TickerTain
TickerTain
NewsroomShortsPortfolioConvergence
NewsroomShortsPortfolioConvergence
←
▶ 7:10 · AI Economics & Business Models · AI Economics & Business Models · AI Economics & Business Models · Kedrosky bets on boring AI applications benefiting from token price collapse
episode briefing
The Information

The “Wildly Concerning” AI CapEx Debt Bubble

2025-11-19 · 5 company · 16 thematic
sentiment
1 bull4 bear0 neu
speakers
paul kadroski

General Partner at SK Ventures who argues AI infrastructure capex is forming a debt-fueled bubble analogous to historical overbuilds, and focuses fund on application-layer companies benefiting from cheap inference.

now playing · AI Economics & Business Models · AI Economics & Business Models · AI Economics & Business Models
AI Bubble / Capex Debateheadwindscore 9/10paul kedrosky
Kedrosky labels AI capex shift to debt 'wildly concerning' and historically inevitable bubble
The shift from equity to debt financing for AI data center buildout mirrors historical capex bubbles (railroads, electrification) where creditworthy companies overborrow, leading to widespr…
AI Infrastructureriskscore 8/10paul kadroski
Big tech shift from equity-funded to debt-funded AI capex is 'wildly concerning'
AI infrastructure funding has shifted from cash flow to debt — bonds, sale-leaseback structures, and private credit partnerships — because perceived capex needs exceed even the hundreds of…
Data Center Infrastructureheadwindscore 7/10paul kedrosky
Private credit and sale-leaseback structures fuel AI data center debt bubble
Hyperscalers are using sale-leaseback, JVs, and SPVs with private credit firms to fund data center buildout beyond cash flow capacity, creating off-balance-sheet leverage.
Semiconductorsheadwindscore 8/10paul kadroski
Nvidia Dominance Parallels 1920s Electrical Giants Pre-Crash
Nvidia's current market dominance mirrors 1920s electrical companies that commanded >50% market cap before the 1929 crash; naive extrapolation of training-driven demand ignores exhausted tr…
AI Bubble / Capex Debateheadwindscore 8/10paul kedrosky
Training demand exhaustion will break capex extrapolation models
Current capex projections naively extrapolate from 2017-2022 when AI data center demand was driven by training. With training corpus largely exhausted, future demand will not match past gro…
Data Center Infrastructureheadwindscore 8/10paul kedrosky
Training demand exhaustion will collapse data center utilization and token pricing
AI data center demand from 2017-2022 was anomalously driven by model training; with training corpus largely exhausted, future inference demand won't support current capex levels, causing ma…
AI Economics & Business Modelstailwindscore 8/10paul kadroski
Cheap tokens from overbuilt infrastructure will power boring enterprise apps
Massive AI infrastructure overbuild will drive token prices to marginal cost (power, light, water), enabling cost-free waste of tokens for mundane but high-value enterprise tasks like suppl…
Data Center Infrastructuremixedscore 8/10paul kadroski
Massive data center overbuild will crash token prices to marginal cost, enabling boring enterprise AI
Vast overproduction of AI compute will drive token prices down to power-light-water marginal cost, making tokens effectively free and allowing wasteful but efficient automation of mundane e…
AI Infrastructuretailwindscore 8/10paul kadroski
Kadroski: Token price collapse from data center overbuild will enable mundane but profitable enterprise AI applications
Massive overinvestment in AI data centers will drive token prices down to marginal cost (power, light, water), making compute effectively free for boring enterprise use cases like supplier…
AI Economics & Business Modelstailwindscore 8/10paul kadroski
Kadroski bets on application layer as token prices collapse to marginal cost
Massive AI infrastructure overbuild will drive token/inference prices down to marginal cost (power, light, water), enabling cheap tokens for mundane enterprise automation like supplier reco…
AI Economics & Business Modelstailwindscore 7/10paul kadroski
Token price collapse to marginal cost unlocks mundane enterprise AI applications
If AI data center overproduction causes token prices to collapse toward marginal cost (power, light, water), tokens become so cheap that enterprises can waste them with impunity on boring b…
AI Economics & Business Modelstailwindscore 8/10paul kadroski
Token Price Collapse to Marginal Cost Will Enable Mundane Enterprise Automation
Massive overproduction of AI compute will drive token prices down to marginal cost (power, light, water), making it economical to 'waste' tokens on boring but high-value enterprise tasks li…
AI Economics & Business Modelstailwindscore 7/10paul kedrosky
Cheap tokens from overbuilt infrastructure will enable boring but valuable applications
If AI data centers are vastly overproduced, token prices will collapse to marginal cost (power, light, water), making tokens so cheap they can be 'wasted with impunity' on mundane enterpris…
AI Economics & Business Modelstailwindscore 8/10paul kedrosky
Kedrosky bets on boring AI applications benefiting from token price collapse
Overbuilt AI infrastructure will drive token prices to marginal cost (power, light, water), enabling cheap tokens for mundane enterprise tasks like supplier reconciliation, creating value i…
Grid & Power Infrastructureriskscore 7/10paul kedrosky
Regulators mistakenly treat data centers as strategic factories, inflating bubble
Regulators view data centers as 'factories of the new industrial revolution' and promote their buildout for national competitiveness, combining industrial policy with existential geopolitic…
AI Regulation & Policyriskscore 7/10paul kedrosky
Regulators' 'factories of industrial revolution' narrative fuels largest bubbles in history
Regulators' framing of data centers as essential national infrastructure in a geopolitical race encourages maximal capex and debt, historically a recipe for massive bubbles.