¹¹ with trailing cash flow already up more than five-fold off the low.¹² The test is not the headline print. It is whether the forward demand the whole buildout is leaning on shows up in one company’s order book. If Micron’s guide confirms, the bill looks affordable. If it wobbles, the cycle just got more expensive for all three. Wall Street’s consensus on the AI buildout: too crowded, too expensive, too late. The stocks fell this week not because the demand soured — because the market finally started counting the cost. That is not the top. That is the capital cycle clearing its throat. The Tape — W2624 Universe of 94 cashflow-memo names, snap dates 2026-06-05 → 2026-06-15. 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 Healthcare spent the week getting repriced by Washington, and the screens are tarring the whole sector with one brush. Regeneron is the name that brush gets wrong. It sits near the top of the board at 11.3x EV/FCF on an 8.8% free cash flow yield — cheaper than UnitedHealth, with no federal prosecutor auditing the numerator. That is the distinction the table cannot draw: UNH’s cash engine is the thing the DOJ is investigating; Regeneron’s is a drug franchise nobody has subpoenaed. Same sector, same de-rate, two completely different reasons for the cheap — one is cheap because the cash might be fiction, the other because it shares a GICS code with the one that might be. The test on the next print is whether biosimilar pressure on the franchise shows up in the cash line, or the de-rate was just guilt by association. 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 90 ranked-eligible names ranked. 0 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable). Source: cashflow-memo master_2026-06-15.csv. NTM growth from FMP analyst-estimates consensus. Composite is a percentile rank, not a recommendation. The Issue — This Week's Brief The Cashflow Memo When Cheap Stopped Being Safe The week being cheap stopped being safe, and the AI buildout kept spending anyway. The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the universe of the Cash Flow Memo. About 13 minutes. No filler. Download the memo at telltales.us. Hunt, Jason, and Mike are back Wednesday on episode 2625. Chapter markers * Time | Segment * 0:15 | Cold open * 0:55 | Theme — the AI buildout’s bill: Oracle, NextEra, Micron * 5:10 | Deep dive — UnitedHealth vs CarMax * 9:25 | Rapid-fire — Moderna, Intel * 11:30 | Close + Consensus Watch 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 early — this is a pilot, and we want to hear what’s working. Ava: Here’s the week. Being cheap stopped being safe. The cheapest large-cap in healthcare spent the week with a federal prosecutor at the door. The retail name everyone screens as cheap is being handed customers by a tariff. And while the market was busy repricing the cheap stuff, the most expensive trade on earth — the AI buildout — just kept writing bigger and bigger checks, and got punished for it anyway. On Wednesday’s main show, episode 2624, Hunt, Jason, and Mike worked through the economics of that buildout — the data centers, and the sheer physics of powering them.[^ep-e2624] We’re going to put the cashflow lens on the bill. Because this was the week the bill started showing up in three different places at once. Theme — the AI buildout’s bill Ava: Start with the most expensive trade in the market, because this week it got complicated. For two years the AI story was demand — is it real, how big, how fast. This week the question flipped to cost. Who delivers it, who powers it, and whose balance sheet carries it. And here’s the contrarian note hanging over the whole thing: as Bill Maris put it on All-In this week, quote, a trillion in spend commitments on $60 billion of revenue, and now you’re going to go to the public and hope that retail is going to pick that up.[^tp-maris-allin-20260609] Hold that thought. Because three companies just tested it. Ava: Start with Oracle, because it did exactly what it promised and got punished for it. It delivered the backlog it said it would — and the stock fell 10% anyway.[^orcl-stock-20260610] Page 2 of the memo: remaining performance obligations hit a record $638 billion, up 363% year over year.[^orcl-rpo-20260610] Cloud infrastructure revenue up 93%.[^orcl-oci-growth-20260610] They did the thing. Marcus — why did doing the thing get them sold? Marcus: Because the market stopped grading the backlog and started grading the bill. Oracle’s trailing free cash flow is negative — minus $21 billion[^memo-orcl-fcf-20260615] — because they spent $48 billion on capex over the last year to build the capacity behind that backlog.[^memo-orcl-capex-20260615] So there’s no multiple to quote. Don’t reach for one; it’s a negative number. What prices Oracle now is one question: does $638 billion of contracted intent convert to cash before the interest on the build eats them. They raised the capital plan into a tape that no longer claps for capital plans. Maris’s line is the bear case in one sentence — and Oracle just walked s

    11 min
  7. Data Centers in Space: How SpaceX Justifies $2 Trillion

    Jun 10

    Data Centers in Space: How SpaceX Justifies $2 Trillion

    Hunt, Jason, and Mike work through an oil market frozen by the US-Iran standoff, then spend the back half stress-testing SpaceX’s ~$2 trillion valuation against Tesla’s — landing on data centers in orbit and a chronic compute shortage as the load-bearing assumptions. Plus Apple’s WWDC miss, the software disruption map, and a strong week of healthcare data. The Cashflow Memo Key Takeaways * Hunt’s working case: if the US-Iran standoff just persists, oil sits near $90 twelve months out despite deep backwardation (mid-$90s spot vs mid-$70s forward); the casualty is natural gas, because roughly two-thirds of dry-gas supply growth is Permian associated gas, so a higher oil price pumps more gas and keeps gas pricing weak. * Exhibit A is the macro tail risk: holding public-debt-to-GNP under 100% requires keeping the deficit flat-to-declining near $1.5T even with higher defense spend — a developed-world problem (Japan is ~180%), not just a US one, and the thing that most threatens the credit markets. * SpaceX’s ~$2T valuation only pencils on space-based data centers: 150kW micro-satellites (one NVL72-rack equivalent, Nvidia silicon, in-sourced solar), a ~1 GW launch cadence targeted by year-end, ~170 Starship launches per gigawatt (a launch every other day) off a 20-30 ship fleet. At ~$20B FCF per deployed gigawatt, ~$100B FCF — a 5% cash yield — looks more reachable for SpaceX than for Tesla at $1.4T, which needs robotaxi plus humanoids to get there. * Compute scarcity is the load-bearing assumption under both bets: xAI’s ~$20B Colossus buildout in Memphis is already leasing capacity to Anthropic and Google at ~$25B of revenue — a near one-year payback that shows how acute the shortage is. The safer, cheaper derivative on chronic compute shortage remains Nvidia. * AI has collapsed the cost to write software but not to support or design it: as the agent becomes the hub, Apple looks exposed (a second straight underwhelming Siri at WWDC, starting to look like IBM or Intel), while horizontal incumbents like Salesforce and ServiceNow can expand share and consumption-priced models gain over seat-based ones. In healthcare, Lilly’s triple-agonist retatrutide showed dramatic weight loss with less muscle loss, plus a gene-therapy LDL result, and Revolution Medicines tripled survival duration in pancreatic cancer by hitting targets long thought undruggable. Show Notes [00:00] Intro: The Cash Flow Memo Mike opens the weekly walk through energy, technology, and healthcare. Download the memo at telltales.us. [00:52] Oil, Iran, and Why Gas Stays Weak With supply off ~2.5-3M bbl/d and the Iran embargo likely to persist, Hunt sees oil near $90 a year out despite heavy backwardation — and natural gas as the casualty, since Permian associated gas grows with oil. [04:24] Exhibit A: The Deficit and the Credit Markets Keeping public-debt-to-GNP under 100% means holding the deficit near $1.5T even with higher defense spend. A developed-world problem, not just a US one. [05:32] SpaceX vs Tesla: Which $2T Bet? Google’s ~$80B equity raise and the SpaceX, OpenAI, and Anthropic financings frame the question: is SpaceX at $2T or Tesla at $1.4T the more defensible valuation? [06:57] Data Centers in Space: The Economics 150kW micro-satellites the size of an Nvidia NVL72 rack, in-sourced solar, ~1 GW launch cadence by year-end — and a deployed cost on par with land, minus the 3-7 year build cycle. [12:30] 170 Launches a Gigawatt: The Physics The hard part is mass to orbit: ~170 Starship launches per gigawatt, a launch every other day, on an 18-hour turnaround across a 20-30 ship fleet. At ~$20B FCF per GW, $100B FCF becomes conceivable by ~2028. [15:43] Colossus, Compute Scarcity, and Nvidia xAI’s ~$20B Memphis buildout now leases to Anthropic and Google at ~$25B revenue — a one-year payback that proves how short compute is. The cheaper, safer derivative is Nvidia. [17:22] Apple’s WWDC Miss and the Agent-as-Hub Threat A second straight underwhelming Siri. If the agent becomes the hub for your data, the device is just a screen — and Apple, staked on privacy, is letting the opportunity pass. Starting to look like IBM or Intel. [20:32] Enterprise Software: Horizontal Wins, Vertical Fragile AI made writing software trivial, not supporting it. Token budgets fold into the software budget; horizontal incumbents (Salesforce, ServiceNow) expand, vertical and seat-based models get fragile, and consumption pricing wins. [23:47] Healthcare: Lilly, Gene Therapy, and Pancreatic Cancer Lilly’s triple-agonist retatrutide (less muscle loss) and a gene-therapy LDL result, Mayo’s AI-assisted earlier pancreatic-cancer detection, and Revolution Medicines tripling survival on previously undruggable targets. [26:03] Wrap-Up Possibly back next week on how SpaceX trades in the aftermarket. Get the Cash Flow Memo at telltales.us. Cashtags $AAPL $CRM $GOOGL $INTC $LLY $MSFT $NOW $NVDA $RVMD $TSLA 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

    28 min
  8. Jun 6

    Weekend Update - W2623

    ▶ Explore this week’s Tape — live, sortable, drill-down → Even the Cash Machines Are Borrowing Now The story this week was not that the AI build is expensive. Everyone knew that. The story was that the companies best equipped to pay for it out of their own pockets stopped doing so — three of them, in the same five trading days. Alphabet, which had not sold a share of stock since its IPO, raised about eighty-five billion dollars in equity to fund its data-center build, with Berkshire Hathaway taking ten billion of it directly (per CNBC).¹ Meta, which did not need a dime, was reported to be lining up a raise of its own — and shed almost seven percent on Friday on the rumor alone (per Bloomberg).² Apple, on page one of the same Cash Flow Memo, did the one thing nobody else in the arms race is doing: it rented. Take the first one, because it is the one with thirty years of history behind it. A company funds its capital spending out of its own cash flow right up until the build outgrows the cash flow. The day it reaches for outside money is the day the build stopped paying for itself. Alphabet’s trailing free cash flow fell about forty-four percent year over year — the capex ate it — so they sold stock.³ The company that described itself as the asset-light, capital-light compounder for two decades just issued eighty-five billion dollars of equity to buy data centers. Asset-light. The cashflow read is in Marcus’s column below — short version: at roughly seventy times trailing free cash flow, this is now a capital-intensive company that happens to own a search engine.⁴ But the tell this week was not Alphabet alone. It was the synchronization. Meta still throws off about fifty billion dollars of trailing free cash flow, still growing north of twenty percent, and trades at thirty-two times — the cheapest large-cap cash machine on the board.⁵⁶ When the best-capitalized company in a sector decides it wants a financing cushion anyway, that is not a statement about that company. It is a statement about the size of the build. We have watched real demand get financed externally before. The late-nineties fiber boom funded a build the internet genuinely needed — and the companies laying the cable raised equity and debt into the same thesis at the same moment. The capacity got used. A decade later. The financing peak and the equity peak landed in the same window, and the build being right did not save the multiple. The lesson was never that the demand was fake. It was that synchronized external financing is what a capital cycle looks like at its most expensive, not its cheapest. Which is what makes Apple the sharpest line on the page. Apple pays Google roughly a billion dollars a year to run the next Siri on Gemini rather than train a frontier model itself — less than one percent of the hundred-twenty-nine billion dollars in free cash flow it generates annually.⁷⁸ Everyone else is spending a hundred times that and selling stock to do it. Dilute, borrow, or rent: only one of the three costs the shareholder nothing. So the demand side is now financed. What gets tested next is absorption — whether the capacity fills before the financing cycle turns. The first data point lands Wednesday after the close, when Oracle reports against a backlog north of five hundred fifty billion dollars and trailing free cash flow that is already negative.⁹¹⁰ Broadcom just showed everyone the grading curve: a record AI-chip quarter, up more than a hundred-forty percent (per Broadcom’s fiscal-Q2 release), and the stock fell fifteen percent anyway because Hock Tan declined to raise the story he had already sold.¹¹¹² At these multiples, delivering is not enough; you have to deliver and beat the build you already financed. Mark the calendar for Wednesday. The thesis breaks if the backlog keeps growing and the cash behind it never shows up. Wall Street’s consensus on the AI build: the balance sheets are fortresses and the capex funds itself. Three of the richest companies on earth sold or floated stock this week to tell you it doesn’t. The Tape — W2623 Universe of 94 cashflow-memo names, snap dates 2026-06-02 → 2026-06-05. Composite is rank-sum percentile of FCF Yield + NTM Revenue Growth (higher = better balance). Banks and finance-book names shown separately. Telltales Yield — Top 10 From the Cashflow Desk — Marcus Graham The name sitting at the top of our own board is the one that does not have to finance anything. Uber leads the composite this week on a 7.5% free-cash-flow yield, growing the forward top line about fifteen percent, at thirteen times EV to free cash flow. It converts cash and grows without selling a data center or a share to do it — the asset-light, self-funding profile Alphabet used to own before it raised eighty-five billion dollars of equity this week to pay for the build. The screens still file Uber under gig economy and the hyperscalers under quality compounders. The cash flow says the labels are backwards. What changes the read is the day Uber has to raise outside capital the way the hyperscalers now are; until then, the top of the board is the part of the market still paying for itself. Telltales Yield — Bottom 10 This Week’s Reporters Note: Oracle’s 46.3% NTM revenue-growth figure is the raw FMP analyst-estimates consensus and looks anomalous against Oracle’s own ~15% forward guide; we read Oracle through its RPO backlog and trailing free cash flow (negative this quarter on the capex cycle), as in The Take above, not this estimate. Sector Medians Debt / FCF Watch (highest leverage on TTM FCF) Weekly Price Movement Top 5 (week-over-week price) Bottom 5 (week-over-week price) Banks (shown separately — FCF metric not meaningful) Finance-book — FCF not comparable Customer-float / captive-finance / reserve businesses (IBKR broker float, KMX CarMax Auto Finance, PYPL customer funds, CRCL stablecoin reserves). The memo’s operating-FCF method overstates their FCF, so they are held off the ranked leaderboard pending the P&L-waterfall rebuild. Data Gaps 90 of 90 ranked-eligible names ranked. 0 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable). Source: cashflow-memo master_2026-06-05.csv. NTM growth from FMP analyst-estimates consensus. Composite is a percentile rank, not a recommendation. The Issue — This Week's Brief The Cashflow Memo Who Pays for the AI Build? The week the AI trade stopped being about the chips and became a question about who pays for them. The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the universe of companies in the Cash Flow Memo. About 13 minutes. No filler. Download the memo at telltales.us. Hunt, Jason, and Mike are back Wednesday on episode E2624. Chapter markers * Time | Segment * 0:00 | Disclaimer * 0:15 | Cold open * 0:45 | Theme — Who pays for the AI build * 4:45 | Deep dive — AI infrastructure’s two-act week * 8:45 | Rapid-fire + the forward week * 11:45 | Close + Consensus Watch Full transcript Disclaimer Ava: The following conversation is intended for informational purposes only. You should always do your own work to determine if an investment is suitable for you. Cold open Ava: You’re listening to the Telltales Weekend Update. I’m Ava Cabot. Marcus: And I’m Marcus Graham — the cashflow desk. Ava: Quick note: the show is produced entirely with AI tools, and both voices you’re hearing are AI-generated. Send feedback through the Substack. We’re still in pilot, so tell us what’s working and what isn’t. Ava: Here’s the week. For two years, the AI trade was a story about chips — who makes the fastest one, who gets the allocation. This week it turned into a different question, and it’s a harder one. Who actually pays for the build? Broadcom printed a record and got punished for it. Oracle’s about to walk into the same exam on Wednesday. And three of the biggest companies on earth spent the week showing you three completely different ways to fund the thing — dilute, borrow, or rent. On Wednesday’s main show, episode 2623, Hunt, Jason, and Mike framed the AI capex boom as the macro tailwind holding up a $31 trillion economy.[^ep-e2623] We’re going to put the cashflow lens on it. Because the tailwind has an invoice attached, and this week the invoice started coming due. Theme — Who pays for the AI build Ava: Start on page one of the memo, because Apple and Alphabet are sitting right next to each other this week giving opposite answers to the same question. Alphabet chose dilution. The asset-light beautiful business, the company that hadn’t sold a share of stock since its IPO, just raised about $85 billion in an equity offering to fund its AI build — with Berkshire Hathaway taking $10 billion of it in a private placement.[^googl-capital-raise-20260601] That’s to cover a capital-expenditure budget of $180–190 billion this year alone.[^googl-capex-2026-20260601] Marcus — what does that raise tell you that the press release won’t? Marcus: It tells you the cash machine stopped covering its own build. Going into this, the memo had Alphabet’s trailing free cash flow falling 44% year over year — capex is eating it alive.[^memo-googl-fcf-20260605] You don’t raise $85 billion in equity when your own cash flow funds the plan. You raise it when it doesn’t. 69 times trailing free cash flow for a company now diluting shareholders to keep up.[^memo-googl-evfcf-20260605] The beautiful business framing is over. This is a capital-intensive company that happens to own a search engine. Ava: Meta took door number two. Mark Zuckerberg spent the week reportedly floating an equity raise of his own to fund $125–145 billion of capex — and the stock fell almost 7% on Friday just on the report that it might.[^meta-capital-rai

    12 min

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

You Might Also Like