All-In PodcastTrump Brokers Gaza Peace Deal, National Guard in Chicago, OpenAI/AMD, AI Roundtripping, Gold Rally
CHAPTERS
- 0:00 – 7:10
Cold Open, Banter, and Reintroducing Brad
The hosts open with trademark banter, stock price jokes, and a comedic recap of how much has changed since Brad’s last appearance, framing him as the ‘fifth bestie’ returning after a long absence marked by SPAC booms, crypto crises, and political shifts.
- •Lighthearted jokes about Uber and Robinhood stock prices and Chamath’s net worth moods.
- •Brad recaps major events since his last appearance: SPAC run-ups, crypto collapses, JCal’s political evolution and move to Texas.
- •Sets a playful tone before pivoting into serious policy, tech, and markets discussion.
- 7:10 – 14:40
Trump’s Gaza Ceasefire Deal and Middle East ‘Moonshot’ Diplomacy
The panel dissects Trump’s newly announced multi-phase Gaza ceasefire deal, positioning it as a rare diplomatic success built on hard pressure on both Hamas and Netanyahu. They then zoom out to Trump’s broader ‘moonshot presidency’ across Ukraine-Russia, China-Taiwan, and the Abraham Accords.
- •Phase one of the deal: ceasefire, unrestricted aid to Gaza, release of all Israeli hostages, 2,000 Palestinian prisoners freed, and Israeli troop withdrawals.
- •Sacks cites Aaron David Miller on Trump showing unprecedented willingness to pressure an Israeli PM, making the US a more credible mediator.
- •Brad brands Trump’s agenda as a series of high-risk ‘moonshots’ (Middle East, Ukraine-Russia, China, re‑industrialization, AI), arguing moonshots are valued in Silicon Valley but politically rare.
- •Speculation about Trump, Kushner, and Wittkop being Nobel Peace Prize contenders if multiple conflicts (Gaza, Ukraine, possibly China/Taiwan) are resolved.
- •Observation that parts of the global left suddenly went quiet on ‘ceasefire now’ rhetoric once Trump appeared to deliver one.
- 14:40 – 32:40
Post-War Middle East: From Oil Dependence to Diversified Powerhouse
Chamath and Brad analyze what a stable post-conflict Middle East could look like, focusing on the economic logic of rapidly monetizing oil and investing in new sectors. They tie regional peace directly to megaprojects in tourism, AI, sports, and new cities.
- •Chamath argues oil is a wasting asset: delayed extraction risks losing value to nuclear, gas, and solar ‘abundance of electrons’.
- •Peace enables states to focus on monetizing reserves and re-deploying capital into services and diversification.
- •Brad notes Gulf states’ ambitions in private equity, renewables, AI, sports ownership, venture capital, and futuristic cities.
- •Stability is essential for attracting foreign capital and tourism; instability makes investors avoid the region.
- •Compliments to Jared Kushner for both mega-LBOs and diplomatic wins that set the region up as a long-term growth pillar.
- 32:40 – 47:20
National Guard in Chicago: Crime, ICE Raids, and Federal vs. State Power
The discussion shifts to Operation Midway Blitz in Chicago, ICE raids, and Trump’s deployment of 300–500 National Guard troops to protect ICE agents from violent protests. This sparks a detailed debate about crime, immigration enforcement, ANTIFA, legal authority, and political framing.
- •Background: 800 illegal immigrants arrested in a month; high-profile raid in South Shore used flash bangs and a Black Hawk.
- •Illinois leadership (Pritzker, Johnson) opposes the raids and Guard deployment, enacts ‘ICE-free zones’, and sues over federal overreach.
- •Sacks links lenient prosecutors and zero-bail policies to repeat violent offenders in multiple cities, arguing federal support is needed where local officials obstruct enforcement.
- •He cites Eisenhower and JFK sending troops to the South as precedent for using federal power to enforce federal law.
- •Chamath praises Maryland’s Democratic Governor Wes Moore for quietly surging state police while publicly criticizing Trump—arguing results (less crime) matter more than who gets credit.
- •Brad gives historical deportation stats (Clinton/Obama at 200k–400k/year) to show both parties have long conducted removals, questioning why it’s now uniquely controversial.
- 47:20 – 1:02:00
Trump 2.0 vs. Trump 1.0: ICE Tactics, Polling, and Moderates’ Support
JCal presents polling data showing Trump’s net approvals sliding on immigration, economy, and inflation, tying the drop to images of violent raids and January 6th–style ‘authoritarian’ vibes. He contrasts the popular Trump 2.0 policy mix with the unpopular Trump 1.0 chaos, arguing the current approach risks midterm disaster.
- •Charts show Trump’s net approval plunging after tariffs, LA protests, Epstein-file decisions, and most recently Chicago ICE raids.
- •Trump’s strengths with moderates: strong border but legal immigration, lower taxes (e.g., no tax on tips), pro-innovation, anti-war, deficit reduction, and deregulation (‘DOGE’).
- •Turnoffs: January 6th violence, images of ICE in masks tackling immigrants, kids in cages, Muslim ban, pardons for January 6th rioters, efforts to overturn election results.
- •JCal argues Americans love the Constitution and rule of law but hate chaos and cruelty; he urges advisers to keep Trump on the 2.0 track.
- •He objects to ICE raids as brutal, inefficient (~$100k/month per deportee), and prone to wrongful targeting and racial profiling; proposes employer-focused fines and compassionate paths for long‑term, non-criminal undocumented residents.
- •Sacks counters that violence comes from ANTIFA and leftist mobs, not law enforcement, and that deportations are a necessary correction after Biden’s ‘open border’ years.
- 1:02:00 – 1:30:00
How Should Immigration Enforcement Work? Alternative Tactics and Legal Lines
The group drills into what ‘non-brutal’ enforcement could look like. JCal proposes employer-centric enforcement and self-deportation incentives, while Sacks insists mass amnesty is untenable after millions entered illegally. They spar over ANTIFA, Proud Boys, and whether federal interventions are being mischaracterized.
- •JCal’s preferred model: target employers with escalating fines ($10k, then $20k per illegal worker), encourage self-deportation, and reserve forceful tactics for actual criminals and gangs.
- •He emphasizes compassion and economic realism: many undocumented workers were effectively invited by prior bipartisan support for cheap labor.
- •Sacks argues Democrats created the problem by opening borders then denying a crisis; Republicans must be allowed to correct it when in power.
- •He notes DC’s successful cleanup under federal authority as proof that with cooperation (vs. Chicago/Portland obstruction) there’s less conflict and more safety.
- •JCal agrees radical groups on both sides (ANTIFA, Proud Boys, Oath Keepers) should be cracked down on; he and Sacks disagree over whether January 6th is relevant to the current ICE debate.
- 1:30:00 – 1:43:20
AMD–OpenAI Megadeal: Bet-the-Farm GPU Strategy and the Gigawatt Era
Attention shifts to AI: AMD’s stock spikes on news of a possible $60B+ GPU deal with OpenAI, including up to 10% of AMD in warrants for OpenAI if capacity is actually used. Brad, Chamath, and Sacks unpack AMD’s strategic gamble, NVIDIA’s dominance, and why AI capacity is now framed in gigawatts rather than chip counts.
- •In 2022 AMD and NVIDIA had similar revenues (~$25B); now NVIDIA is on track for >$200B while AMD is just ~$33B, with NVIDIA capturing nearly all incremental AI data-center revenue.
- •Lisa Su’s MI350 lagged badly; MI450 is AMD’s last credible shot to compete. The warrant structure is effectively a ‘bet the farm’ that MI450 will gain traction.
- •Deals are now quoted in gigawatts because power is the binding constraint and chip generations change quickly; a GW normalizes across H100, GB200, TPUs, Trainium, etc.
- •Brad notes Wall Street estimates are far below the ‘100 GW, $4–5T buildout’ hype; official NVIDIA forecasts show a slower growth trajectory, reflecting skepticism.
- •They discuss training vs. inference, with Sacks questioning whether NVIDIA’s lead in training could be diluted if specialized inferencing chips/ASICs win most of the 99% inference market.
- •Consensus that demand is likely understated because new AI applications (agents, video, etc.) will emerge, echoing how early internet fiber was eventually fully used.
- 1:43:20 – 1:53:20
Power, HBM, and Jevons Paradox: Who Really Controls AI Growth?
Chamath reframes AI’s bottlenecks around electricity and constrained inputs like HBM memory, analogizing to Rothschild’s quote about controlling the money supply. The group explores massive data center capex, hardware reliability issues, and the interplay between efficiency gains and exploding token demand.
- •NVIDIA’s architecture bets heavily on HBM; HBM capacity is dominated by SK hynix/Samsung, giving them leverage as key chokepoints.
- •Chamath interprets OpenAI’s Korean memory deals as Sam Altman securing HBM allocation in advance, giving him power to allocate that scarce capacity to others in exchange for equity returns.
- •Rule of thumb: ~$50B of total investment per 1 GW data center (chips, power, land, shell); 10 GW ‘mega-centers’ could cost ~$500B each over time.
- •Chamath notes high failure rates (9–20%) at the server-rack system level as we push physics and nanometer scaling, adding operational complexity.
- •They discuss Jevons paradox: as tokens get cheaper (more efficient chips, sparser models) total token consumption may explode, driving yet more demand for compute and power.
- 1:53:20 – 2:15:00
Is AI a Bubble? TAM, Dark GPUs, and Round-Tripping Concerns
The besties debate whether AI resembles the dot‑com dark‑fiber bubble, while also tackling fears that equity-for-GPU commitments constitute dangerous round-tripping. They largely conclude that AI demand is real, GPUs are fully utilized, and vendor-financing structures are economically substantive so far.
- •Brad points out there are ‘no dark GPUs’—unlike unused dark fiber, all current GPU capacity is over-subscribed.
- •They estimate potential AI TAM: hundreds of millions of business users paying ~$3k/year and consumers paying subscription-style fees, implying multi-trillion-dollar annual markets.
- •Chamath notes that a trillion in AI capex implies many trillions of opex (power, upgrades, operations) over 20 years, so total economic footprint is massive.
- •Brad and Sacks stress that round-tripping only becomes problematic when there’s no real end demand (sham deals); current AI deals ride on companies like OpenAI growing revenue from ~$5B to potentially tens of billions.
- •Chamath analogizes GPU forward-flow and equity structures to long‑standing auto dealer floor-plan financing and factoring; legal, audited, and industry-standard when real customers exist.
- •They acknowledge some players will blow up if a correction hits, but expect winners’ gains to outweigh losers, as in prior tech booms.
- 2:15:00 – 2:26:40
Gold and Silver’s Surge: Tether, Central Banks, and De-Dollarization
The conversation turns to gold’s record-breaking rally and silver’s outperformance. Rather than a single ‘fear trade’ story, they attribute the move to new forms of structural demand, stablecoins, and geopolitical hedging against US dollar weaponization.
- •Gold breaks above $4,000, up ~50% YTD, outpacing bitcoin and equities; silver rallies even more, partly due to industrial use.
- •Chamath emphasizes net new buyers: Tether’s gold-backed stablecoin, central bank rebalancing, macro funds wary of central banks and fiat.
- •Brad shows charts implying correlations with both US debt and S&P risk assets but warns of ‘chart crime’—overfitting narratives to noisy data.
- •Sacks highlights China’s 11-month streak of adding gold, building ~74M ounces and $283B in reserves as they diversify away from Treasuries.
- •He ties de-dollarization to US sanctions policy and Russian asset freezes, prompting BRICS to explore gold-backed settlement certificates for large trade flows.
- 2:26:40 – 2:35:00
Polymarket and the Institutionalization of Prediction Markets
They close with ICE (Intercontinental Exchange) investing in Polymarket at a multi‑billion valuation, paving the way for US institutional and retail access to political and event betting. The panel sees this as the professionalization of ‘wisdom of the crowds’ and a step toward everything becoming tradable.
- •ICE invests roughly $2B at around a $9B valuation; Polymarket is expected to launch widely in the US, with even a market on its own go‑live date.
- •Chamath relays ICE CEO Jeff Sprecher’s view that ‘everything is becoming a market’—sports, knowledge, equities, and debt all moving toward tokenized, liquid trading.
- •Potential disruption to HFT shops, sportsbooks (FanDuel/DraftKings), and traditional exchanges as on‑chain, token-based markets undercut cost structures.
- •Brad celebrates that prediction markets ‘professionalize wisdom of the crowds,’ citing 1907 experiments where crowd estimates beat experts, and arguing real-money markets are even more accurate.
- •They note earlier failed attempts like Intrade and frame Polymarket’s success as a lesson in persistence, regulation-friendly execution, and timing.
- 2:35:00
Sign-off, Running Gags, and Meta Commentary
The episode wraps with jokes about ‘steel-manning,’ Texas ranch life, law-and-order ‘security blankets,’ and inside references to the All-In universe. They tease future segments and rib each other about their political influence and portfolios.
- •Self-deprecating jokes about who really calls ‘balls and strikes’ and who’s benefiting most from Trump-era policies.
- •Running bits about JCal’s guns and ranch in Texas, NRA-ish vibes vs. Silicon Valley roots.
- •Teasing future science vs. markets segments with Friedberg vs. Brad.
- •Final nod to Sacks as an informal adviser to the Trump administration on free markets and AI.
- •Housekeeping plug for All-In.com and the newsletter before sign-off.
