Video summary

Aswath Damodaran (NYU) | Private Credit Is the Biggest Loser When AI Corrects | #15

Main summary

Key takeaways

Finance

Finance-focused summary (markets, investing, valuation, credit)

“Big market delusion” (AI theme) & how investors should think

  • Damodaran argues that in large, underfunded markets (e.g., AI), both entrepreneurs and capital providers tend to be overconfident. This leads many individual pods/companies to overestimate success.
  • Even if each company is internally consistent, the combined set of valuations and revenue expectations across AI businesses can exceed what the real market can support, creating collective overvaluation.
  • Portfolio implication: AI/growth investing becomes a low-odds game:
    • If you target an 80% success probability, you should avoid the space.
    • A VC-style approach implies being right about roughly ~1 out of 5 (or 2 out of 10 in VC terms). Winners must be big enough to cover losses.
  • Practical caution: prefer teams least prone to overconfidence and that take precautionary steps if they’re wrong.

AI “value chain” and profitability timing

  • Subscriptions alone don’t solve AI economics: production is expensive, especially data centers, power, and data.
  • Example cited: running a Claude “fable” costs roughly ~$6,000 per hour, illustrating cash cost intensity as products become more advanced.
  • Core point: the value chain isn’t making money yet. Profits may accrue to different layers over time.
  • The next 2–3 years (possibly longer) could be chaotic, as big tech and LLM/platform providers fight for who captures value.

Valuation vs pricing; multiples are “lazy” when uncertainty is extreme

Damodaran draws a sharp distinction:

  • Valuation: forecasting business economics and risk over time.
  • Pricing: setting a number based on what others will pay when you avoid confronting uncertainty.

For young companies (e.g., SpaceX, OpenAI, Anthropic), he argues many investors/VCs avoid true valuation and instead “price” comparables, creating spillovers (e.g., IPO pricing influenced by SpaceX pricing).

He also challenges “too-large” market-size narratives using explicit numbers:

  • Global publicly traded companies’ revenue: ~$142T
  • Employee expense: ~$20T–$25T
  • Even in a dystopian case where every employee is replaced by AI agents, the implied ceiling is ~$25T—used to challenge narratives like a “$26T AI opportunity.”

Valuation traps in semiconductors/chips

  • Not all chip businesses are mature:
    • Nvidia and ASML are treated as mature.
    • ARM is cited as an example of a younger chip business.
  • The issue isn’t simply “chip maturity”—it’s uncertainty, which makes investors uncomfortable and blocks non-lazy forecasting (no “extrapolate the last 10 years” crutch).

GPU depreciation / accounting focus

  • Damodaran criticizes focusing on GPU depreciation schedules in AI valuation.
  • Depreciation affects earnings, but for young growth companies he argues it’s often irrelevant because investors use earnings-driven multiples or EBITDA heuristics—which he views as “lazy and sloppy.”
  • By contrast, depreciation matters more in infrastructure/mature businesses where CapEx is front-loaded.

Credit + equity interplay in AI/data centers (private credit warning)

Shift toward debt/off-balance-sheet finance

The discussion frames AI/data center financing as increasingly debt-focused, with more private credit and off-balance-sheet structures.

Damodaran’s main thesis: private credit is the “biggest loser”

  • He argues private credit is vastly overrated “for intelligence.”
  • Overreach cycle: what began as a niche (lending where banks couldn’t lend) grew into a much larger business, leading to sloppiness and attracting bad actors.
  • Return profile asymmetry: lenders earn interest (often high but capped) and typically don’t share upside, leaving them exposed when AI equity valuations compress or projects fail.
  • Risk management principle: lenders must price at a fair rate that compensates for default risk. If private credit functions like a “lender of last resort,” it implies rates were too low.
  • He warns private credit can take others down structurally during market corrections.

Equity narrative driving cost of capital = dangerous

  • He criticizes debt that’s priced based on equity narratives/market caps rather than the borrower’s current cash flow capacity.

  • Key point (direct quote):

    “You can’t make interest payments with potential and promise—you need cash flows.”

  • References include:

    • SpaceX: lenders asking “should they borrow more?” are described as “insane” while the business is losing roughly ~$2.5B right now.
    • CoreCivic: mentioned as a credit name being monitored; unsecured debt discussed with CDS around ~500 pips (as phrased in the subtitles).
    • CoreWeave: appears as a company with significant borrowing; he questions why it wouldn’t raise equity instead.

Venture debt / convertibles vs straight debt

  • He’s skeptical of venture debt due to structural misalignment:
    • Startups’ value is in future growth/promise, but debt places survival risk on the line.
  • For young companies, he recommends convertibles:
    • Better for the company (e.g., lower coupon; preserves cash).
    • Better for the lender (more protection if equity investors try to advantage themselves).
  • “Financing should act its age”:
    • Borrowing too early can destroy value by risking survival and cutting off growth.
    • Borrowing too little when tax shields exist can forfeit value for mature companies.

Overcapacity and bad-business accounting metrics (NAV / book value)

Structural overcapacity framework

Damodaran distinguishes:

  • Overcapacity in growing markets: can be a competitive advantage (e.g., building large factories early).
  • Overcapacity in shrinking markets: “born in hell,” with no attractive exit; risk becomes more fatal.
  • For investors, overcapacity is only acceptable if it aligns with market demand—otherwise it becomes a cliff risk.

Book value / NAV delusion

  • Book value is often an accounting artifact, not liquidation value.
  • Selling below NAV can be a feature of bad businesses where earnings power deteriorated.
  • Exceptions where book-ish value can approximate liquidation value:
    • Real estate (with liquidation/tax considerations).
    • Holding companies with marked-to-market public holdings (e.g., SoftBank; Alibaba is mentioned as an example of public holdings marked-to-market).

Distress risk, optionality, and how credit pricing breaks down

Valuation approach for stressed firms

  • He argues discount rates aren’t the right knob for truncation risk (risk that there is no year 6).
  • He proposes a two-path framework:
    1. Going-concern value: traditional DCF using a traditional cost of capital; cash flows may improve as smaller firms learn to survive.
    2. Failure/liquidation value: value failures as liquidation scenarios; subtract debt; equity is limited liability and can drop to zero.
  • Emphasis: weight cash flows by survival probability, rather than forcing everything into a higher discount rate.

When equity becomes like a call option

  • In deep distress:
    • Equity behaves like a call option on the firm’s assets.
    • Therefore debt behaves like an implied put option.
  • He warns lenders can end up on the wrong side if they act passively when optionality dominates.

Failure/distress “costs”

Distress costs are often under-discussed, including:

  • Losing customers
  • Worsening supplier terms
  • Underinvestment (especially in levered firms that look healthy until the “Pandora’s box” opens)

Practical investing/credit guidance and key mistakes

Debt level guidance

  • If you must make a leverage mistake, he argues it’s safer to have too little debt than too much.
  • He links overleveraging to a debt spiral:
    • Higher debt costs can reduce revenues/capacity to retain employees → distress.

Credit investor objective

  • Credit investors should focus on whether they will be paid back, not on excess return on capital (contrasting equity’s typical focus).

Common sophisticated-investor mistakes

  • Many “sophisticated” investors price, not value:
    • They use multiples/screens and avoid grappling with business-model uncertainty.
  • He warns that financial modeling can become an excuse that replaces business understanding:
    • Models are tools, but shouldn’t “run you.”

Pricing vs value divergence

  • He claims market fear/greed (and resulting moves in an equity risk premium) can affect value even if fundamentals don’t change.
  • He references his own equity risk premium work as a gauge of fear/greed, citing examples such as:
    • 2008
    • 2020 COVID
    • Early tariff announcement
    • Start of the Iran war

Instruments / tickers / entities mentioned

Companies / tickers (explicit or implied)

  • SpaceX, OpenAI, Anthropic
  • Nvidia (NVDA), ASML, ARM
  • CoreCivic (credit example)
  • CoreWeave (borrowing discussion)
  • SoftBank (holding company/book value context)
  • Alibaba (public holdings marked-to-market context)

Credit instruments / measures

  • CDS (notably discussed as “around 500 pips” for CoreCivic’s unsecured debt example)

Macro/business metrics

  • Equity risk premium

Methodologies / frameworks explicitly described

  • “Big market delusion” winner selection (behavioral / portfolio)

    • Assume low odds.
    • VC-style mindset: tolerate frequent misses; winners must be large.
    • Look for teams least prone to overconfidence and that plan for downside.

  • Valuation vs pricing framework

    • Valuation: forecast evolving business economics under uncertainty.
    • Pricing: use comparables/multiples when you won’t forecast economics.

  • Young company valuation basics

    • Step back from hype to unit economics:
      • What is the product/service?
      • What does it cost to produce?
      • What earnings/cash flows are realistic?

  • Two-path distressed-company valuation (failure risk / truncation risk)

    • Compute:
      • Going-concern value (DCF with a traditional cost of capital)
      • Failure/liquidation value (liquidation proceeds vs debt; equity limited to zero floor)
    • Treat failure as truncation risk, not just “higher discount rates.”

  • Capital structure: “financing should act its age”

    • Young: prefer convertibles / avoid straight debt overreach.
    • Mature: use debt where tax advantages exist; avoid over- or under-borrowing.

Key numbers / figures mentioned (as stated in subtitles)

  • AI market / revenue ceiling
    • ~$142T global publicly traded company revenues
    • ~$20T–$25T employee expense
    • Dystopian “every employee replaced by AI agents” ceiling: ~$25T
    • Reference to a “$26T AI opportunity” narrative
  • AI operating cost example
    • Running Claude “fable”: ~$6,000 per hour
  • SpaceX
    • Loss: ~$2.5B “right now”
  • CoreCivic
    • Unsecured debt CDS: ~500 pips (approx.)

Disclosures / disclaimers

  • No explicit “not financial advice” disclaimer appears in the provided subtitles.

Presenters / sources

  • Aswath Damodaran (NYU Stern), professor focusing on corporate finance and valuation.
  • Host/other speaker: referenced only as a podcast/show host (name not shown in the subtitles) for Fixed and Floating – the Credit Podcast.

Original video