# Special Situations Report — Weekly Call (readable notes)

- **Title:** Big Mag7 earnings week (AMZN, AAPL, META, MSFT), Nvidia's $250B guarantee for OpenAI, an AppLovin (APP) add, the Betterware de Mexico (BWMX) special-situation add, and a temporary Iran peace — plus the Taco Index calling the de-escalation, S&P short interest at 3.7% of free float (highest since 2010), the Alphabet (GOOGL) deep-dive (record net income but the first negative FCF quarter since the IPO, $811B of purchase commitments, and the "unhealthy search revenue" ad-quality teardown), Dan Niles' mega-cap preview, the new Section 301 tariff regime, the Utz (UTZ) buyout, hyperscaler FCF incineration, and a Q&A on semis-on-dips, the EchoStar/SpaceX pair unwind, agency REITs (DX/RWT/RWTS/NLY/AGNC), the Clarity Act, IBM and RSP-over-SPY
- **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-26 (Sunday call; the deck/transcript title it "2026-07-26 - SSN")
- **Source URL:** the premium recording is distributed via the SSR Discord + email (no public YouTube video; no Discord permalink captured for this week)
- **Source PDFs (this folder):** `transcript.pdf` (premium transcript), `report.pdf` (the condensed report + actionable items), `7-24-26.pdf` (the full weekly strategy deck — the call walks it page by page: p4 short interest, p5 Alphabet FCF, p37 hyperscaler CDS/gilt yields, p38-39 Tesla capex/negative FCF, p40 Roper, p43 the AppLovin model, p46 ServiceNow, p47 South Korea's leveraged single-stock ETF rules, p54 end of Jensen Huang's open-weights essay, p58 global consensus growth, p59 BlackRock on rates), `Applovin Valuation Model and Risk Framework.xlsx` (the APP model referenced as "the model on page 43"), and the three Q3 macro outlooks read at 52:25 — `BlackRock Q3 Fixed Income Market Outlook.pdf`, `MS Q3 Market Outlook.pdf`, `State Street Q3 Market Outlook.pdf`
- **Note:** readable notes auto-extracted from the premium PDFs; fillers (um/uh/you know/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 (the Ref column is the section time as plain text). Auto-transcript garbles corrected to the right entity and listed here: **"SSN" = SSR** (Special Situations Report); **"Besant" = Bessent** (US Treasury Secretary); **"Li Puan" = Lip-Bu Tan** (Intel CEO); **"Varin Rubin Ultra" = Nvidia's Vera Rubin Ultra** (the platform Nebius deployed in Finland); **"SKHY" = SK Hynix**; **"Echo" = EchoStar (SATS**, renamed "ECHO"**)**; **"the RWTS / the Redwood S" = the Redwood Trust baby bond (RWTS)**, distinct from the RWT common; **"Dynex" = Dynex Capital (DX)**; **"agenic / Gentic AI" = agentic AI**; **"open claw" = OpenClaw** (the agentic-AI formalization he dates to January 30); **"sword about 90%" = soared ~90%** (Utz on the Intersnack bid); **"anything below 13 here is kind of Bible" = buyable**; **"Max7" = Singh's own term for the Mag7 / mega-cap cohort**; **"Taco Index" = the Signum Global Advisors composite** (Andrew Bishop); **"CXMT" = ChangXin Memory Technologies** (the private Chinese DRAM maker); **"USSTR" = USTR** (Office of the US Trade Representative); **"sectorial" = sectoral**; **"sanguin" = sanguine**. Two garbles left deliberately **untickerized**: **"SCX was up"** in the memory-rally list is most likely **Seagate (STX)** but the spelling is too far off to trade on, and **"Sandisk"** in the same list is SNDK. Internal inconsistencies in the source (both left as spoken, flagged here): Alphabet's cloud revenue is given once as "$24.7 billion up 82%" and later as "$27.5 billion up 82%" (the report uses $27.5B), and its negative FCF quarter is called both "the first since 2016" and "the first since going public more than 20 years ago" (the report and the surrounding commentary use the latter). The transcript footer's "Date of Transcript: 2024-05-21" and the "2026-01-26" filename are stale template artifacts — the real call date is **2026-07-26** (Sunday).

---

2026-07-26 - SSR - Big Mag7 Earnings Week (AMZN, AAPL, META, MSFT), NVDA 250B Guarantee for OpenAI, APP Add, BWMX Add, Temporary Iran Peace

## Introduction and Nvidia AI Data Center News - 00:00
- Breaking minutes before the call: **Nvidia (NVDA)** is in talks to **guarantee a ~$250 billion AI data-center build for OpenAI** — "one of the biggest AI data center guarantees ever."
- That, plus a **temporary halt to missile launches** between the US and Iran over the last two days, "is probably going to result in some strong short covering."

## Market Overview and US/Iran Conflict Impact - 00:47
- Pre-open tape: Nasdaq +1.36%, S&P +0.8%, Russell small-cap +1.2%, crypto rallying, **US 10-yr down 5 bps** (risk-positive), **crude −5%**.
- Caveat: "we've seen these rallies fade overnight if there's anything said negatively about the Iran conflict, but at least for now this is looking very positive for tomorrow." (Later in the call the Nasdaq gain tapers to ~1.1%.)

## Q&A and Attachments Note - 1:28
- Questions go into the SSR Q&A tab; all attachments are in the email (the deck + the three Q3 macro outlooks + the AppLovin model saved in this folder).

## Economic Calendar and Earnings Week Preview - 01:51
- Week's data: durable goods, consumer confidence, personal income & spending, **PCE inflation**, UMich sentiment.
- The 10-yr was up 5 bps last week, pushing **July rate-hike odds to 40%** after the Iran re-escalation. "Bessent was probably crying in the bathroom at the end of last week" — then the weekend changed it.
- **Real** 10-yr yields hit their highest since **2023**; initial jobless claims the **lowest since 1969**, which had supported long-end yields.
- "The biggest earnings week of the quarter": AMZN, AAPL, MSFT, META, QCOM, WDC, LRCX, FTNT, ARM — plus non-tech at scale (V, KO, BA, GLW, SPGI, PYPL — "under an M&A bid", SHEL, ABI, YUM, ABBV, CVX, SONY, ENB, MRNA, SIE).

## Short Interest and the Taco Index - 03:43
- Deck **page 4**: **S&P 500 short interest ~3.7% of free float — one of the highest levels since 2010**, i.e. hedge funds pressing shorts into a market near record highs. That's the fuel for a short-covering rally.
- The **Taco Index** (Signum Global Advisors, led by Andrew Bishop) combines four inputs: **Brent crude, the 10-yr yield, Strait-of-Hormuz crossings and the S&P level**. Backtests show a combined **2.3–3.4 standard-deviation move (avg 2.9σ) has historically forced presidential action** to push energy prices and yields down.
- Signum's window: a "taco" **no earlier than July 26, no later than July 30** — the de-escalation landed **July 27**. "Almost right on the ball."

## Google Earnings Analysis and Hyperscaler Spending - 05:24
- **87% of Alphabet's record $112B Q2 net income came from markups on equity investments** (SpaceX, Anthropic). Ex-markups, net income was up only ~2% sequentially — hence last week's selloff (shares −6% Thursday).
- Deck **page 5**: capex guidance raised to **$195–205B** (from ~$190B); **Q2 free cash flow −$5.9B, the first negative quarter since going public** (>20 yrs; the transcript also says "since 2016" — an internal inconsistency).
- Cloud **$27.5B, +82% YoY** but at a lower margin; **backlog $514B** (from $460B), majority recognized as revenue within 24 months. Search ads **$63B** (slight miss). **YouTube $11B** (beat, snapping a multi-quarter losing streak). **Other Bets $382M** (miss). Q2 capex **$45B** vs $44B est. Revenue ex-TAC **$103.6B** vs $101B est; EPS **$9.11**.
- Investment portfolio: ~$100B of quarterly gains; **$94B SpaceX stake, $80B of it under short-term sale restrictions** ("they can't sell").
- **Purchase commitments ballooned to $811B — up $500B in three months** (chips, data centers, electricity, inventory, content licenses); ~$201B short-term. "Which scared the street."
- Gemini MAUs **950M** (above estimates, below Q1). Gemini 3.6 Flash expanding, **Gemini 3.5 Pro delayed**, development redirected to **Gemini 4** with a roughly monthly release cadence.
- The four hyperscalers telegraphed **$725B of 2026 AI spend** in April — "if Alphabet is any indication, the sum will be larger by the end of this coming week."

## Dan Niles' Market Outlook and Max7 Analysis - 13:57
- Scoreboard for Niles' prior week's call: **Max7 spenders −6%** (one huge down day on Google's print) while the **semiconductor index +1%**.
- Oil was +8% last week, now −5%; 2s/30s yields +9–15 bps last week, now −5 bps — "which is why we're seeing that countertrend rally." **November 3 midterms put a powerful impetus on the administration to get oil down** (oil was ~$70, spiked past $100, now mid-$80s).
- **Three positives:** (1) agentic AI — "the advent of agentic AI on January 30 with the formalization of OpenClaw" — drives **10–100× more token production/consumption**; (2) microprocessor vendors benefit as **GPU:CPU ratios move from 8:1 toward 1:1**; (3) hyperscalers "cannot afford to lose the AI race," which keeps him **more bullish on the recipients of capex than the spenders**.
- **Three negatives:** (1) the shift from **token maximization (March) to token minimization (June)** as enterprises adopt open-source Chinese models — "even Microsoft is"; (2) that AI cost control comes **at the expense of non-security software companies, IT services and headcount**; (3) the **cost of money** — wider credit spreads (Oracle et al.) and compressed equity multiples raise the cost of capital. **11 major central banks** went from cutting (2.7% Jul-2025 → 2.4% Feb) to **raising, averaging 2.6%**; their 10-yr sovereign yields went 3.2% → **3.8%**.
- **META** — World Cup engagement helps, but Q3-25 revenue grew 27% vs 22% in Q2-25 = a tough comp; a public cloud / LLM API launch could push **capex guidance higher** (a negative). Valuation in the teens vs the S&P's ~22× 2026 PE limits downside.
- **MSFT** — Azure has a high bar after GCP accelerated 63% → 82%; **Microsoft 365 Copilot** is the AI-cost-control tailwind (native to where enterprise work already happens); but Microsoft owns **27% of OpenAI**, which Niles is negative on — "caught between Google in consumer AI and Anthropic in enterprise."
- **AMZN** — AWS also has a higher bar; higher oil pressures e-commerce logistics and consumer purchasing power; the four-day Prime event moved to late June **borrows Q3 demand into Q2**.
- **AAPL** — benefits from *others'* AI capex because it isn't spending on its own (Google is its partner). But **Q3/Q4 revenue and margin estimates are too high given semiconductor/memory cost inflation, and 37× PE is expensive — "he's not a buyer of Apple here."** Longer term still bullish on the foldable upgrade cycle.

## Deeper Dive into Google's Pricing Strategy - 19:12
- Max Anderson (spends ~$500k/yr on Google Ads): search revenue growth is "artificial and extremely unhealthy… mostly them raising prices while they lose eyeballs to LLM search."
- **(1) Search volume cannibalization** by non-monetized LLM queries.
- **(2) Deprecation of the second-price auction.** Historically, bid $5, next-highest bids $1 → Google charged **$1.01** (a penny above the second-highest bid), which incentivized advertisers to bid their true maximum. Google **silently deprecated it** and now charges **as much as your bid/budget caps allow** regardless of competing bids.
- **(3) Exact-match dilution.** "Close variant" now means whatever Google wants, so ads run across thousands of unrelated keywords that the targeting settings explicitly forbid — with no way to turn it off. "Exact match is broad match and broad match is meaningless spam."
- **(4) Budget caps ignored by ~2×** — $1,000/day campaigns suddenly spending $2,000, no refunds or recourse.
- The verdict: "These are the actions of a company whose core business is in decline but desperately needs to pump quarterly earnings… At the altar of AI capex, Google is sacrificing its golden goose of search." Google ran as a **benevolent monopoly for 25 years** — capturing a small fraction of the value it created — "this is no longer the case."

## Special Situation: Betterware de Mexico (BWMX) - 23:38
- **BWMX** — a Mexican direct-selling household-products group (furniture, sinks, electronics, utensils, laundry, cleaners, bath, personal care). "Very, very compelling after an acquisition."
- **The core disconnect: operational earnings surging while the multiple shrinks.** EBITDA ~**2.0B pesos (2024) → ~4.3B pesos (2027E)** — a double. ~3B pesos of free cash flow on an ~11B market cap ⇒ a **high-20s% FCF-to-equity yield**.
- NTM EBITDA estimates **+50% over the last year while EV/EBITDA fell 30% → 3.8× 2027E EBITDA**. Converts **>80% of EBITDA to free cash flow**.
- **Dividend ~8.6%** (≈$0.36/share quarterly, 250M pesos), **26 consecutive quarters** of payouts.
- **Tupperware LatAm accretion:** in its first consolidated month it contributed **11% of net revenue and 16% of EBITDA**; management estimates pro-forma trailing-12-month **EPS 36% higher** with Tupperware integrated. Leverage **1.6× net debt/EBITDA**, down from a brief post-close 2.6×.
- **Why the mispricing:** (1) the **direct-selling discount** — the Street habitually applies 3–6× EV/EBITDA to direct-to-consumer selling networks on perceived distributor-churn risk; (2) emerging-market/FX (peso, real — a tailwind here); (3) **M&A digestion** — the deal just closed, so people are waiting to see the cost savings.
- Q2 print: revenue **+16.8%**, EBITDA **+25%**, EBITDA margin **+140 bps**, net income **+20%**, diluted EPS **+14.9%**. Betterware segment revenue +~4%, Jafra +~4.5%. Q2 FCF 578M pesos (~74% EBITDA conversion). **2026 guidance raised from 4.8% growth to 18–22%**, EBITDA margin held **>19%**.

## Market Momentum, Hedge Fund Positioning, and US/Iran War Update - 28:23
- **"We also added to app"** — **AppLovin (APP)**, ~**50% upside**; strong adoption from e-commerce companies testing the ad engine as it enters a new vertical. (The model is on deck **page 43**; the spreadsheet is saved in this folder.)
- UBS: the momentum sell-off may be near its end — hedge funds cut long positions in momentum/semis by **~5% of gross market value**, one of the largest reductions on record. "Much of the speculative positioning has already been unwound."
- The remaining risk is the pressure on companies to justify AI spend — **Tesla (TSLA) −14% last week** on accelerating capex + a profit miss; "Elon spent some time on the call thanking Micron for selling him memory chips at a decent price and the market didn't really like that."
- **Morgan Stanley predicts 25% memory price rises on AI demand; Micron's HBM production is sold out through 2027.** On July 21: Sandisk +9%, SK Hynix +9%, **WDC +8%, Micron +8%** — "we took that opportunity to trim a little bit and then we added back at the end of this past week, and it looks like Monday's going to be a good day."
- Since mid-June hedge funds **reversed 80% of their cumulative net buying** in semis/semi-cap equipment — but **started buying again last week** ("also a good sign"). Global semi net allocation: 10% at the start of the year → **24% record in June → 19%** ("healthy correction"). US semis in the Goldman prime book: 7% → 14% → **11%**.

## New US Tariffs - 31:08
- **Why the regime changed:** the Supreme Court struck down the post-Liberation-Day tariffs in **February 2026**; the White House used a temporary **10% global tariff under Section 122** of the Trade Act of 1974, statutorily capped at 150 days, which **expired July 24**.
- Replacement: **USTR invoked Section 301**, justified on **forced-labor prohibitions** in supply chains ("which is unreal — most of these countries do not have forced labor at all"). ~60 trading partners covered.
- **10.0% tier** (20 partners judged to have taken legal steps on forced-labor import bans): UK, EU, Canada, Mexico, India, Argentina, Bangladesh, Cambodia, Indonesia, Malaysia, Pakistan, Sri Lanka. **12.5% tier** (non-compliant): China, Japan, South Korea, Australia, Brazil, Singapore, Saudi Arabia ("surprising that Saudi's there"), Vietnam, Philippines, South Africa, Thailand, UAE.
- **Carve-outs:** agricultural food products, fertilizers, oil/natural gas and certain goods not produced domestically are exempt; goods already under national-security tariffs (steel, aluminum, autos) are excluded from duplicative levies. For **China the duties are additive** on top of existing tariffs, with no carve-outs.

## Utz Brands Acquisition - 34:09
- **UTZ soared ~90%** after agreeing to be taken private by **Intersnack at $14.25/share**; shares closed ~$7.25 Monday and traded as high as $14.04 Tuesday. Close expected **Q4**.
- **"I think anything below 13 here is kind of buyable. There's not a lot of antitrust risk here."**

## Iran War Update and Market Impact - 34:37
- Iran's IRGC claimed it destroyed **Amazon's central data-center facility in Bahrain** with cruise missiles and warned it would strike more named tech facilities.
- As of Sunday night, **Iran halted retaliatory strikes after Trump called off US strikes for a second consecutive night** — "Bessent saw the US 10-year, then freaked out, and that's why we have this rally overnight."
- **Price increases since the war started:** sulfur +142%, European natural gas +97%, heating oil +62%, jet fuel +44%, diesel +38%, gasoline +38%, rice +36%, WTI +30%, Brent +29%, cotton +23%, wheat +18%, palm oil +14%. "So this is a big problem."
- July 22: the White House said for every ship Iran shoots, the US destroys a bridge or power plant; the IRGC hit three tankers attempting the mined southern route (one severely exploded, two turned around) and US bombing resumed 12 hours later. House Republicans passed a **$95B budget plan** to fund the war and restock ammunition. "Trump says Russia and China told him they would not sell weapons to Iran. Don't believe that at all."
- **75% chance of an effective ceasefire by August 31** ahead of the midterms — "which they need."

## General Market Commentary and Cash on Sidelines - 36:33
- BCA Research: **cash on the sidelines is at a record low.**
- Goldman's new oil-upside-risk piece skipped — "if there's no peace, this will be more real… given that peace talks are back on the table, I'm going to skip the oil fearmongering here, even though all the data is true about shale."

## Open Source AI's Impact on Memory Demand - 37:12
- **Bank of America refutes** the "open-source AI is bad for memory" thesis. Closed models **amortize demand across shared HBM pools** concentrated in a handful of data centers; **open models create a new memory footprint per deployment** — 10,000 self-hosting enterprises replicate the weights across **10,000 separate HBM pools**, each with its own KV cache.
- As **1M-token context windows** become commonplace, the **KV cache alone can exceed 40 GB per session**. So low-cost Chinese APIs drive **more** inference demand, enterprise self-hosting and DRAM sockets worldwide.
- **AMD and Anthropic signed a deal for tens of billions of dollars of AI servers over the weekend**; Nvidia signed (or is about to sign) the $250B OpenAI backstop — "which was very shocking because they were creating their own open-source AI, but I guess they need someone to buy their GPUs."

## JPMorgan CIO on AI's Impact - 38:33
- The core of every JPM business unit (fraud, payments, lending) still **relies on critical software platforms like ServiceNow, "which crushed earnings last week."**
- Point solutions are easier to replace with AI now that software is easy to build — "the challenge is to scale and secure it and to make sure people use it." **Open source is extremely risky** because attackers can test public code themselves. JPM's ~**1 exabyte of data** is the competitive advantage it's starting to test.

## Micron/Meta White Paper and DRAM Shortage - 39:38
- A **Micron/Meta white paper** shows a **38× slowdown once shuffle data exceeds available DRAM and spills to SSD** — evidence the DRAM shortage is still binding. "Elon Musk just stopped his earnings call to thank Micron for reasonable prices… he knows he's burning cash now, so he has to play nice."
- **Reddit (RDDT)** fell July 22 on rumors it might walk away from the **$60M/yr Google/Gemini data deal** — "I think it's just a negotiation tactic. Reddit probably wants something closer to **$150–200M per year** because it's the sixth-largest website on the planet," and may believe Gemini benefits more from its data than it gets back as a cited source (hurting user and monetization growth).
- **Q2 revenue tracking +55–60% vs 46% consensus; Q3 tracking >70% YoY** — "we'll probably see some revenue acceleration" from user growth, ARPU, bigger ad budgets and better ad products. With **$2.8B cash and zero debt**, RDDT trades ~**17× NTM earnings with EPS expected +42% over two years — "those estimates are probably too low."**

## STMicroelectronics Earnings Miss - 41:17
- **STM −17%**, the biggest intraday decline since July 2025, on a miss and weak guidance — one of the reasons for last week's selloff. **"Luckily, we didn't have any shares here."**

## Hyperscaler Free Cash Flow Incineration - 41:42
- "The sheer incineration of free cash flow generation at the hyperscalers has been one of the most important market stories of the past year" — an **epic wealth transfer from the capex spenders to the semiconductor companies receiving the spend**, visible in the performance divergence.
- Total hyperscaler capex is expected to **exceed $1 trillion by 2028**; Goldman, Morgan Stanley and JPM have all flagged that capex forecasts were too low last year. Consensus expects **aggregate hyperscaler free cash flow to go negative in 2027**, funded from balance-sheet cash or incremental debt/equity — "but what if the cost of that debt and equity goes up, as we saw last week? Spreads are widening and stock prices are falling. The price to pay for glory would become an increasing challenge."
- The street is **sanguine about FCF improving in 2028**, assuming AI revenue grows faster than the cost to serve. Singh's view: the **capital-light, low-marginal-cost legacy model is gone** — higher spending to drive growth means **lower margins and lower ROIC** (Google guided to exactly that on cloud). "We have been arguing for the past year that AI would gradually make these hyperscalers lower-ROIC companies."
- Market math: the **five hyperscalers are ~16% of the S&P 500** and the **semiconductor cohort ~19%**. A capex cut would rally stocks on renewed discipline — "but it doesn't seem like that's going to happen anytime soon."
- **Max7 multiples compressed from 32× (October 2025) to 22×** even as earnings soared on one-time gains and capex-spreading accounting. The Bloomberg Max7 index fell **4.8% last week** (~$750B of market cap, the most in several years). Deck p37: hyperscaler CDS + the gilt yield; p38–39: **Tesla −14%** on capex and its negative free cash flow.
- **Roper (ROP)** (deck p40) — **+10%, the most since 2020**: adjusted EPS **$5.38 vs $5.29**, revenue **$2.11B vs $2.10B**, full-year EPS guidance raised to **$22.15–22.30**. TD Cowen: "the lack of an issue is a positive in itself right now"; Citi expected a positive reaction to the beat-and-raise. **"We own some shares."**
- **AppLovin (APP)** — "we're adding to our growth position… there are so many new e-commerce platforms, it's not really AI-disruptible. I think AppLovin has about **50% upside**. You can see the model on **page 43**."

## ServiceNow Earnings and South Korea ETF Rules - 48:51
- **Nebius (NBIS)** deployed its **first full version of the Vera Rubin Ultra** — a finished facility — "so maybe we see a bounce there."
- **AMD's Lisa Su expects the AI accelerator market to reach $1.4 trillion by 2030, "which is absolutely absurd."**
- Deck p46: **ServiceNow (NOW)** earnings were "absolutely phenomenal," but people faded the rally — "very unclear why, probably they weren't bullish enough… The market produces an endless stream of information but very little insight. Most of what crosses your screen every day will have almost no bearing on what a business earns five to 10 years from now." **"ServiceNow benefits as the AI organization engine and we think that it will actually continue to grow."**
- Deck p47: **South Korea is tightening rules for leveraged single-stock ETFs.**

## CXMT Memory Pricing and Intel's Turnaround - 50:05
- **CXMT's** memory prices are rising faster than Samsung's — **"there is no cheap memory."** A 64GB DDR5 server module runs ~**$1,240**, so CXMT is not a cheap alternative: "there's really no alternative to SK Hynix, Samsung and Micron for the short term." Chinese memory/chip names sold off last week and are presumably rallying now.
- **Intel (INTC)** — "the long-awaited turnaround is in progress." **Q3 revenue guidance $15.8–16.8B vs $15.1B expected**; **Lip-Bu Tan** is positioning Intel as a key beneficiary of AI data-center spend with **CPUs taking off — data-center segment sales up over 59%**. Shares +3% pre-market then sold off. **"Intel's probably a buy on the dips here… but it's run so much that I don't want to be aggressive in buying it."**

## Token Costs, Nvidia/OpenAI Deal, and Open Source AI - 51:32
- **Token costs have erased the entire agentic-AI bounce** (hyperscaler chart + the Silicon Data LLM token-expenditure index).
- **Nvidia is in talks with OpenAI to guarantee $250 billion of financing for a data center** — "one of the biggest guarantees I've ever seen." A **10-gigawatt project SoftBank is developing in Southern Ohio**.
- **Jensen Huang's first post on open-source AI models** — his **open-weights / American AI leadership** essay (ends on deck **page 54**): the argument is that this is very different from the tech bubble and **demand goes up** because of open-weight models. **"It's a really good read, you should read it."**

## Macro Outlooks and Productivity Differences - 52:25
- Attached: **State Street's** Q3 outlook (US / European / Asian growth), **Citi** on productivity differences across Europe, Japan and the UK — "US productivity growth by region is so high"; global consensus growth forecasts driven by AI investment (deck p58); and **BlackRock on rates — "it's time to buy bonds. They would have been right today"** (deck p59). (The BlackRock, Morgan Stanley and State Street Q3 outlook PDFs are saved in this folder.)

## Q&A: Is Google and Amazon a Sell or Buy on Dips? - 53:06
- **"I believe the semiconductors are buy on dips right now."**
- **The hyperscalers — wait.** "When Meta reports, Microsoft reports, their capex will likely have gone up like Google, and that could be a **sell-the-news event**. So wait for sell-the-news first from Amazon, Microsoft this week, like Google, which sold off last week on higher capex spend."

## Q&A: Adding to Warner Brothers? - 55:52
- **WBD** — "Yeah, with the **15% spread**, especially with the news this week, we'll be adding to it."

## Q&A: Which Agency REITs Do You Like? - 56:37
- **The EchoStar/SpaceX pair unwind:** "We were long Echo, short SpaceX. SpaceX has actually fallen more than Echo, so that spread has come in. Not horrible — but at this point, if you want to just take the entire trade off, **take your short-covering profits on SpaceX and sell Echo**, you can just get out of the whole position."
- **Agency REITs:** the most recent add was **RWTS — "the Redwood S is about a 10% yield."** On the commons: **"we like Dynex (DX) here — a very, very high yield"**, and **"we've been adding to Redwood common (RWT) below five — it was a great add from a three handle."** Also **"AGNC is interesting."**
- **"If you believe rates have peaked, all of them are pretty much adds here."** **NLY after the dividend raise is also an add.** "I personally think the 10-year, especially if we get some sort of temporary peace for a couple of months ahead of the midterms — I think all of them are likely adds here."

## Q&A: Any Thoughts on the Clarity Act? - 57:14
- Delayed by scrutiny over the administration publicly disclosing how much money it made off crypto. **BlackRock, Charles Schwab, Fidelity, Goldman and Grayscale are all pushing passage**; Senator Lummis says it would close gaps that groups like North Korea's Lazarus exploit.
- "There needs to be some resolution around government corruption around crypto before it can pass." **Passage probability ~38% on Polymarket** — "and that's why crypto is kind of where it is. But it probably means crypto is kind of interesting here, especially if rates have peaked, because what's being priced in is actually quite low."

## Q&A: Annaly's Dividend Raise - 58:49
- **NLY raised its quarterly dividend to $0.75** — it had been $0.70 for about six quarters and $0.65 for eight before that. **"Quite positive… about a 13.4% dividend yield here."**

## Q&A: Is Echo a Sell? - 01:01:04
- Repeats the pair unwind: long Echo / short SpaceX, the spread has come in, "take your short-covering profits on SpaceX and sell Echo — you can just get out of the whole position."

## Q&A: View on IBM? - 01:03:30
- **IBM** sold off dramatically "but I think it could decelerate further because of more spending on AI. **Next quarter could miss as well. I would wait for another quarter or two to add to IBM.**"
- On quantum: "not a big contributor to earnings, so I'm not a big fan of quantum overall… you'll continue to see spend focused on AI products and less on IBM. **I just don't see a reason to really own it. It's also not super cheap.**" Expects "three or four quarters of just subdued spend and then a bounce later on, just like you've seen with some of the lower-quality software companies."
- The alternative: "find some of the cheapest AI names that have sold off and do a **short-term trade**, because AI capex is only accelerating — there's probably one more run there — and then focus within tech on some of the cheaper software companies."
- **"Roper, for example, was one of the cheaper industrial software companies, and then they crushed — they're up 10%, but nobody was looking at that name. So being contrarian in this type of market is definitely a really important thing."**
- **The contrarian trade right now:** "wait for the Max7 earnings, because everyone else is increasing their capex spend. The market won't like that. But when the market realizes that all the Max7 guys are raising capex spending — which it seems like is the case — you'll see another bid for the semiconductor names."

## Q&A: Rates vs Yields? / RSP vs SPY - 01:05:55
- "By rates I mean **yields** — yields matter more than Fed funds. The 10-year is down to **4.63%. I think the 10-year might have peaked at 4.7 if this war is over.**"
- Closing question (Gerardo): **RSP vs SPY** — **"RSP should outperform SPY simply because Max7 is underperforming right now."**