Big Technology Podcast — Inside The AI Bubble: Debt, Depreciation, and Losses — With Gil Luria

Host: Alex Kantrowitz
Guest: Gil Luria, Head of Technology Research at D.A. Davidson
Date: November 14, 2025

Episode Overview

This special report of Big Technology Podcast, hosted by Alex Kantrowitz, features Gil Luria for a deep-dive into the current state of the "AI bubble." The episode critically examines whether the massive investment in AI is justified, or if it risks mirroring past economic bubbles. Core discussion points include the roles of debt, asset depreciation, speculative behavior by tech giants and startups, and potential systemic risks. Luria uses his financial expertise to break down healthy versus unhealthy AI investment styles, placing particular focus on companies like OpenAI, Meta, Oracle, and CoreWeave, and explores broader market and societal implications.

Key Discussion Points and Insights

The AI Bubble Debate: Bubble or Opportunity?

[02:15–04:21]

• Alex introduces the "AI bubble" discussion, noting contrasting views:

  • Some see it as a moment of insatiable demand and boundless opportunity, referencing bullish forecasts from industry leaders such as AMD CEO Lisa Su and Anthropic's profitability targets.
  • Others are wary of unsustainable speculation.

• Gil’s take: “Both things are true.”

  • AI is revolutionary, akin to the internet or industrial revolution (“All you need to know... is just use them.” — Gil, 04:21).
  • Demand is huge, and large players like Microsoft, Amazon, and Google are investing responsibly, using cash with risk in mind.
  • Unhealthy, bubble-like behaviors are emerging, mainly among companies overleveraging and relying on speculative customers (e.g., CoreWeave and Oracle).

Healthy vs. Unhealthy Investment Behaviors

[04:21–08:13]

• Healthy behaviors:

  • Large, diversified firms with robust customer bases (Microsoft, Amazon, Google) investing from cash flow.

• Unhealthy behaviors:

  • Startups like CoreWeave borrowing heavily to build data centers, often dependent on contracts with speculative customers (OpenAI, also a loss-maker).

• Quote:

  “CoreWeave is the poster child for the bad behavior… borrowing money to build data centers for another startup... both losing tremendous amount of cash...” — Gil (06:03)

The Danger of Debt in AI Data Center Build-Outs

[08:13–12:19]

• Debt vs. equity in financing:

  • Debt is suitable when backed by predictable cash flows or strong assets.
  • AI data centers, funded with debt in anticipation of large but speculative AI contracts, don’t meet these criteria.

• Oracle’s risky exposure:

  • Betting its balance sheet on OpenAI’s growth promises, which are also promised to others—total commitments far exceeding OpenAI’s current or realistic revenue potential.

• Quote:

  “If you’re borrowing money to make a speculative investment based on a speculative customer, that’s bad behavior...” — Gil (11:41)

Systemic Risk: Could This Spark an Economic Crisis?

[12:19–15:21]

• Impact depends on the debt’s scale:

  • Tens of billions: losses contained to lenders and investors.
  • Hundreds of billions: systemic risk, potential economic contagion (analogous to the 2008 financial crisis).

• Cites “The Big Short” as a cautionary tale for this precise mechanism.

Who’s Lending and Why? Incentive Problems

[15:21–19:16]

• Major banks (US Bank, JP Morgan, Mitsubishi) are issuing loans to speculative ventures like CoreWeave.

• Why?

  • Institutional mandates to “have more AI” exposure, misaligned incentives, and lack of personal downside for dealmakers.

• AI market and pricing is too nascent and volatile to underwrite responsibly.

• Quote:

  “It’s because they don’t have the downside, right? ... you get a big bonus... if the data center is worthless in three years, you don’t care.” — Gil (18:13)

The Problem of Asset Depreciation (Michael Burry’s Critique)

[19:16–26:43]

• Michael Burry warns: Companies are extending the “useful life” estimates for AI hardware (e.g., Nvidia chips) to 5–6 years, artificially boosting earnings.

• Tech advances are so rapid that older chips become economically obsolete within 2–3 years, not 5–6.

• If depreciation is accurately stated, many companies would see profits overstated by 20–30%, risking investor shock and devaluation.

• Quote:

  “These chips will only generate meaningful revenue for three years, [and] their profitability would decline very dramatically.” — Gil (24:50)

Counterarguments on Depreciation — Is Burry Wrong?

[25:08–30:34]

• Some argue that hardware lifespan is longer (supported by warranties and anecdotal usage).

• Gil counters: economic lifespan depends not on operational survival but on ability to generate revenue—once newer chips outclass them, “it’s like used cars during COVID... not a sustainable situation” (29:01).

• Eventually, chip value equals their capability, and older chips lose most of their worth.

• Quote:

  “You can have a working chip... but will it be able to do enough computation to generate revenue? That’s a completely different question.” — Gil (29:54)

Implications for Company Valuations

[30:34–32:31]

• If companies are forced to accelerate depreciation, reported profits would drop instantly, and stock values could be cut by 40% or more.
• Even Big Tech giants would feel earnings pressure, but smaller and leveraged players would be at existential risk.

Is the AI Party Over If OpenAI Delivers?

[33:28–35:41]

• Three future scenarios:

  • Pessimist: AI is another hype cycle, not transformative.
  • Optimist (most likely): AI drives productivity/GDP growth, justifying significant but prudent investment.
  • Maximalist: Near-future superintelligence—winner takes all.

• Quote:

  “Most of us are in this optimist camp: we should invest a lot, just be thoughtful and careful about how we do it.” — Gil (35:19)

• OpenAI’s success could sustain the ecosystem, but its recent aggressive expansion and overcommitment pose big risks.

The “Ugly Underside” of Big Tech AI Profits

[43:28–49:24]

• Major AI revenue for Microsoft and Big Tech comes from AI startups (especially OpenAI, which operates at a loss).

• Microsoft’s Azure benefits directly, but this source is risky as it relies on OpenAI’s negative gross margins and future pricing adjustments.

• Pragmatic view:

  • The “Uber analogy”—early losses may convert to high-value, high-priced, sticky services in the future.
  • But AI’s market is not winner-take-all (unlike Uber); competitive forces mean price hikes may take time or never become feasible for all.

Meta, Special Purpose Vehicles, and Hidden Risks

[49:24–52:30]

• Meta, lacking business customers for AI infrastructure, is borrowing through “special purpose vehicles” (SPVs)—a trick reminiscent of Enron.
• SPVs allow Meta to obscure true debt levels and risk exposure; this alarms the market and recalls past corporate scandals.

Market Game Theory & The AI Prisoner’s Dilemma

[52:30–56:46]

• “Game theory” dynamics are keeping prices low and losses high:

  • Dominant AI players (Microsoft, Google, Meta, Amazon) are acting as if it’s “winner take all”, incentivizing underpricing and massive spending to take or preserve market share.
  • This dynamic could prevent profitability and keep losses chronic, forcing smaller competitors out.

• Quote:

  “They are willing to do anything to win... only the biggest, most deepest pocket player can win... OpenAI has no chance because they can’t make it through another year or two at this level of spending.” — Gil (55:01)

Collateral Crunch & Systemic Risk

[56:46–59:05]

• If AI asset values collapse (“collateral crunch”), a tightening of credit markets (“credit crunch”) could follow, echoing financial crises.
• Big Tech expects smaller, overleveraged players to go bankrupt—allowing giants to pick up assets at bargain prices when the dust settles.

The Looming Power Bottleneck

[59:05–61:19]

• Physical power (electricity) now a key constraint—companies are running out of ways to plug in new hardware.
• Market will likely “find a way” with creative, if expensive, workarounds (e.g., private generators, premium pay for electricians).
• Quote:

  “If you’re an electrician right now, or an HVAC technician, boy are you making bank... flown on private jets... to install a data center.” — Gil (61:15)

Notable Quotes & Memorable Moments

• On unhealthy lending:

  “It’s because they don’t have the downside... If the deal goes sour... you don’t care. You’re not giving your bonus back.” — Gil Luria [18:13]

• On asset risk:

  “If you’re borrowing money to make a speculative investment based on a speculative customer, that’s bad behavior.” — Gil Luria [11:41]

• On AI’s market reality:

  “You can have a working chip... but will it be able to do enough computation to generate revenue? That’s a completely different question.” — Gil Luria [29:54]

• On potential for systemic risk:

  “Once that capacity even gets close to catching up, then the old chips... will be worth a fraction. This is where the market’s going to end up.” — Gil Luria [28:22]

• On competitive game theory:

  “They are willing to do anything to win... only the biggest, most deepest pocket player can win... OpenAI has no chance...” — Gil Luria [55:01]

Key Timestamps

• 02:15 — Alex sets up the bubble/boom argument
• 04:21 — Gil: “AI is clearly revolutionary, but there’s unhealthy bubble behavior too.”
• 08:13 — Debt in data center investments explained
• 11:41 — "Speculative investment based on speculative customers = bad behavior"
• 15:56 — Who’s lending and why?
• 18:13 — Lender incentives and risk
• 19:16 — Michael Burry’s depreciation fraud warning
• 24:50 — Economic, not just physical, lifetime for chips
• 29:54 — Working chips vs revenue-producing chips
• 31:04 — The valuation risk of correct depreciation
• 35:38 — Gil’s three scenarios: pessimist, optimist, maximalist
• 43:28 — Big Tech’s profits tied to AI startup losses
• 45:34 — Uber pricing analogy for AI services
• 49:56 — Zuckerberg as AI maximalist
• 51:12 — Meta’s use of special purpose vehicles (“infinite money glitch”)
• 52:30 — AI prisoner’s dilemma and persistent losses
• 55:01 — Game theory and OpenAI’s limitations
• 56:46 — Systemic “collateral crunch” risk
• 59:57 — Power bottleneck as a new AI constraint
• 61:19 — Electicians and HVAC techs: hidden winners in the AI boom

Conclusion

Gil Luria's nuanced analysis provides sobering warning signs, spotlighting unhealthy speculative trends amid genuine technological revolution. He differentiates robust, sustainable investments from behavior reminiscent of classic bubbles—flagging debt-fueled expansion and aggressive asset depreciation as critical vulnerabilities. Ultimately, he suggests that while the underlying technology is real and world-changing, the current financial environment includes clear echoes of past bubbles, risking systemic disruption if correction doesn’t come soon. The real winners may be those with patience, cash flow, and discipline—or perhaps, more unexpectedly, the electricians powering the AI revolution.

Final word:

“This is all going to make a great movie one day. Gil. Yes, hopefully not as devastating.” — Alex Kantrowitz [61:19]