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

Chamath frames the bipartisan Epstein Files Transparency Act as a symbolic ‘compact’ between citizens and government: when an issue animates millions, releasing long‑sealed files—carefully redacted to protect victims and open investigations—signals responsiveness, akin to declassifying JFK, MLK, Amelia Earhart, or UFO files. The hosts argue the lack of major Trump revelations suggests the files may be more damaging to Democratic elites and intelligence institutions than to Republicans, which could explain the prior reluctance to leak during the Biden years.

Jason describes meeting Epstein multiple times at TED ‘billionaires’ dinners’, noting how Epstein embedded himself with top scientists, tech founders, and billionaires through donations and tax advice. The group questions how he amassed so much money (e.g., $168M from Leon Black for ‘tax advice’) and why he obsessively networked with scientists and power players. JCal predicts the files will eventually show some intelligence agency involvement—possibly multiple (CIA, Israeli, Russian contacts already appear in released emails)—and that this is a key reason the full story remains suppressed.

Chamath explains Tether’s core model: users in unstable currencies convert local cash into dollar‑pegged USDT; Tether invests the backing dollars in U.S. Treasuries and keeps the yield. With about $183B in circulating USDT, roughly $135B in Treasuries, and additional holdings in Bitcoin, gold, and land, Tether may be throwing off billions in largely profit‑margin interest income with a very small staff. The hosts highlight that 500M+ users—largely in Africa, Central America, and Asia—prioritize currency stability over earning 4% yield, and that this stablecoin adoption reinforces U.S. dollar hegemony abroad.

Current U.S. policy, shaped heavily by bank lobbying, prevents stablecoin issuers from directly paying interest on user balances—protecting traditional banks’ net interest margin. Crypto firms are experimenting with ‘rewards’ as a workaround, while a ‘Clarity Bill’ in Congress seeks to formalize market structure. The hosts see David Sacks’s policy work as pivotal in shifting U.S. crypto from hostile, offshore-driven structures (like early Tether) to regulated, onshore models, and predict intense future competition from Stripe, Visa, Circle, and others, which should compress Tether-like margins over time.

Friedberg walks through GAAP Accounting Standard 360 to rebut Michael Burry’s claim that big tech is ‘cooking the books’ by using 5–6 year useful lives for GPUs. Under GAAP, depreciation must match actual useful life, not technological obsolescence, and only accelerates when assets are truly retired or rendered incapable of revenue-generating use. Since older chips are still producing revenue years later, longer depreciation schedules are justified. The cash flow statement already reveals the capex, so nothing is hidden—investors can choose to value firms on free cash flow instead of EBITDA.

The hosts see Google’s Gemini 3 trained on TPUs as proof that hyperscalers can build viable non‑Nvidia alternatives for training and inference. Chamath expects a fragmented ‘decode’ layer: Grok’s custom silicon, Google TPU, Microsoft and Amazon chips, Meta’s efforts, and specialized low‑power edge inference chips for robotics and vision. Friedberg projects a ‘black swan’ around 2026 from Huawei and Chinese fabs deploying advanced, AI‑designed chips at high volume and low cost, potentially challenging Nvidia in key workloads, especially in non‑U.S. markets.

Jason argues that early fears of Google search being ‘slaughtered’ by ChatGPT were overblown; Google’s search volume and ad revenue are growing, and Gemini now leads many benchmarks. Enterprise adoption is tilting toward Anthropic and open-source models because startups fear OpenAI will compete at the application layer (e.g., Sora vs. image/video startups, coding tools vs. Cursor). His ‘pair trade’ concept: be long Google, Grok, and Anthropic, and structurally short OpenAI’s perceived valuation and dominance as the market fragments and distribution moats (OS, browser, phone) favor incumbents.