All-In PodcastWhy Burry's AI short misreads data center depreciation
Burry claims hyperscalers cook the books on data center depreciation; the cash flows are apparent in GAAP filings, and Palantir has no clear alternative.
CHAPTERS
- 0:00 – 4:20
Media’s Math Error and Michael Burry’s AI Shorts
The hosts open by mocking CNBC’s misreporting of Burry’s Palantir short size and use it to critique mainstream financial illiteracy. They then examine Burry’s broader thesis that AI hyperscalers are using aggressive depreciation to inflate earnings.
- •CNBC reportedly mis-stated Burry’s Palantir short as $900M instead of $9M due to misunderstanding options contracts.
- •Hosts argue many financial journalists lack basic experience with assets and derivatives, leading to uncorrected fact errors.
- •Chamath floats a tongue-in-cheek conspiracy that the misreporter could be intentionally trading around their own mistaken headline.
- •Misreporting amplified fear in an already frothy, then-correcting, AI market.
- 4:20 – 14:00
Accounting Corner: Data Center Depreciation and AI Infrastructure
Friedberg walks through GAAP depreciation mechanics and assesses Burry’s accusation that hyperscalers are ‘cooking the books’ by extending useful lives of servers and networking equipment. He and Chamath argue that real-world utilization patterns and technical advances justify longer depreciation schedules.
- •Under GAAP, CapEx is depreciated over an asset’s useful life; changing from 3 to 6 years roughly halves annual depreciation expense.
- •Example: If Google spends $70B in CapEx, moving from 3-year to 6-year depreciation shifts ~10% of operating profit on paper.
- •Google has progressively extended useful lives (servers from 3→4 years; networking from 3→5→6 years) after internal analysis.
- •7–8-year-old TPUs/GPUs at Google and elsewhere are still at 100% utilization, supporting longer useful lives.
- •Chamath argues Burry lacks the technical depth to understand kernel improvements, attention mechanism advances, and chip design changes that increase longevity and utilization.
- 14:00 – 25:50
The Role and Reality of Short Selling
Chamath reflects on his hedge fund experience to contextualize Burry’s shorts and the broader short-selling ecosystem. He contends that genuine fraud detection is rare and that most shorts are about manufacturing panic rather than protecting markets.
- •Chamath’s hedge fund made substantial gains on long positions (Tesla, Amazon, Salesforce) but little on shorts.
- •He recalls only one clear short success: a fraudulent telecom whose corporate HQ was literally a pizza parlor.
- •Short reports often blend valid criticism with sensationalism aimed at rapid price drops.
- •As public company oversight has tightened, large-scale fraud has become much harder, further weakening the rationale for most shorts.
- 25:50 – 35:00
Palantir’s Extreme Valuation and Competitive Moat
The group dissects Burry’s Palantir short amid the company’s massive price-to-sales multiple. Friedberg emphasizes forward cash flows over trailing revenue, while Chamath argues Palantir’s uniqueness and lack of churn risk underpin its premium valuation.
- •Palantir’s run-rate revenue is ~$3.5B versus a market cap around $480B, implying ~137x sales.
- •Comparable high-growth software names (Datadog, Snowflake, Cloudflare, CrowdStrike) trade at much lower multiples.
- •Friedberg stresses that investors should price in long-term earnings power, not just last year’s sales.
- •Chamath calls the Palantir short “stupid,” claiming there are no true substitutes; most peers like MongoDB or Snowflake are well-run but not unique.
- •He argues valuation premia are rational when cash flow durability is high and churn options are minimal.
- 35:00 – 40:50
White House Dinner and All-In Event Promotion Interlude
The hosts take a lighter detour as Chamath describes a White House dinner with Trump and other financial leaders. They pivot into promoting the All-In holiday party and their tequila line.
- •Chamath attended a Trump-hosted dinner for financial leaders and was invited to the Oval Office bill signing.
- •They joke about Trump’s cologne and White House gift-shop fantasies.
- •All-In promotes its San Francisco holiday party and a casino night featuring Tony Hinchcliffe and a celebrity DJ.
- •Their tequila brand has begun shipping to buyers.
- 40:50 – 47:30
Housing Affordability Crisis and 50-Year Portable Mortgages
The conversation shifts to America’s housing affordability crisis, centering on a proposed 50-year, portable mortgage and the backlash it triggered. Chamath frames affordability—housing, healthcare, student loans—as the linchpin issue for upcoming elections.
- •Trump administration floated 50-year mortgages and portable mortgages to reduce monthly payments and increase mobility.
- •Critics label long-duration mortgages as ‘debt slavery’ due to higher lifetime interest costs.
- •Portable mortgages could unlock housing supply by letting owners keep low-rate loans when moving, reducing the ‘mortgage lock-in’ effect.
- •Average age of first-time homebuyers has risen sharply from 28 (1991) to 40, with most of the jump in the last 4 years.
- •Chamath identifies three priority affordability issues: housing access, Obamacare-driven healthcare cost inflation, and student debt relief.
- 47:30 – 59:10
Government Distortions: Rent Control, Fannie/Freddie, and Housing Supply
Friedberg dissects how local rent caps and federal lending backstops combine to distort housing markets. Jason contrasts supply-squeezed coasts with Texas metros, arguing that aggressive building is the only durable path to affordability.
- •LA City Council capped annual rent increases to 90% of CPI, with a 1% floor and 4% ceiling.
- •Rent control and heavy regulation reduce the upside for investors, discouraging new builds and renovations.
- •Federal agencies (Fannie/Freddie) backstop ~$8T in home loans, adding liquidity that ultimately bids up prices.
- •Government ‘help’ via subsidies and guarantees in housing, education, and healthcare often inflates costs instead of expanding real access.
- •Jason cites Austin/Houston as counterexamples: abundant new units drove Austin rents down ~20% in three years.
- •He describes a young couple on ~$130k income easily renting cheaply and buying a $300–500k home around Austin at $200–300/sq ft.
- 59:10 – 1:08:40
Student Debt, Personal Responsibility, and Putting Universities on the Hook
The hosts connect student loans to broader affordability woes and recount a White House discussion about potential reforms. They advocate both structural changes to university incentives and greater individual financial discipline among students.
- •Bill Ackman’s suggested reform: universities should absorb the first-loss chunk on student loans (e.g., $20–40k), forcing them to underwrite ROI.
- •Such a system would discourage universities from pushing expensive, low-earnings degrees.
- •Jason argues students must exercise agency: work during school, live cheaply off-campus, extend graduation timelines, and choose high-demand fields.
- •He criticizes entitlement attitudes among grads who expect to live in core global cities despite modest earnings and high debts.
- 1:08:40 – 1:18:40
H‑1B Abuse, High-Skill Immigration, and Auction-Based Reform
Trump’s comments on H‑1B visas spark a deeper dive into visa abuse and potential fixes. The hosts try to reconcile the need for top technical talent with the reality that large firms game the lottery-based system.
- •Trump defends the need for imported specialized talent, arguing you can’t simply retrain the unemployed for advanced missile factories overnight.
- •Chamath describes how big firms file hundreds of thousands of H‑1B applications to raise their lottery odds, crowding out smaller firms.
- •He proposes: (1) curbing bulk-application abuse and (2) attaching a high per-visa fee (e.g., $100k) to signal true scarcity and value.
- •Jason supports auctioning a portion of visas so Big Tech pays market rates, using revenues to fund domestic vocational retraining.
- •They criticize inconsistent enforcement, including heavy-handed deportations at a Hyundai plant even as the administration courts foreign manufacturing investment.
- 1:18:40 – 1:31:20
Solar Storms, CMEs, and Technological Vulnerability
Friedberg delivers a detailed explainer on the week’s major coronal mass ejections and the resulting G5 geomagnetic storm. He outlines both why this event was relatively benign and how a larger Carrington-level storm could cripple modern electronic infrastructure.
- •The sun’s 11-year cycle leads to periods of intense activity where magnetic fields snap and eject plasma clouds (CMEs).
- •Three large CMEs, two overlapping, hit Earth and triggered a G5 geomagnetic storm—the top of the scale.
- •Charged protons at >100 MeV spiked by ~1000x, prompting airlines to avoid polar routes due to radiation concerns.
- •Storms can disrupt GPS, communications, and induce currents that damage satellites and power grids.
- •A Carrington-scale event today could destroy satellites and transformers, effectively rolling back key systems to a pre-digital state.
- •Friedberg predicts that a long-term shift from electron-based to photon/quantum computing would greatly reduce vulnerability to such storms.
- 1:31:20
Tech Diaspora, Network States, and Global ‘Escape Hatches’
The episode closes with observations from Tokyo and Southeast Asia, where many tech figures are exploring alternatives to the U.S. for residence and innovation. The hosts riff on network-state experiments and tongue-in-cheek visions of ultra-luxury micronations.
- •Jason notes a growing tech expat presence in Tokyo and Riyadh, driven by fears of future U.S. ‘confiscatory’ policies.
- •He visits Malaysia’s Forest City, where Balaji Srinivasan is hosting an in-person ‘network school’/proto-network state.
- •Participants pay a single fee covering housing, food, and amenities to live alongside like-minded builders.
- •Some founders see the U.S. as late-cycle and look abroad for ‘Wild West’ environments to build in.
- •Chamath jokes about founding a vicuña-and-wagyu-only enclave, underscoring the satirical side of the discussion even as the underlying trend is real.
