# Special Situations Report — Weekly Call (readable notes)

- **Title:** Non-farm payrolls miss (57k) → "bad news is good news" rate relief, the yen at a 40-year low, the memory / AI selloff (SK Hynix −30%, Micron sell-the-news despite a blow-out), the Shopify (SHOP) long thesis (~60% upside to $190), a special-sits/M&A sweep (Comcast NBCU spin, Rocket Lab/Iridium, S&P Mobility spin MGGL, Martin Marietta/Lhoist, the Allied Gold (AAUC) 34% Zijin arb), Jana in Everpeer (ex-Pure Storage), Burry's new shorts (NVDA/TSLA/CAT/AMAT), Altana's life-insurer short (LNC/MET), Blue Owl (OWL) 5% redemption gates, the Q2 Russell reconstitution "style blurring", Apple's memory-cost hit, and Sable Offshore (SOC) post-refi
- **Show:** Weekly SSR research call (premium subscriber recording — no public video)
- **Guest/host:** Jay Singh (founder, Special Situations Report; ex-Goldman Sachs)
- **Date:** 2026-JUL-05 (Sunday call; PDFs title it "2026-07-05")
- **Source PDFs (this folder):** `transcript.pdf` (premium transcript), `report.pdf` (written SSR condensed report + actionable items), plus supplemental research Singh references: `DB Emerging Themes.pdf` (Deutsche Bank), `Mag7 Underperformance - Apollo.pdf` (the Apollo presentation on Mag7 cash flow / capex), `State of the AI Economy.pdf`, `Shopify Valuation and Risk Framework.xlsx` (the SHOP DCF model)
- **Note:** readable notes auto-extracted from the premium PDFs; fillers (um/uh/false starts) removed, wording otherwise verbatim from the recording. Section headers carry the recording's `MM:SS` cue. No public video, so the per-name table has no deep-links. Auto-transcript garbles corrected to the right entity (Everpeer = rebranded Pure Storage/PSTG; MGGL = the S&P Mobility spin; Druck = Druckenmiller; Wycoff = Witkoff; Lhoist limestone; Bonkiero/Agbaio ≈ Bonikro/Agbaou; "Vestin's tested agreements" smoothed).

---

2026-07-05 - SSN - NonFarm Payrolls Miss, Rates, Yen Collapse, AI Selloff, Shopify Growth, Special Sits Update

## Introduction and Weekly Overview - 00:00

Happy Sunday everyone. Hope you had a wonderful weekend with your families.

This week we cover several topics: non-farm payrolls and interest rates — how the Fed is less likely to hike, which is why we had a brief rally after the payrolls print. We talk about Shopify, which we think has 60% upside from $119.50 on Friday's close to about $190 a share over the next one to two years.

We also cover large-cap special situations from last week: the Rocket Lab deal, Comcast, the Gold arb spread (about 34%), and the S&P spin-off (SPGI spin). On page 17: investor bets against life insurers and continued private-credit redemptions. Then the US market's Q2 recap, Tesla sales recovering on the oil spike (we think it's temporary), the JPMorgan software valuation matrix, and a big AI/semiconductor discussion — there was a large sell-off in memory, SK Hynix down 30% from the local highs. The yen fell to the lowest level in 40 years. A quick Iran update (very brief — I'm tired of it). Druckenmiller adding to SE, which we had a model snapshot for last week. Burry shorting Caterpillar, Tesla, and others. Kevin Warsh's speech in Europe (page 44) where he sounded more dovish. The BIS warning that debt-fueled spending on AI is raising the risk of a global financial crisis (a little overdone). Page 48: why Apple stock has sold off relative to the Mag 7. Page 49: Samsung and SK Hynix could announce as much as $1.3 trillion of capex over 10 years on Monday. Page 50: the momentum sell-off in Goldman's momentum names. Page 52: part two on AI — Chinese AI models being adopted by Microsoft. Page 53: a snapshot of the coming SK Hynix IPO. Page 54: Bloom Energy's deal with Brookfield, and the SMCI scandal. Page 55: the Sable Offshore debt refinancing (some of you got involved around $3 a share) — we sold much higher, then the stock collapsed on an equity offering to about $3; to get long we sold some 2.5-strike and 2-strike puts with very high premium, where we think we'll make a 70–80% return on capital. That trade is still on. Page 56: will the Mag 7 cash flow ever recover — the risks, tying into the Apollo presentation in the attachments.

Very light economic week ahead: services PMI, ISM services, and existing home sales.

## Non-Farm Payrolls and Market Reaction - 04:01

To dive right in: non-farm payrolls were a miss, and we are now back to "bad news is good news" — any bad economic data means the Fed is less hawkish than expected in June during Warsh's first speech. So the worse the unemployment news and the lower inflation falls, the better it is for the market.

Non-farm payrolls came in at 57,000 versus an estimate of 113,000. The unemployment rate fell to 4.2% — but because 750,000 people left the labor force, which is not good. The unemployment rate falling isn't always good; so many people left that it means they're tired of looking, especially white-collar labor with savings. Participation fell from 61.8 to 61.5. Private payrolls were only 49,000. April and May were revised down about 74,000. The last four months were the weakest print: January 160k, February −156, March +214, April 148, May 129, June 57. There was a surprising number of layoffs in hospitality (leisure and hospitality lost 61,000; building materials lost 12,000), offset by health, social, and professional services.

US dollar inflation swap rates have fallen very sharply from the April peak back to roughly where they were before the Iran war — the market is fast to price these in. Rate-hike odds dropped as Warsh's Europe speech played out more dovish. Overnight we've had a rebound in Asian semiconductor stocks: Nasdaq futures up 1.4%, S&P up 40 bps, Russell up 25 bps. But the market is very fickle and can turn overnight, so we can't depend on the futures.

## Shopify (SHOP): An AI-Powered Growth Story - 07:45

One of our growth analysts identified Shopify as an AI beneficiary — they're incorporating AI tools successfully in the business model. Of the SaaS software companies, it's another that has sold off a lot this year, and over the next two years we think it grows into its valuation. Shopify keeps expanding on attractive e-commerce tailwinds, international expansion, continued payments penetration, and growth within Shopify audiences and merchant solutions. A key beneficiary of the pandemic-driven shift to e-commerce, it saw multiple compression after COVID, and the debate has shifted to the durability of GMV growth and levers to expand its take rate (its margin).

With an expanding solution portfolio and the recent pullback, SHOP looks attractive with room for operating leverage. There's a simple DCF model (in the attachments) — total merchants, net new merchants, subscription revenue growth, merchant-solutions revenue growth — flowing down to operating income, cash from operations, and free cash flow used to forecast the stock price.

What is Shopify? A cloud-based commerce platform for merchants to create omnichannel experiences — big brands like Lululemon and Vuori sell through it. It serves Canada and the US primarily; a leading global commerce platform providing internet infrastructure to start, scale, and manage retail operations. Dual revenue model: (1) **Subscriptions** — recurring monthly SaaS fees for store themes, inventory tools, analytics; (2) **Merchant Solutions** — transaction-driven revenue (the bulk of top-line growth): payment processing via Shopify Payments, shipping logistics, and cross-border checkout (Shop Pay).

Shopify powers millions of businesses across 175+ countries (concentrated in US/Canada), from independent entrepreneurs to Skims, Supreme, Vuori, and Mattel. EMEA is the fastest-accelerating expansion vector. International Shop Pay volume was up 70% year-over-year in Q1. In its latest shift, Shopify integrated generative-AI tools like its **Sidekick** commerce assistant into its catalog architecture to optimize storefront management and search-to-order conversion.

We think AI is a tailwind for Shopify — rapid innovation at the top of the funnel is driving GMV share gains and market share from other solutions. (GMV = gross merchandise value: the total dollar value of products sold across the platform; Shopify takes a percentage, so it's strongly correlated to revenue.) By incorporating AI tools it has attracted more businesses that see the ability to monetize the flywheel. In Q1, GMV grew ~35% year-over-year (30% constant currency), an acceleration from 31% in Q4 — the fourth consecutive quarter of 30%+ GMV growth — with: (1) 100M+ merchant count, big merchants doubled over two years (enterprise traction); (2) European GMV growth of 35% (impressive because Europe isn't growing); (3) offline retail accelerating to 33% (brick-and-mortar). Looking forward, generative-AI investment should drive the next leg: (A) improving top-of-funnel for new merchants (President Harley Finkelstein's "agentic entrepreneurial evolution"); (B) improving share of e-commerce as AI-driven search drives twice the new buyers of traditional search — mitigating the risk that people search on Gemini instead of Google (Shopify has partnered with LLM optimizer tools); (C) broadening usage among existing customers — merchants using Sidekick weekly were up 4x in Q1, 12,000 custom apps built, and nearly half of all Shopify flows in Q1 were built with Sidekick; (D) leading into agentic commerce via the **Universal Commerce Protocol (UCP)** — last month Amazon, Meta, Microsoft, Salesforce, and Stripe all joined the UCP council, which Shopify started.

The AI tailwinds have only just started, and the company has a track record of accelerating share gains during commerce disruptions. The one risk is rising token cost (as with all AI-using tech), but they've used it wisely — margins haven't eroded much. On the last call management called out rising token costs as a weight on subscription gross margins, touching a nerve for investors worried about structural software costs. In our view the rising costs are best seen as investments in future growth — more merchants, more GMV, better conversion — all driving the core flywheel. At the same time Shopify aggressively used AI internally: it trail-blazed 20% top-line growth with flat headcount well before Silicon Valley peers. In the quarter, further opex leverage drove a 260 bps year-over-year improvement in non-GAAP operating margins despite the AI hit, more than offsetting an 80 bps gross-margin degradation, translating into 31% free-cash-flow growth.

Bottom line: this best-in-class athlete is using GenAI to expand its market opportunity, gain share, and execute efficiently versus e-commerce software peers — a durable runway for outsized free cash flow. Q1 specifics: GMV $101B, +35% (30% cc), above the $99B consensus; North America GMV accelerated (its main market); total revenue ~$3B (+34.3%, 32% cc, ~1% ahead); international GMV +45%; Europe +45% (35% cc); offline +33% (from 29%); B2B GMV +80%; Shopify Payments penetration up; Merchant Solutions ~$2.4B (+39%); take rate beat (~2.4%). Free cash flow ~$476M, 750 bps ahead of the street's $443M, a ~15% FCF margin (70 bps above consensus). Operating margin 16.5% vs 15.5% consensus (would have met consensus but for AI spend in gross margin).

The bull case: the underlying engine is compounding robustly ($101B GMV +34%, $3.2B revenue, 15% FCF margin). The bear case: the stock isn't cheap on traditional cash-flow metrics — Wix is much cheaper — but this company grows much faster, so it trades at a slight premium; the market has brought growth expectations down from mid-30s to high-20s, and a miss of that high-20s could still correct it. Hopefully management was honest that they'd do high-20s, and a lot of the growth decline is priced in.

## Special Situations and M&A Update - 20:40

**Comcast (CMCSA)** shares were up 20% after announcing a plan to separate its media/entertainment business (NBCUniversal and Sky) from broadband and wireline. The spun-off NBCUniversal will include theme parks, Universal Film and TV Studios, NBC and Telemundo, Peacock, Bravo, and Sky. It's a tax-free spin-off to Comcast shareholders in roughly one year (subject to board approval, tax opinions, regulatory approvals, financing). NBCUniversal keeps the same dual-class structure; Comcast retains ~20% for up to a year, to monetize tax-efficiently over time. If it pulls back, we'd get involved.

**Iridium (IRDM)** entered a definitive agreement to be acquired by **Rocket Lab (RKLB)** in a cash-and-stock deal for ~$8B enterprise value: $27 cash plus Rocket Lab shares via an exchange ratio subject to a collar (67.5 to 112.5), closing mid-next-year. Rocket Lab actually rallied 8% because it took out a competitor. In other news, **BioLife (BLFS)** drew takeover interest from **Repligen**; **Theravance** entered a definitive all-cash agreement to be acquired by **Zymeworks** at $17/share; and **Martin Marietta (MLM)** agreed to combine with **Lhoist North America** in a $13.5B deal including debt (using $7B cash plus ~$6.5B of stock).

## Allied Gold (AAUC) — Zijin Arb Spread ~34% - 28:25

Major arb spreads (page 12): NSC/UNP still ~14%; WPP/Sky ~16% — still wide because the UK ruled against the Getty/Shutterstock merger; we're getting out of the WPP/Sky position on Monday close to break-even or a small profit (implied vol has come down). Because of that UK exposure, Warner Bros./Peacock/Sky also carries UK risk that they'll be forced to sell UK assets.

The Rocket Lab/Iridium and Comcast deals: Comcast's spin unwinds the 2010s distribution-plus-content integration, like AT&T's disastrous Time Warner deal; the market was fatigued with big media/cable tie-ups, so equity gave Comcast a 20% rally on the value unlock. Martin Marietta's $13.5B Lhoist deal: $7B cash + $6.5B stock.

**Jana Partners** built a new position in **Everpeer** — formerly **Pure Storage (PSTG)** — now owning more than 1 million shares as of Q1, though the current size is unclear. Everpeer gained ~9% to ~$79. Jana began the position in Q1 2026 but didn't include it on its 13F because it requested a temporary SEC disclosure delay (Berkshire used to do that). Everpeer offers flash-based data storage plus software for enterprises, competing with NetApp and Dell, and it's fairly cheap versus peers relative to ~20% growth. It rebranded from Pure Storage in March toward a broader hardware-plus-software data platform to capture the AI wave. The stock was already up ~18% year-to-date; earnings growth accelerated in its last three quarters, but it faces rising flash-memory costs (both Everpeer and NetApp raised prices in the past six months). For fiscal 2028 (really 2027, January fiscal year), it's supposed to do about $1.1B of free cash flow — not deep cheap (~4% FCF yield) — but doing that while growing 20% is why Jana finds it interesting, likely a big TAM.

**AAUC Canadian Gold spread** (a subscriber question): **Allied Gold Corporation** (US shares AAUC, Canadian shares same ticker) is notably wide because of concentrated geopolitical, regulatory, and macro risk. Chinese company **Zijin Mining** agreed to buy it all-cash when gold was peaking in Q1 at C$44 / US$32 a share; the stock now trades ~$23 US and ~$33 CAD. Why so wide? (1) **Chinese outbound regulatory approvals + extended timeline** — China will likely approve it (it's the world's biggest gold buyer), but the closing date, after clearing Canada's Investment Act (ICA) in May, was pushed from May 29 to July 29, 2026 (three weeks out); MOFCOM/SAFE/NDRC filings are outstanding and could extend further, and arbs hate extensions. (2) **Extreme West-African exposure** — unlike Alamos's tier-one Canada/Mexico assets, Allied's mines are in Ivory Coast, Ethiopia, and Mali (Sadiola in Mali, Bonikro/Agbaou in Ivory Coast, Kurmuk in Ethiopia). Mali is under a military junta; resource nationalism is rising across West Africa; arbs fear local governments hold up the deal to demand tax penalties or carried interest. (3) **Execution and financing squeeze** — Allied reported a $58M Q1 loss (growing quickly), higher all-in sustaining cost than African peers, and its Kurmuk mine is in a critical capex phase toward first gold. (4) **Gold-sector volatility** — gold sold off ~20% in a quarter, so will Zijin pull out or restrike? It's a strong cash offer and Zijin has the cash; the stock has sold off to roughly its pre-announcement level, so the risk-reward is quite decent if you look past the fear.

## S&P Global Spinoff and Hedge Fund Activity - 32:39

On Wednesday **S&P Global** completed the separation of its mobility division into an independent public company, **Mobility Global**, trading regular-way under ticker **MGGL**. We think this spin is good for the S&P stock — it makes it more of a pure play.

**Lee Robinson's** new hedge fund **Altana** (which took $20M and rallied 1,000% shorting banks) is launching a new fund to short life-insurance companies — names like **Lincoln National (LNC)** and **MetLife (MET)** with too much private-credit exposure that aren't regulated as closely as US banks. (I warned about life insurers a few months ago.)

Several firms imposed caps: **Blue Owl (OWL)** — its OCIC fund ($34B assets) saw 18.8% redemption requests, capped at 5%; Blue Owl Technology Income Corp saw 38% redemptions, capped at 5%.

## US Market Q2 2026 Recap and Analysis - 34:18

A historic Q2 for global equities, plus the largest-on-record FTSE Russell reconstitution. Recent performance and enthusiasm, combined with scrutiny around AI capex and monetization, has created a more complicated environment.

Broad indices: emerging markets +24% (oil falling), Russell 2000 +21%, Russell 1000 Growth +16%, S&P 500 +15% — a very strong quarter. Semiconductors +88% (100% YTD), GS AI data centers +70% (115% YTD), S&P 500 AI winners +65% (62% YTD), non-profitable tech +43% (48% YTD, the "momo"), high momentum +31% (56% YTD, sold off hard Friday), Mag 7 only +11.7% (−1.7% YTD), hyperscalers +10% (−6% YTD), S&P 500 ex-AI only +7% (6% YTD).

MSCI All Country +15%, the best quarter since Q4 2020 — remembered for AI-buildout enthusiasm. The S&P 500 posted nine consecutive weekly gains and +17% cumulative from early March to early June (second-best streak since 1950). The S&P's 15% total return ranks in the 93rd percentile of quarterly performance since 1950. We remain constructive on earnings momentum but acknowledge a confluence of tailwinds and a potential pull-forward, which with summer seasonality may lead to digestion in H2.

The dominant driver was earnings momentum — reported growth topped 20% year-over-year in Q1, and 2026/2027 EPS estimates rose more than 6% during the quarter despite an oil shock. But earnings growth was concentrated in three AI-capex beneficiaries — semiconductors, capital goods, and technology hardware/equipment — which combined for 65% of the S&P's quarterly return. AI plays outperformed non-AI by ~60% during the quarter; semis returned 88% while the average S&P company returned ~11% and hyperscalers lagged at 10%.

## Market Bifurcation and Concentration - 39:26

Notable weakness in June closed the quarter. Beyond the bifurcation, speculative areas like non-profitable tech and trailing-momentum names returned 40% — elevated risk appetite. A market of haves and have-nots: just 10 companies (32% of the index) accounted for more than 60% of the quarterly return; only 30% of companies outperformed the index and 33% actually declined. Very difficult from a trading perspective.

## Index Reconstitution and Style Blurring - 40:16

The Russell reconstitution took on greater significance this year given the magnitude of changes and blurring lines between market segments. Starting with Russell 1000 Growth vs Value: the Mag 7's capitalization is migrating from growth to value. This began a year ago when Alphabet, Meta, and Amazon were classified as both value and growth (included in both) — dangerous, because both indices then hold big tech.

This year: Alphabet moves entirely back to growth (after a short stay in value); Amazon moves almost entirely to value (showing it's cheap); Microsoft and Apple see half their weight shift to value. FTSE Russell's formula assigns growth/value by historical growth rates and current valuation, with a mandate to keep an equal split of large-cap market cap between the two. The market-cap expansion of the largest growth companies raised the bar for "growth," pushing many into value. The Mag 7 is now 17% of the Russell 1000 Value index — a surprisingly large allocation for a group that never exhibited value traits (slower growth, cyclicality, capital intensity) — though capital intensity is now changing. This broadens the investor universe but brings the Growth and Value indices closer together stylistically, reducing diversification benefit.

Large vs small cap: 43 companies are graduating from the Russell 2000 in this rebalance. The maximum market cap for small-cap qualification keeps rising, now $5.7B (up from $1B post-GFC, more than double the $2.9B average since 2000); the weighted-average market cap reached $7.5B (from $2.1B a decade ago) — which 10 years ago would have been larger than 30% of the S&P 500. The 43 graduates are many key AI-capex beneficiaries — this cohort largely carried the Russell 2000 this year (nearly half its total return). Their graduation, like a championship team losing its best players, leaves the index searching for new contributors, potentially less fundamentally attractive. The small-cap index will likely rely more on traditional drivers (easing financial conditions) with less tech.

## Second Half of the Year Outlook - 45:54

Global equity markets are at an interesting intersection after a historical Q2 — elevated enthusiasm, most pronounced in highly cyclical areas with strong recent momentum and the most leverage to the AI-infrastructure theme. Underlying earnings momentum remains a key pillar, but this environment warrants a more disciplined H2. Crowded positioning, high leverage, lofty expectations, and historical outperformance from high-momentum have historically opened the door to rapid resets. We remain constructive on equities but believe investors should prioritize durable earnings growth and use rebalancing and diversification — not crowd into a few high-growth names.

## Tesla Sales and JPM Software Quadrant - 47:04

**Tesla (TSLA)** sales recovered on the oil crisis: deliveries ~480,000 versus expectations of 406,000 — likely because China and Europe were afraid of the oil crisis. We'll see how it fares in H2.

JPMorgan's software analyst has a great quadrant graph. Consensus longs: Microsoft, Datadog, Snowflake, CrowdStrike, Palo Alto, Cloudflare, MongoDB (already performed). Microsoft moved into "battleground" and became a buy this quarter. They're highlighting Autodesk and even Palantir. Consensus shorts: names people think are easy to replace (which isn't true) — Figma, Adobe. Hedge-fund hotels include Twilio and Zoom (Zoom because of its Anthropic exposure) — be careful with hedge-fund hotels.

## AI's Impact on the Stock Market - 48:24

Micron and Nvidia are estimated to account for 40% of S&P 500 EPS growth this quarter — unbelievable. The stock market is effectively two asset classes: AI stocks and non-AI stocks. There are 41 AI-related stocks in the S&P 500 (Broadcom, Oracle, Palantir, AMD, Salesforce, Uber, Qualcomm, Arista, Adobe, Micron, Palo Alto, Intel, CrowdStrike, etc.), about 45% of the index.

In the May 2022 Flow Show, B of A's Michael Hartnett showed the market is the most concentrated in a single theme in 150 years — the last comparable was late-19th-century railroads. AI and non-AI stocks are diverging sharply. This matters for two reasons: (1) index investing — nearly half of index holdings are now AI; (2) market signaling — since the Iran war began February 28th, the S&P is up 7% in four months (implying a healthy economy), but the 459 non-AI stocks are only up 1.5%, a very different economic read.

## Memory Chip Lawsuit and Market Dynamics - 50:37

Samsung, SK Hynix, and Micron were sued over alleged DRAM price-fixing amid record memory costs. I don't think they have a case, but the suit claims coordinated curtailment of DDR3/DDR4 production since 2022, creating shortages; these three control ~90% of global DRAM, and laptop and Apple prices have risen because of it.

**CXMT** is the biggest DRAM producer in China now, aiming to add capacity quickly and pressure SK Hynix/Samsung/Micron. But it lacks ASML lithography equipment (US export controls), faces manufacturing bottlenecks across process steps (progress in one doesn't solve others), and trails incumbents by several generations in DRAM process technology; the HBM gap is even larger. Its presence is concentrated in China with very limited global share; broader adoption depends on quality, qualification, geopolitics, and customers' willingness to diversify from the three main suppliers.

## The Memory Chip Supercycle - 53:24

What does this mean for the supercycle? Some may misread CXMT as purely negative for memory — that misses the point. Even as CXMT adds supply, the DRAM market is still extremely constrained and supportive of suppliers near-term; CXMT may struggle even to satisfy domestic Chinese demand, let alone flood the market. And its memory isn't cheap — Chinese memory pricing has also gone through the roof, tracking the global DRAM strength; CXMT is benefiting from the same shortage pricing, not acting as a deflationary force. View it as a long-term structural competitor, not a cycle killer. Near-term DRAM undersupply is too large for CXMT's incremental output to loosen. The supercycle remains defined by constrained supply, rising memory content, HBM wafer absorption, and accelerating AI demand. DRAM prices are still expected to rise from here; NAND prices are expected to peak next year.

## Micron Financials Deep Dive - 57:08

We used a new AI tool to build the Micron chart book — cumulative returns and catalysts, revenue by segment (mostly DRAM, prices still rising) and geography (US growing on AI), gross-margin contribution by product, forward P/E (all of Micron's growth has come from earnings, not multiple — multiples compressed), and earnings growth vs multiple contraction.

Micron delivered a historic beat-and-raise, but the stock was still a sell-the-news by week's end. Results were driven by an explosive ASP increase across DRAM and NAND on AI demand far outpacing supply. The stock rallied 16% on June 25th then sold off post-release. Key metrics: revenue growth 346%; DRAM ASPs +250% (about half of that price); gross margin 85%; all four business units posted record revenue; data-center revenue above $25B; HBM4 revenue surpassed $1B; data-center SSD revenue exceeded $5B (more than doubling).

Earnings drivers: the ASP surge, plus **strategic customer agreements — 16 SCAs with four large customers averaging five-year terms, covering 20% of DRAM volume and 33% of NAND volume, with $22B in cash deposits**; accelerating HBM ramp; supply tightness that is structural, not cyclical. Structural constraints: long fab construction lead times, HBM's 3–4x higher wafer consumption vs standard DRAM, and suppliers prioritizing DRAM over NAND. Capital allocation: free cash flow of $18.3B; a 30% dividend increase; intent to return 100% of excess cash via buybacks beginning December 2026 when CHIPS Act restrictions expire. FY26 capex was raised to ~$27B; FY27 capex is expected to exceed ~$10B per quarter — insane. On the call, analysts probed the SCA ceiling-floor pricing: management said floor prices yield gross margins well above prior cycle peaks, and declined to project when supply catches demand. Management said CXMT mostly sells within China.

## Post-Selloff Market Dynamics - 61:02

After the AI sell-off, Samsung said it would raise Q3 DRAM prices by another 20% — which could spark a fourth AI-name rally, because it signals a near-term supply shortage. That's why memory stocks and the Nasdaq (+1.4%) bounced overnight. Samsung's DRAM ASP increase is especially pronounced relative to SK Hynix, so maybe SK Hynix raises next ahead of its US IPO. Despite Meta scaring the market about selling excess data-center capacity, consensus is Meta still increases capex next call. UBS raised its DRAM/NAND price forecast and sees shortages lasting until 2028; it sees no meaningful impact from Meta's comments (HBM procurement intact across Nvidia, Google, AMD, Meta), and views the memory pullback as positioning, not deteriorating fundamentals — which is what we said: retail euphoria (Korean single-stock levered ETFs, options volume) meant a pullback was due, and it wasn't small (SK Hynix −30% in a week).

Chinese smartphone brands Xiaomi, Oppo, and Vivo cut shipment targets ~30% because of the semiconductor shortage. The semi sell-off continued Thursday (Micron down to ~992; Friday was a holiday). **Anthropic** said it's in talks with Samsung to manufacture its custom ASIC (like Amazon's Trainium and Google's TPUs), so Nvidia and AMD sold off. SK Hynix (000660 Korea) is down ~30% from the local high but still up 17% since May.

## Yen, Iran, and Investment Moves - 64:17

The yen (page 39) fell to the lowest in 40 years — bizarre because volatility was so low on the way down (40-day realized vol just over 3.75). The drivers all push higher but the market is too scared to act. Filtering since 1970 for times USD/JPY was at least 20% above its 10-year average and at 10-year highs threw up a handful of days in 2015 and 58 days since 2022; today's 40-day vol is the lowest of the bunch. The opposite exercise (10-year lows, 20% below average) generated 172 observations with average 40-day vol of 10 (min 6). Sterling and Swiss franc show the current low yen vol is anomalous — the lowest reading for these extremes. Speculators have built the second-largest yen short position in history.

Iran: on Monday the White House said Witkoff and Kushner will attend a Doha meeting with technical talks on the sidelines — a 60-day ceasefire, still needing agreement on the biggest item (the Strait tolls). The market isn't too concerned with WTI at $68.50 and Brent at $71.80. One concern: strategic petroleum reserve weekly stocks are the lowest since the 1980s, so a resumption of war would not be good.

**Druckenmiller** added to one of the names we discussed and liked last week — **SE (Sea Limited)** at today's valuation. Because he's the GOAT, worth mentioning (his 13F is a couple months old). **Michael Burry** disclosed new shorts: he was right on Palantir (down 50% from where he shorted it), and disclosed new shorts against **Nvidia, Tesla, Caterpillar, and Applied Materials**, calling the semiconductor index a pure form of overvaluation rarely seen. He shorted Caterpillar for the first time ever after the stock surged 86% in H1 even though sales aren't growing that fast.

M2 money supply hit a record, and surprisingly gold and crypto sold off despite that. Year-to-date maximum drawdowns from the high: 9% S&P, 13% Nasdaq, 11% Russell, 10% Dow.

## Fed Rate Path and Market Warnings - 67:33

Traders now see 2027 cuts after the poor labor data; rate hikes peaked at end of May and are down a bit. After 27% year-over-year Q1 earnings growth, Q2 S&P earnings growth is expected at 22% — the highest estimate heading into earnings season since 2021 (Goldman). Is the Fed pricing too hawkish? Since Warsh's first appearance, OIS settled on only 1.5 hikes by year end (it was as high as ~3 at one point; I never believed that, and B of A was wrong). His pivot in Europe seemed to surprise some. Have we gone too far the other way? PCE was benign, the labor market looks stable rather than overheating, NFP was quite bad, and the oil drop should feed through headline inflation. Warsh's preferred trim-mean measure strips out the most extreme moves (Apple's 20% Mac/iPad hikes, Netflix's 8–12.5% rises) — the Dallas Fed trim-mean PCE shows inflation coming down sharply, even as core services ex-housing was up a lot this year and is probably coming down this month. On the other hand, Citadel Securities says investors underestimate Warsh's commitment to lowering inflation, warning higher rates weigh on risk assets — but we already saw that with the tech sell-off. It's still possible we see one hike this year (bad for the market), but the market would say "one and done."

## BIS Warning and the AI Debt Debate - 69:xx (page 44)

The BIS issued one of the sharpest warnings yet on the debt behind the AI boom — several hundred billion raised for AI data centers where you don't even know the payoff. Circular financing adds a weak point (chipmakers, hyperscalers, AI labs, and compute providers fund each other while booking future sales from each other), making real demand harder to read. Hyperscaler bond issuance topped $100B in 2025, while off-balance-sheet vehicles shifted data-center obligations toward private credit. An AI-capex slowdown could hit suppliers first, then credit markets, then households (US stocks are ~64% of the MSCI Global Index; household equity exposure is at all-time highs). Private credit raises systemic risk — direct lenders quadrupled AI/IT exposure in five years to ~15% of portfolios. America still dominates AI venture capital (~$200B last year vs $15B Europe, $13B China, $13B UK).

Macro trader **Matt Dratch** takes the other side: the AI boom is justified. Bears compare AI capex to the late-1990s telecom fiber bubble, but that's harder to defend once AI revenues consistently exceed the economic cost of the assets (we're starting to see that) — though it could change quickly if more US companies use Chinese open-source models. Once management teams show incremental infrastructure earns attractive returns, incentives shift from caution to acceleration (as with early AWS/Microsoft Cloud). AI revenue has overtaken AI capex depreciation. Memory is only 18% of total data-center build cost (was 2% in 2021, up 9x in price), vs logic (Nvidia) 42%, cooling 10%, power 15%, building 15%. Nvidia's chips consume more memory each generation (H100 → H200 → B200/B300 → Rubin).

## Apple's Memory-Cost Hit - 71:xx (page 48)

DRAM is short-term working memory (in iPhones/Macs/iPads); HBM is the stacked, faster version used in AI accelerators. **Apple (AAPL)** is now asking Washington for permission to buy DRAM from CXMT — one reason memory names sold off last week. CXMT sits on the Pentagon's 1260H Chinese military company list, so Apple must ask permission; Apple fears Commerce could upgrade CXMT to the Entity List (a total trade ban) that would paralyze iPhone 18 assembly lines. Apple sold off because of this unprecedented supply-chain crisis — chipflation and factory cannibalization. HBM requires far more silicon wafer capacity and complex steps than the low-power DRAM used in phones/laptops, so Micron/Samsung/SK Hynix diverted lines to lucrative HBM and enterprise server DRAM, starving consumer tech. Morgan Stanley says average memory prices skyrocketed 6x over the past year; Jefferies warns DRAM prices surge another 40–50% in Q3. Tim Cook described it as a "once-in-a-century flood." Apple lost $263B in market value — its largest drop in over a year (−6% in a day) — after a blanket 20% price increase across MacBook and iPad lines (base iPad Pro $999 → $1,199). Samsung and SK Hynix said they could invest over 1.3 trillion (won) over 10 years. Micron's long-term contracts are structured much better than SK Hynix, which is why SK Hynix is more sensitive to the sell-off. Goldman's high-beta momentum index was down 18% last week, the biggest since COVID.

One risk to the AI trade: Chinese models charge as little as 18 cents per million tokens vs $4 for top models (95% below Anthropic), per Citi. Open-source processing on OpenRouter rose to 65% in June vs 34% in January; DeepSeek gained attention on much lower token prices. Cheaper AI is a new enterprise priority as usage-based bills turn model choice into cost control. Gartner estimates AI coding costs pass the average developer salary by 2028.

The **SK Hynix IPO** prospectus (page 53) will be exciting. **Bloom Energy** (page 54) sold off with the Russell rebalance (big in Russell 2000, small in Russell 1000). The **SMCI (Super Micro)** scandal: Taiwan's Keelung District prosecutor expanded its investigation into illicit diversion of advanced semiconductors to China — detentions after raids across 12 locations (six private residences), two local Super Micro employees detained. Super Micro is uninvestable at this point. Individuals under investigation expanded from three to nine; offices of Albatron and Chief Telecom were raided; the group is accused of forging shipping documents to smuggle ~50 high-end AI servers with Nvidia chips to China, Hong Kong, and Macau (one shipment routed through Japan). Exporting chips to China wasn't a criminal offense under Taiwanese law, so prosecutors used document forgery and customs fraud; Taiwan's ruling DPP is drafting a Foreign Trade Act amendment to criminalize the diversion of restricted silicon with broader seizure powers.

## Sable Offshore (SOC), Mag 7, and Meta's Strategy - 81:12

**Sable Offshore (SOC)** raised capital to repay its Exxon term loan, and the stock sold off on the equity-and-convert raise. The original term loan was ~$956M including PIK (deferred) interest; they paid a ~$30M amendment fee on June 22 to push maturity out a month, then raised a $675M high-yield term loan B, $345M of convertibles, and $115M of stock — sending the stock to ~$308... i.e. ~$3.08 at the lows. It has now raised capital three ways and extended maturity, eliminating near-term default/bankruptcy risk from Exxon, so shares rallied toward the convert exercise price near $4 (a 30% rally from the lows). The stock needs momentum to get above $4; maybe a put-selling opportunity (we're already long via sold 2/2.5-strike puts at high premium, ~70–80% return on capital).

Will Mag 7 cash flow recover? If you believe sell-side estimates, the Mag 7 is very undervalued because AI spend translates to revenue growth — but we don't know that. Free cash flow fell off a cliff in 2025–2026 on ~$725B of capex this year alone. Amazon's FCF fell 95% in Q1 ($44B in a quarter); Meta's FCF compressed against $19B quarterly capex, with FCF turning negative by 2027; Alphabet guided capex as high as $185B. Consensus 2028–2030 FCF explodes (no one can predict it). AI-related debt issuance is skyrocketing (IG gross supply and private credit, page 57).

There was relief from **Meta (META)** after it announced selling excess compute and becoming a cloud company (competing with SpaceX, Amazon, Microsoft) — but it's probably a small amount of compute, not a full pivot, which is why the stock rallied high-teens then sold off the next day. AI super-intelligence is still Meta's top priority; it's committed hundreds of billions to data centers but hasn't executed well. Nebius and neo-clouds sold off on fears they'd compete with Meta (Nebius −15%, CoreWeave −13%, Hut 8 −11%, WYFI −11%, Cipher −9%, Core −8.25%, TeraWulf −7.8%, IREN −6%, Applied Digital −4%). Zuckerberg told employees Meta's AI bets haven't come to fruition, just like the metaverse. Meta is committed to up to $245B, mostly AI infrastructure this year, while cutting 10% of its workforce. SemiAnalysis: with Bloomberg headlines suggesting Meta could become a neo-cloud, CoreWeave and Nebius sold off on overcapacity fears — but Meta is doing this so the market doesn't penalize it, because it thinks capex will accelerate in 2027 and be shockingly high (it contracted for over five gigawatts of cloud/colo capacity in H1 alone, excluding self-build). Not a full neo-cloud shift.

Chip and memory stocks took a beating: SanDisk −16%, Micron −14, Western Digital −11, Seagate −11, Applied Materials −7, Arm −7, Intel −6, Broadcom −4.

## Q&A - 87:16

**Robinhood (HOOD)** — 87:16 — Rebounded from $70 to $112 (year highs ~$150). It does ~$2B operating income, ~$101B market cap, ~41% revenue growth (why the multiple is so high), but growth slows to 13% this year and 25% next. ~$2 EPS at $112 = ~55x P/E (~50x operating earnings) — not cheap. It's taking share and announced a workforce reduction, and LTM growth is there, but I don't think it has a lot of upside from here; the 10% headcount cut is priced in and retail growth probably peaked. I'd like it in the 70s/80s. Don't like it here unless growth accelerates or they add new products.

**Meta renting out excess compute** — 89:09 — We think they're doing this so investors don't penalize them more, because there's a strong chance Meta's capex actually goes up. It's a tactic ("we're smart fiduciaries; if AI doesn't succeed, we have this option"), but in reality they'll spend hundreds of billions on AI — Zuck hasn't given up, so we can't rely on the excess-capacity narrative.

**Gundlach and gold** — 89:55 — Gundlach says he'll buy gold at $3,500 with both hands. We've been buying more gold; I'm already over 10% in gold and silver, and I wouldn't go more than 15% for diversification. I'll stop the call there — happy Fourth of July.