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

Watch on YouTube ↗  |  November 22, 2025 at 01:28  |  1:01:52  |  All-In Podcast
Speakers
Jason Calacanis — Angel Investor / Founder, LAUNCH
David Friedberg — CEO, The Production Board
Chamath Palihapitiya — CEO, Social Capital

Summary

Recorded live from The Venetian in Las Vegas during F1 weekend, the besties open with the near-unanimous congressional vote to release the Epstein files and speculate about the unexplained source of his money. Chamath then walks through Tether's stablecoin economics after a dinner with founder Paolo Ardoino, which turns into a debate about whether 95 percent margins survive competition and a falling rate cycle. The main market segment is a detailed rebuttal of Michael Burry's claim that hyperscalers inflate earnings by depreciating AI chips too slowly, followed by Gemini 3, Google's TPUs, the coming fragmentation of inference silicon and Huawei as a 2026 risk to Nvidia. The show closes with Friedberg on why he returned to being an operating CEO and a poker conversation with Alan Keating.

  • Congress voted 427 to 1 to release the Epstein files; the besties expect embarrassment across both parties and suspect intelligence-agency involvement.
  • Tether is described as an extraordinary business: about 500 million users, 30 million added per quarter, roughly 183 billion dollars of USDT with 135 billion in Treasuries and margins above 95 percent.
  • The group argues stablecoin economics face competition from Stripe, Visa and the banks, and a direct revenue hit when interest rates fall, with Circle the listed exposure.
  • Nvidia reported revenue up 62 percent year over year, net income of 31.9 billion dollars and guidance of 65 billion dollars for the current quarter.
  • Friedberg rebuts Michael Burry point by point: chips still generate revenue in year six, GAAP allows it, and three-year schedules would cut Google's profit only 10 to 12 percent.
  • Chamath argues decode and inference silicon fragments quickly across TPUs, Groq, Microsoft, Amazon Inferentia and Meta, and names Huawei as the underpriced 2026 risk to Nvidia.
  • Jason takes the other side of the search-disruption trade, going long Google and short OpenAI on rising search volume, better AI ad targeting and eroding startup trust in OpenAI.
  • Friedberg explains concentrating on Ohalo as CEO after failed venture-studio bets, and Alan Keating discusses live reads, bet sizing and mastering fear.
Ideas
Jason Calacanis Angel Investor / Founder, LAUNCH 21:22
Stablecoin margins compress as rates fall.
Stablecoin issuers earn essentially all of their profit from the interest on customer float parked in Treasuries, which produces margins above 90% with barely 100 employees. That structure cannot hold. Stripe has already bought a stablecoin provider, Visa and effectively every large payments company will issue their own coin, and the banks are fighting the same fight over net interest margin, so a business this profitable invites competition until the spread is ground down. The second and more mechanical problem is rates: because the revenue is simply the yield on the reserves, a falling rate cycle takes the revenue down with it. Both forces are headwinds for the whole stablecoin space, and Circle is the listed way that space is owned.
CRCL MED
David Friedberg CEO, The Production Board 27:14
Burry's AI depreciation fraud claim is wrong.
Michael Burry's claim that big tech is inflating earnings by depreciating AI chips over six years instead of three is wrong on the accounting itself. Under GAAP (accounting standards 360) useful life is set by whether the asset is still producing value for the owner, not by whether a newer, better asset exists. Old H100s are still generating revenue in year six, they are not being retired, maintenance costs are not spiking, throughput requirements are not forcing early retirement and the chips are not being held for sale, so none of the conditions that force accelerated write-downs are met. Even in the extreme case, moving Google to a three-year schedule would only reduce its net profit by roughly 10 to 12 percent, which is not the order of distortion a book-cooking accusation implies. And nothing is concealed: capex and cash movements are fully disclosed in the cash flow statement, so any investor who wants to value these companies on free cash flow rather than GAAP earnings can already do so.
NVDA GOOGL HIGH
Chamath Palihapitiya CEO, Social Capital 35:59
TPU strength as decode compute fragments.
The AI model market is not winner-take-all; it is a nascent market now going through a sorting function, with separate winners in science, in enterprise coding and in consumer chat. The evidence is that the traffic allocation that once favored a single company is breaking apart: Google went from roughly 8 percent to 16 percent of the entire chat market in a single month. On the enterprise side the deciding variable is model quality, but on the consumer chat side the advantage pivots on inherent distribution, meaning the operating system, the browser and the phone. That structurally favors Apple, Google and Microsoft in owning the consumer assistant layer.
GOOGL AAPL MSFT MED
Jason Calacanis Angel Investor / Founder, LAUNCH 38:03
Google's search franchise grows despite AI.
Taking the other side of the consensus that AI destroys Google's search business: AI improves advertising targeting and increases the number of searches, so even if revenue per search declines, higher query volume plus better targeting per query grows the franchise rather than shrinking it. The feared slaughter by ChatGPT has not happened, with searches and revenue both rising, Gemini 3 has retaken the benchmark lead and prediction markets put Google at 89 percent to end the year with the top model. Google is firing on all cylinders while the challengers lose share, so the search franchise grows from here.
GOOGL HIGH
Chamath Palihapitiya CEO, Social Capital 40:47
Huawei is the underpriced 2026 Nvidia risk.
The black swan missing from the 2026 equation is Huawei. There is lithography technology inside China that is not publicly discussed and that will be deployed through Huawei and the fabs being built across the mainland, which would let Huawei produce chips at very low cost, in high volume and on reasonably short lead times, the exact opposite of the expensive, long-lead-time parts the market is paying up for today. Chip architecture is also being redesigned with AI, which compresses the catch-up period further. Expect announcements during 2026 and real competitive impact by 2027; that is the underpriced risk sitting against Nvidia.
NVDA MED
Up Next
49:41
Why Living Standards Are About To Collapse, Says Economist | Steve Hanke
The David Lin Report · Nov 21
33:58
Layoffs Surge, Delinquencies Soar; How Bad Will It Get? | Danielle DiMartino Booth
The David Lin Report · Nov 20
26:15
Sentiment Hits 40-Year Low, Brace For Massive Job Losses? | Joanne Hsu
The David Lin Report · Nov 19
39:24
What Will Trigger The Next Great Depression? | John Tamny
The David Lin Report · Nov 19
23:28
Government Shutdown Cost Economy Billions, Congressman Explains Impacts | Rep. Stutzman
The David Lin Report · Nov 18

This All-In Podcast video, published November 22, 2025, features Jason Calacanis, David Friedberg, Chamath Palihapitiya discussing CRCL, NVDA, GOOGL, AAPL, MSFT. 5 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Jason Calacanis, David Friedberg, Chamath Palihapitiya  · Tickers: CRCL, NVDA, GOOGL, AAPL, MSFT