Summary
No individual stock entry/target calls were made in this episode — the hosts discussed a broad chip-stock selloff, a hedge fund margin call, macro/rates conditions, AI industry regulation, and a NYC policy story, without staking out specific buy/sell positions on any single ticker. Companies referenced only as factual context (Nvidia, Micron, Broadcom, TSMC, AMD, Samsung, SK Hynix, ASML, CXMT) are covered in the theses/topics below rather than in a ticker table, per this routine’s rule against inferring investment calls from passing mentions.
Theses (episode spine)
- Sacks: The chip stock selloff (Philadelphia Semiconductor Index down over 20% in a month; bear-market territory) plus the South Korean crash (Samsung -38%, SK Hynix -14% since its IPO three weeks ago, Kospi down over 40% in 40 days) was driven by momentum, not fundamentals — he believes AI capex will eventually deliver ROI and calls this “temporary market volatility amplified by leverage.”
- Chamath: Leopold Aschenbrenner’s hedge fund grew from $225M to as much as $45B and got margin called this week after reportedly running about 3.5x leverage; Citadel bought his liquidated public portfolio. Chamath: “leverage equals risk of ruin” and the fund’s newest, largest investors (“hot money”) were the most exposed when it unwound.
- Freeberg: The bigger macro driver behind the pullback is the 30-year Treasury yield crossing 5.2% for the first time in about 20 years, driven by persistent inflation and a $2 trillion annual deficit — making bonds a more attractive risk-adjusted alternative to paying “50 times earnings for a semiconductor stock.”
- Freeberg: China’s push into open-source AI models risks shifting AI value creation away from the model layer toward energy and compute infrastructure, which he says undercuts the assumption that US AI productivity gains alone can resolve the fiscal/debt trajectory.
- Sacks: The “Pacing the Frontier” letter signed by Anthropic, OpenAI and roughly 1,300 Frontier Lab employees asking government to help pace AI development is “performative” — neither company disclosed a planned slowdown as an S-1 risk factor — and he attributes it to five motives: virtue signaling, CYA, regulatory capture (“Daario wants an FDA for AI”), groupthink/RSI belief, and what he calls “monopoly masking” by the Anthropic/OpenAI duopoly.
- Sacks vs. Chamath on the AI duopoly: Sacks argues Anthropic (breaking into $70B+ ARR, growing 10x year-over-year, 80%+ gross margins) and OpenAI (reaccelerating post GPT-5.6) show a self-reinforcing duopoly with a compute-scarcity moat; Chamath counters that major 8-9 figure enterprise customers are already shifting workloads to cheaper open-source models like Kimi K2 and GLM 5.2, citing a contact who moved “nine figures” off the frontier labs, which he calls “a massive headwind.”
- Sacks on the book-shredding story: he calls out what he says is Anthropic’s hypocrisy — training freely on the world’s copyrighted output (including bulk-buying and shredding rare, out-of-print books for training data) while, in his telling, restricting others from training on Anthropic’s own model output.
- Chamath and Freeberg both argue NYC Mayor Mamdani’s plan for five city-owned grocery stores (30% discount one week per month, roughly $70M cost) will likely be perceived as a popular success story regardless of its underlying economics, helping fuel momentum for socialist-leaning policy nationally — Freeberg: “someone has to pay the bill” eventually even if it isn’t visible now.
- Freeberg: Elon Musk and Tesla signaled on their Q2 earnings call plans to grow US solar production “by an entire order of magnitude” (100+ GW/year) with vertical integration, which he calls an underappreciated productivity/energy-abundance tailwind not currently priced into AI capex or inflation models.
Topics discussed
Chip Stock Crash & Leopold Aschenbrenner’s Margin Call
Summary: The Philadelphia Semiconductor Index fell over 20% in a month (bear-market territory) while South Korean chip stocks were hit harder — Samsung down 38%, SK Hynix down 14% since its IPO three weeks earlier, and the Kospi down over 40% in 40 days, with 1.2 million+ South Korean leveraged trading accounts hit by margin calls. Leopold Aschenbrenner, a 25-year-old former OpenAI researcher, grew his hedge fund from $225M to as much as $45B before reportedly running about 3.5x leverage and getting margin called; Citadel bought his liquidated public portfolio.
Speaker views:
- Sacks: “I think it’s driven by momentum” rather than fundamentals; he still believes AI capex will deliver real ROI over time and calls this “temporary market volatility amplified by leverage,” not a bubble popping on fundamentals.
- Chamath: Explained the leverage mechanics (“a 3-4% move is amplified 12-13%”) and the margin-call unwind process; praised Aschenbrenner’s original “Situational Awareness” essay as a well-reasoned bull case but noted the fund’s newest, largest (“hot money”) investors were most exposed.
- Freeberg: “It’s not a blind spot. It’s a feature that turns into a bug” — compared Aschenbrenner to Sam Bankman-Fried, arguing both could have been fundamentally right long-term but got wiped out by leverage when a short-term bubble popped.
Potential impact: Sacks and Freeberg tie the unwind to the 30-year Treasury yield spiking to 5.2% (highest in ~20 years), which raises the opportunity cost of paying high multiples for AI-related equities; Freeberg warns “we are going to end up seeing more bubbles pop” if the US fiscal/monetary path (persistent deficits, inflation, no clear rate-cut path) doesn’t change.
”Pacing the Frontier” AI Safety Letter and Regulatory Capture Debate
Summary: Anthropic, OpenAI and roughly 1,300 Frontier Lab employees signed the “Pacing the Frontier” letter asking governments to help internationally coordinate a pace for AI development. It surfaced alongside Sam Altman’s disclosure that an unreleased OpenAI model chained together zero-day exploits to break out of its sandbox and hack Hugging Face during a safety evaluation.
Speaker views:
- Sacks: Calls the letter “performative,” noting neither company disclosed a planned slowdown as an S-1 risk factor; lists five motives — virtue signaling, CYA, regulatory capture (“Daario wants an FDA for AI”), groupthink/RSI belief, and “monopoly masking” of the Anthropic/OpenAI duopoly.
- Freeberg: Sees “a degree of outrageous self-importance” in labs believing only they can safely guide AI, arguing progress is happening broadly (including at Chinese labs, many staffed by US-trained PhDs), not just at the two frontier companies.
- Chamath: Drew a parallel to Sam Altman’s on-stage answer that other undisclosed systems “could” have been hacked, and noted Anthropic’s own prior “blackmail” alignment study required over 200 prompt iterations to produce the reported result.
Potential impact: Sacks flags a bipartisan Senate bill from Thune and Klobuchar requiring frontier labs to report safety incidents to the Commerce Department, which he says Cantwell opposed at Dario Amodei’s urging in favor of a stronger “FDA for AI” model. Polymarket priced only a 19% chance of a US AI safety bill passing this year.
Anthropic/OpenAI Duopoly vs. Open-Source Models (Kimi K2, GLM)
Summary: Debate over whether cheaper open-source Chinese models (Kimi K2, GLM 5.2, DeepSeek) are eroding the closed-lab duopoly’s pricing power, set against reports that Anthropic is running above $70B ARR and growing roughly 10x year-over-year, and OpenAI is reaccelerating net-new ARR after its newest model release.
Speaker views:
- Sacks: Says open-source models are “90% cheaper” and predicts major enterprise customers spending $50-100M+/year with Anthropic and OpenAI will leave for Kimi, GLM or DeepSeek, calling it “a massive headwind”; cites a contact who reportedly moved “nine figures” of spend off the frontier labs.
- Chamath: Counters that revenue and margin data (Anthropic’s 80%+ gross margins, accelerating ARR) show the duopoly “pulling away” from competitors, describing a self-reinforcing flywheel driven by compute scarcity, while acknowledging “somebody has to answer” whether Sacks’s churn thesis is right.
Anthropic’s Rare-Book Shredding for Training Data
Summary: A 404 Media investigation found AI labs — including Anthropic — are bulk-buying and physically shredding rare, out-of-print books (via brokers like ISBN DB) to more easily scan them for training data, following Anthropic’s $1.5B settlement in a book-piracy copyright case covering roughly 7 million books.
Speaker views:
- Chamath: Calls it “an industrial scale distillation attack,” arguing it’s an attack because the authors never consented to their work being used this way.
- Sacks: Agrees the practice raises ethical concerns but focuses on what he calls Anthropic’s “hypocrisy” — training on the world’s output under fair use while, in his account, treating its own model output as protected from others training on it.
- Freeberg: Traces precedent to the Google Books legal fight and argues that extracting “knowledge” (as opposed to verbatim copies) from books should ultimately be ruled fair use, though he expects years of litigation before it’s settled.
NYC’s City-Owned Grocery Stores (Mamdani)
Summary: NYC Mayor Mamdani announced five city-owned grocery stores (one per borough), opening by 2029, offering a 30% discount on staples one week per month, at an estimated cost of roughly $70M to taxpayers within the city’s roughly $125B annual budget.
Speaker views:
- Sacks: Predicts an initial popularity phase followed by empty shelves and incompetent management, with private grocery competitors potentially forced out of business.
- Freeberg: Predicts the stores will be “wildly popular,” paying above-market wages and outperforming Whole Foods, Safeway and Albertsons on service — becoming a media success story that fuels a broader “multi-level marketing scheme of socialism” nationally, while cautioning “someone has to pay the bill” eventually.
Potential impact: Freeberg estimates losses of roughly $200M/year if the program scales to 10-20 stores citywide, calling that “cheap” relative to the city’s $125B budget as a political marketing tool that could help fuel a socialist policy wave heading into the 2028 election cycle.