AI capex boom-bust risk

The 'trillions' buildout and bubble hand-wringing in the CNBC pieces signal potential overinvestment relative to near-term demand, echoing concerns that an AI data-center surge could overshoot and trigger telecom-style bust dynamics.

The rapid proliferation and scale of new AI-driven data centers, alongside rising local backlash and siting constraints described (Alabama, Memphis, NoVA, Georgia), evidences a fast investment surge whose social and infrastructure bottlenecks could set the stage for an overbuild-and-correction cycle.

The rapid acceleration and leverage—$1.5 trillion needed from investors/developers, construction lending doubling vs. 2024—signal potential overbuild and credit-cycle vulnerability in a single asset class (data centers), aligning with concerns about an AI-driven investment overshoot and subsequent bust.

Noah Smith’s warning about opaque private credit financing AI data centers, Moses Sternstein’s counterpoint about funding from hyperscaler profits, and Kling’s added concern about banks’ exposure collectively map to the risk of overbuilt AI infrastructure and who bears losses if returns disappoint—key elements of a potential AI investment bust.

Describing an AI buildout 'still growing' and surpassing prior telecom/internet booms signals classic overbuild dynamics, raising the prospect of a telecom-style bust once demand lags capacity.

The article compares today’s data-center surge to past telecom/railroad overbuilds and asks whether this capex boom could crash the economy.

The rush toward multi‑GW data centers and a $500B infrastructure pledge heighten the risk of overbuild and potential demand overshoot, echoing telecom‑style boom‑bust dynamics.

Meta’s commitment to invest hundreds of billions in compute and bring 1–5 GW superclusters online exemplifies the scale of AI data-center capex that could overshoot near-term demand, raising telecom‑style boom‑bust risks.