Breaking Down Michael Burry’s Big Nvidia Short Thesis and Open AI’s Massive Loss Projections | Jack & Max

Episode Summary

Executive Summary: The episode centers on Michael Burry’s critique of the AI boom, especially NVIDIA, arguing that depreciation schedules, circular financing, and overheated sentiment may be masking weak economics. The hosts debate whether Burry’s short thesis is material now or premature, with more concern directed at OpenAI’s huge projected losses than NVIDIA’s profitability. They also touch on AI’s macro impact and the growing likelihood of a December Fed rate cut.

Main Topics: Michael Burry’s AI short thesis (Priority: 5/5): The hosts discuss Burry’s Substack launch and his claim that AI markets may be in a bubble, focusing on his criticism of NVIDIA and the broader AI trade. Depreciation and earnings quality (Priority: 5/5): They examine the argument that AI firms are extending chip depreciation lives, artificially boosting earnings by understating the real economic decay of hardware. Why NVIDIA is the target (Priority: 4/5): A major debate is whether NVIDIA is the right company to short, given that it is the supplier rather than the ultimate spender in the AI capex cycle. OpenAI’s economics and venture-backed losses (Priority: 5/5): The discussion shifts to HSBC projections showing OpenAI’s massive revenue growth but even larger cumulative operating losses, raising questions about sustainability. Circular financing and AI capex sustainability (Priority: 4/5): The hosts worry that AI spending is being financed by unprofitable startups and venture capital liquidity, creating a potentially unstable funding loop. Macro and Fed implications (Priority: 3/5): They end with a brief macro update on rising odds of a December Fed cut and the possibility that AI spending could bolster GDP even in a weak real economy.

Key Arguments: Burry’s depreciation critique may have merit, but it may not matter to near-term stock prices if AI demand remains strong. The strongest bubble risk is not NVIDIA itself but the unprofitable AI companies buying its chips, especially OpenAI. NVIDIA is difficult to justify as the primary short because its customers are highly profitable hyperscalers funding purchases with free cash flow. If AI spending slows or becomes cyclical, NVIDIA’s revenue and valuation multiples could compress sharply. The more serious fragility may be circular financing: venture-backed firms losing huge sums while relying on future capital raises to keep buying compute. HSBC-style projections imply OpenAI can grow revenue massively and still lose vast amounts of money for years, which may be unsustainable. The AI bull case depends on falling inference costs and rapid model improvement, but the hosts question whether that is enough to offset the spending burden. Burry’s edge in 2008 came from novel loan-level data; here, the hosts argue much of his evidence is already public and not clearly new. AI spending could materially inflate GDP even if the broader economy weakens, creating a disconnect between macro data and lived conditions.

Data Points: NVIDIA depreciation life: 2-3 years historically, extended to 5-6 years in some cases - Central to Burry/Chanos argument that AI chips are being depreciated too slowly NVIDIA net income since 2018: $205 billion - Used to argue NVIDIA has generated enormous profits over the period NVIDIA free cash flow since 2018: $188 billion - Supports the point that the company is highly cash generative Share-based compensation since 2018: $20.5 billion - Mentioned in Burry’s critique of NVIDIA’s capital allocation and dilution Stock repurchases since 2018: $12 billion - Used to assess whether buybacks reduced share count meaningfully Change in shares outstanding: 47 million more shares outstanding - Evidence cited to question the effectiveness of NVIDIA buybacks OpenAI revenue in 2025: $12.5 billion - HSBC projection referenced in discussion of OpenAI economics OpenAI revenue in 2026: $35 billion - HSBC forecast of rapid growth OpenAI revenue in 2027: $67 billion - HSBC forecast continued acceleration OpenAI revenue in 2030: $214 billion - HSBC forecast showing extreme upside in TAM realization OpenAI operating loss in 2025: $17.7 billion - Projected losses despite strong revenue growth OpenAI operating loss in 2026: $80 billion - Illustrates scale of expected spending and losses OpenAI operating loss in 2027: $104 billion - Shows losses continuing to widen OpenAI operating loss in 2028: $108 billion - Peak loss level in the HSBC projection OpenAI operating loss in 2029: $108 billion - Losses remain elevated OpenAI operating loss in 2030: $76.6 billion - Losses begin to decline only at the end of the forecast period ChatGPT users by 2030: 3 billion - HSBC bullish demand assumption despite projected losses OpenAI valuation: $500 billion - Referenced as part of the scale of market enthusiasm around the company Fiscal AI discount: 30% off through December 1st - Sponsor promotion repeated several times during the episode Fiscal AI enterprise plan: Less than $1,700 per year - Used to compare cost versus Bloomberg terminal pricing Bloomberg terminal comparison: Less than one-tenth the cost - Positioning Fiscal AI as a cheaper alternative for stock-focused investors December Fed cut probability: 85% - Market pricing mentioned near the end of the episode

Pivotal Quotes: "This is the most important topic that has happened in macro event, really, in the past five, maybe 10 years." — Host: Opening framing of Michael Burry’s Substack and the AI debate "I think the center of the bubble, if this is a bubble, is not NVIDIA." — Max: Argument that the real risk is in unprofitable AI buyers like OpenAI, not the chip supplier "The eye of the storm of the bubble hurricane is OpenAI, not NVIDIA." — Host: Direct statement that OpenAI is the most vulnerable part of the AI trade

Implications: Listeners should watch the AI trade through the lens of buyer economics, not just chip-seller profits. If OpenAI-style funding becomes harder, the ripple effects could hit NVIDIA, Oracle, and broader markets, while strong AI capex may also distort macro data like GDP.