---
id: "concept-stranded-assets"
type: "concept"
source_timestamps: ["\\\"§ Industrial Policy", "Capital Flows", "and the Geopolitics of AI\\\""]
source_url: "https://hbr.org/2025/10/is-ai-a-boom-or-a-bubble"
source_title: "Is AI a Boom or a Bubble?"
tags: ["capital-allocation", "market-correction", "infrastructure", "tech-bubbles"]
related: ["concept-geopolitical-ai-acceleration", "claim-enterprise-lag", "prereq-dot-com-bubble", "question-enterprise-demand-timing", "claim-bubble-timing-distortion"]
definition: "Overbuilt AI infrastructure, such as data centers and chips, that becomes economically unviable if enterprise adoption and demand fail to meet speculative projections."
sources: ["futures"]
sourceVaultSlug: "hbr-seg-futures"
originDay: 2
articleStem: "hbr-foci-74-ai-boom-or-bubble"
sourceUrl: "https://hbr.org/2025/10/is-ai-a-boom-or-a-bubble"
sourceTitle: "Is AI a Boom or a Bubble?"
---
# Stranded AI Assets

The risk that massive, **preemptive** investments in AI infrastructure (chips, data centers, foundational models) result in oversupply and unused capacity if enterprise demand fails to materialize at the projected rate.

The author draws a direct parallel to the [[prereq-dot-com-bubble|dot-com crash]]: speculative capital flooded into telecom infrastructure based on exponential internet-demand projections, ultimately leaving behind vast networks of unused **"dark" telecom fiber** when adoption lagged. Today's venture funding and sovereign-wealth investments (accelerated by [[concept-geopolitical-ai-acceleration|geopolitical acceleration]]) are similarly betting that demand will eventually catch up to front-loaded capital — an open bet tracked in [[question-enterprise-demand-timing|will enterprise demand materialize in time?]]. The mechanism ties directly to [[claim-enterprise-lag|lagging enterprise adoption]].

> **Enrichment / counter-perspective:** The risk framing (overbuild now, demand later) is strongly supported by historical analogy and current capex data — analysts estimate 15–25% of S&P value tied to AI expectations could correct if earnings disappoint (IR-Impact). **BUT** IR-Impact and NBER work also stress that chips and data centers **retain long-run economic value** — dark fiber was eventually absorbed. The realistic risk is *timing and capital efficiency*, not complete economic uselessness; "fully stranded" vs "temporarily under-utilized" is genuinely uncertain (see [[claim-bubble-timing-distortion]]).


## Related across articles
- [[claim-capex-obsolescence]]
- [[action-plan-ai-bust]]
- [[question-ai-boom-or-bust]]