Epstein Files Fallout, Nvidia Risks, Burry's Bad Bet, Google's Breakthrough, Tether's Boom

Discussion of the near-unanimous congressional vote to release Epstein files, Larry Summers' involvement, and Jason's theory that Epstein was an intelligence asset. The crew debates why files weren't released during Biden administration and speculates on connections to CIA, Mossad, or Russian intelligence based on Epstein's pattern of targeting scientists and powerful individuals.

• •House voted 427-1 and Senate unanimously to release Epstein files; Trump signed bill saying 'give them everything'
• •Larry Summers stepped down from OpenAI after emails showed he sought dating advice from Epstein until 2019
• •Jason's theory: Epstein was likely an intelligence asset given his focus on top scientists, cameras everywhere, and unexplained wealth sources
• •Leon Black paid Epstein $168M in one year for 'tax advice' - far exceeding normal advisory fees
• •The unanswered question: Where did Epstein's money actually come from, and why was he obsessed with networking scientists and billionaires?

Chamath shares insights from dinner with Tether CEO Paolo Ardoino, revealing Tether's massive scale: 500M users, 30M new users per quarter, $183B in treasuries generating $7-10B annual profit at 95%+ margins. Discussion covers stablecoin mechanics, the fight over interest-sharing with users, and Tether's evolution from regulatory challenges to legitimacy.

• •Tether has 500M users globally, growing 30M users per quarter, with $183B in US treasuries
• •Business generates $7-10B annually at 95%+ profit margins - 'never seen a business like that'
• •Stablecoin users trade local currency for dollar-pegged tokens, Tether earns interest on treasury holdings
• •US legislation prevents stablecoin issuers from sharing interest with users (banks lobbied for this)
• •Tether cleaned up operations, moving from attestations to audits, desperately wants US regulatory approval

Deep dive into Michael Burry's claim that big tech is 'cooking the books' by depreciating AI chips over 6 years instead of 3. Friedberg breaks down GAAP accounting standards, explaining why Burry is wrong - companies can depreciate assets over their useful life as long as they generate revenue. The debate reveals Burry's misunderstanding of how depreciation works versus actual value creation.

• •Burry claims tech companies should depreciate AI chips over 3 years, not 6, which would reduce Google's profit by 10-12%
• •GAAP standards allow depreciation over useful life - if chips still generate revenue in year 6, 6-year depreciation is correct
• •All financial information is visible in cash flow statements - companies aren't hiding anything, investors can choose valuation metrics
• •Chamath argues output token revenue matters more than energy costs - companies already gate usage based on profitability
• •Friedberg: 'This conversation lacks technical literacy' - Burry confuses physical utilization with value creation

Analysis of Google's Gemini 3 release, which regained the lead on most LLM benchmarks and was reportedly trained entirely on Google's TPUs rather than NVIDIA GPUs. Discussion covers the fragmentation of AI chip market, with every major player (Microsoft, Amazon, Meta) developing custom silicon, and predictions for Huawei's potential disruption in 2026-2027.

• •Google's market share jumped from 8% to 16% in chat market; Polymarket gives them 89% odds to finish year as top LLM
• •Gemini 3 reportedly trained only on Google TPUs, not NVIDIA GPUs - major validation of custom chip strategy
• •Market fragmenting: Grok, TPU, Microsoft, Amazon Inferentia, Meta all developing specialized decode chips
• •Friedberg's 2026 prediction: Huawei will announce breakthrough lithography enabling cheap, high-volume chip production in China
• •Jason's contrarian take: Google's search franchise will grow with AI, not shrink - better targeting and more searches offset lower per-search revenue

Candid discussion on the psychology of investing with your own capital versus managing LP money. Chamath reveals higher returns but wider dispersion investing solo, while Friedberg shares his failed venture studio experience and decision to return as CEO of Ohalo after watching Oppenheimer. Both emphasize the accountability and discipline that comes from managing others' capital.

• •Chamath: Returns better investing own money, but dispersion increased - holds losers longer, cuts winners less aggressively
• •Managing LP money created discipline: 'Never lose money, return capital quickly, run the upside' - traded 7x volatile for guaranteed 3.5x
• •Friedberg's venture studio delusion: MetroMile and Eatsa both failed, realized being chairman doesn't work - 'frustrating watching CEOs ignore advice'
• •Watching Oppenheimer in IMAX made Friedberg cry and realize: 'I wasn't doing what I should be doing with my life'
• •Friedberg went all-in as Ohalo CEO after $40M research investment showed breakthrough results - LPs were thrilled

High-stakes poker legend Alan Keating joins to discuss his famous bluff against Doug Polk with 4-2, the psychology of reading opponents, and how mastering fear creates edge. Keating explains his investment philosophy of making bets 'where if it doesn't work out, I'm in trouble' and the beauty of thriving in chaos beyond where everyone else is prepared.

• •Keating's edge: exceptional live reads and comfort with fear - 'people make bad decisions when they're scared'
• •Famous Doug Polk hand: called $600K all-in with 4-2 based on tonality tells and bet sizing patterns from earlier hands
• •Investment strategy mirrors poker: makes bets where failure means 'a little bit of trouble' - pain is a motivator
• •Calls Chamath for advice, then triples down anyway - uses conversations to 'earmark all the reasoning' for later reflection (super forecasting)
• •Philosophy: 'There's beauty in the chaos after everyone's prepared' - no interest in spaces where everyone has a plan