“Big Short” Investor: One Company Could Break The AI Boom

Finance-focused Summary (Markets, Investing, Macro, Risk, Company Financials)

Core Thesis: AI Concentration + Macro Rates Are the Key Near-Term Risks

Guest Steve Eisman argues markets are increasingly driven by rates/yields and oil, with the AI trade acting as the dominant narrative risk.

• Biggest near-term risk: concentration risk in AI hyperscalers’ revenue coming from just two model providers:
  • OpenAI + Anthropic are cited as responsible for about ~70% of AI hyperscaler revenue.
• Eisman warns that if either Anthropic or OpenAI “gets into trouble,” it could spill across “the whole chain” of the AI ecosystem.

---

“Token Maximization” vs. Open Models (Business Model / Competitive Dynamics)

Eisman says the industry’s earlier “token maximization” phase—aggressively spending until budgets were exhausted—is largely over.

• He describes prior behavior as: using AI “until you feel sick… track you… leaderboard… use it 24/7.”
• As businesses became more cost-aware, that pattern reportedly changed.

Competitive shift:

• Open-scale models are taking more share.
• This pressures the duopoly (OpenAI/Anthropic) and reduces “protective barriers.”

Regulatory angle (Eisman’s speculation):

• He suggests some firms may be motivated to create a crisis to push regulation that disadvantages open models, restoring a duopoly.

---

Anthropic Financials (Reuters) and IPO Timing Logic

Using a Reuters-based discussion:

• Revenue: $4.6B “last year” (>1,000% growth vs. 2024)
• Operating losses: over $8B (described as the key real number)
• Net loss: “almost $42B,” but includes non-cash revaluation effects from financial instruments

Eisman’s caution/interpretation:

• He downplays the predictive value of 2025 and focuses on whether profitability is sustained later.
• He frames IPO timing as speculative (“conspiracy theory”):
  • He suggests the IPO may align with a stronger first half:
    • Q1/Q2 strength from the cost-insensitive “token maximization” period
    • potential slowdown into Q3/Q4 after “token maximization” ends
• He wants comparison of H1 2026 vs. H2 2026, rather than earlier periods.

---

Risk of a “Price War” and the Lack of Long-Lasting Barriers

Eisman claims competitors are beginning to enter a price war, which he calls “a very bad thing.”

• He contrasts this with older tech monopolies (e.g., Google Search barriers).
• His view: current AI model providers have no durable moats—leadership can change quickly.

---

Bubble / Valuation Commentary: Narrative Matters More Than Multiples

Eisman is skeptical that the market fully prices concentration risk.

• He says he has “no [expletive] idea” what is already priced in.
• Instead, he reframes risk as narrative-driven:
  • If the AI narrative holds → markets can rise.
  • If the narrative collapses → a sharp correction can happen regardless of valuations.

Valuation reference:

• He cites the Shiller ratio ~41 as the second-highest ever (only higher at the end of 1999),
• but says valuations are not his main problem.

---

Macro: Fed, Yields, and the “Rubicon” at 5% (10-Year)

Macro events mentioned:

• Fed raised rates for the first time in 3 years
• Treasury yields hit multi-year highs
• Inflation cited at 3.4%

Eisman’s framing on rates:

• “5% on 10-year bonds” is described as the “Rubicon.”
• He argues: unless rates fall below 5%, a market correction seems inevitable (in his quoted view).

Transmission mechanisms:

• Higher mortgage rates reduce housing demand
• Higher corporate borrowing costs slow or raise costs (notably data center financing)
• Deficit/credit stress adds to economic “creaking”

---

AI Debt and Off-Balance-Sheet Structures (Potential Credit-Rating Preservation)

Eisman states roughly $500B in AI debt is being raised this year.

He focuses on the mix of:

• on-balance-sheet vs off-balance-sheet debt (he says he doesn’t know how much is off-balance-sheet)

Why off-balance-sheet structures matter (his argument):

• Such structures may help preserve credit ratings because rating agencies and financial statements may not fully capture the risk.

Meta example (credit/rating logic):

• A “deal worth almost $30B” for a data center in Louisiana reportedly involved debt not reflected on the balance sheet (based on his read of 10-K audit language).
• He analogizes the logic to historical cases, explicitly mentioning Enron.

AI infrastructure credit sensitivity:

• Example: CoreWeave borrowing costs ~9% earlier; could be ~11% with higher rates.
• Higher rates may make it harder to charge customers, potentially hurting weaker-financed firms’ viability.

---

Data Center / Credit Risk Signals (Oracle + Blue Owl; FMOA Warning)

Eisman discusses a report that Oracle issued a force majeure notice to Blue Owl regarding a New Mexico data center project:

• Project referred to as “Jupiter.”

Eisman’s interpretation:

• He believes Oracle is acting to protect its credit rating.
• Oracle’s rating is cited as “BBB minus” (one level above junk).

Entities he says he’s most worried about:

• Oracle (BBB-/near-junk)
• CoreWeave
• other weaker AI-infrastructure borrowers

---

“Crowding Out”: AI Debt Competing with Treasuries

Eisman argues AI debt contributes to higher 10-year Treasury yields via crowding out:

• U.S. Treasury deficit cited as $6T this year
• Add ~$500B AI debt → yields rise toward about ~5.21% (compared with ~4.75% earlier in his discussion)

He dismisses alternative explanations for yields (including an “inefficient market” framing of a proposed Treasury intervention):

• He references “Operation Twist” and says it failed
• He also challenges arguments that Treasury markets were “wrong”

---

Practical Investing Stance and Hedging

Oil/rates hedging:

• Eisman says he “doesn’t do anything” to hedge oil/rates timing.
• Rationale: if geopolitical regimes change suddenly, rate/oil hedges could lose money.

AI paper short positions:

• He previously opened short positions on “four of my AI papers,” but does not recommend it.
• If doing such shorts, he says positions must be closed by end of January due to tax treatment.

---

Short Position Example Outside AI: FICO

Eisman mentions a short in FICO.

Rationale:

• He argues FICO raised prices dramatically: +1600% over ~5 years
• He claims regulators pushed alternatives:
  • VantageScore referenced as gaining share

Catalyst:

• He says Rocket Mortgage (largest mortgage lender implied) announced it would use VantageScore only
• He claims VantageScore share could rise from ~8% to ~50%

Market reaction:

• FICO stock fell “almost 20%” pre-open

Framing:

• He treats this as a short less tied to macro/AI cycles—“in its own universe.”

---

Step-by-Step / Methodology Frameworks Mentioned

Narrative Monitoring Framework (Implicit / Repeated)

• Track whether the AI narrative is strengthening or breaking
• Watch for market share shifts:
  • specifically whether open-scale models take share from expensive models
• Use forthcoming filings (notably Anthropic’s S-1) to validate what’s happening

Risk-Focus Framework

• Concentration risk: if OpenAI/Anthropic are stressed → broader ecosystem risk
• Credit risk: assess debt quality and whether risk is disguised via off-balance-sheet structures
• Macro trigger: watch whether 10-year yields stay above/below 5% (“Rubicon”)

---

Key Numbers and Explicit Cautions

• ~70%: AI hyperscaler revenue concentration tied to Anthropic + OpenAI
• Anthropic
  • $4.6B revenue (“last year,” >1,000% growth vs. 2024)
  • > $8B operating loss
  • ~$42B net loss (with major non-cash revaluation components)
• Macro / rates
  • Inflation cited at 3.4%
  • “Rubicon”: 10-year at 5%
  • 10-year yields discussed around 5.21% vs 4.75%
• Debt / credit
  • $500B AI debt raised in the year discussed
  • CoreWeave: ~9% then possibly ~11% with higher rates
  • Oracle rating cited as BBB minus
• Valuation reference
  • Shiller ratio ~41 (2nd highest ever; higher only at end of 1999)

Stated stance:

• He does not recommend proactive oil/rate hedging to profit from guessed geopolitical outcomes.
• Short-hedge approach (AI papers) exists but must be closed by end of January due to tax treatment.
• He urges investors to focus on:
  • direction of the AI narrative
  • rate/oil regime in the near term

---

Disclosures / Disclaimers

• The episode includes typical investment/sponsorship disclosures (e.g., “investing involves risks,” fee/terms language).
• No clear “not financial advice” statement appears in the provided subtitles excerpt, but the content is presented as commentary/discussion.

---

Tickers / Companies / Assets / Sectors Mentioned

Companies (primary)

• Anthropic, OpenAI, Nvidia, Meta, Google, Microsoft
• Oracle, CoreWeave, Blue Owl, ByteDance
• SpaceX (and its AI brand Grok)
• FICO, Rocket Mortgage

Financial instruments / macro assets

• 10-year Treasury yields, mortgage rates, Treasuries
• S&P index and the Shiller ratio

Credit rating referenced

• BBB minus

Sectors / projects

• AI hyperscalers, data centers
• Housing sector, mortgage/credit scoring ecosystem
• Operation Twist (policy reference)

---

Presenters / Sources (as named)

• Steve Eisman — investment analyst, portfolio manager (guest; “Big Short” legend)
• Prof G Markets / Prof G Media — host/show entity (host not explicitly named in subtitles)
• Reuters — cited for Anthropic financial/business details
• CNBC — referenced for Eisman’s prior viral interview clip
• E*TRADE (Morgan Stanley) — podcast sponsor
• Fundrise (VCX) — paid sponsorship
• Hims — sponsorship
• Morgan Stanley — “Hard Lessons” podcast sponsorship