AI accountability news — digests, data trackers, not hype
Radical Transparency

AI Bubble Tracker

AI Bubble Tracker — featured image

The Bubble Tracker is LLM Watch’s living dashboard for AI infrastructure ambition versus cash-flow reality. Capex announcements move markets; durable revenue and adoption move slower — we keep both in view.

What we track

  • Hyperscaler and lab capex commitments (cloud campuses, GPU clusters, “Stargate-scale” programs).
  • Enterprise revenue and retention signals that substantiate — or contradict — the spend.
  • Revisions from banks and vendors when the gap widens.

How to read a tracker update

Every update should answer: what was promised, what was spent or booked, what revenue or usage offsets it, and what remains unverified. No chart without a source.

Related analysis lives in Markets and The Weekly Take.


Live tracker

AI Bubble Tracker

Updated: 2026-09-21T06:41:35Z

Hyperscaler capex guidance keeps climbing toward roughly $730B-$933B in 2026-era spend, but the clearest demand-side signals in the last two weeks come from Anthropic and OpenAI adoption/revenue data, not from the infrastructure spenders themselves.

Entity Capex / spend signal Revenue / demand signal Gap read
Microsoft~$175B 2026 capex guidance midpoint; $35B reported in latest quartern/d disclosed AI-specific revenue split; Azure AI growth cited qualitativelySpend disclosed in hard numbers; AI-specific revenue attribution remains opaque
Alphabet~$195-205B 2026 capex guidance; $24B reported in latest quartern/d disclosed AI-specific revenue split; Search/Cloud AI growth cited qualitativelyCapex guidance far more concrete than any AI-attributable revenue figure
Amazon~$220B 2026 capex guidance, highest among the fourn/d disclosed AI-specific revenue split; AWS growth cited qualitativelyLargest capex commitment paired with the least granular AI revenue disclosure
Meta~$130-145B 2026 capex guidance; $19.4B reported in latest quartern/d disclosed AI-specific revenue split; ad-efficiency gains cited qualitativelySpend rising sharply while AI monetization remains bundled into core ad business
7-company hyperscaler/AI buildout set$657.1B trailing-4Q capex; $213.7B latest quarter (per Supercycle tracking)n/d aggregate AI-specific revenue disclosed across the groupCapex tracked in granular detail; matching revenue evidence is largely absent
OpenAIn/d direct capex disclosed; reliant on partner/cloud infrastructure deals~$40B annual revenue run rate (Rhodium, Sept 2026)Revenue run rate growing but still a fraction of partner capex committed on its behalf
Anthropicn/d direct capex disclosed; reliant on partner/cloud infrastructure deals~$65B annual revenue run rate; 43.8% enterprise adoption share (Ramp, Sept 9 2026)Leads OpenAI on both revenue run rate and enterprise adoption in latest reads
Combined 7-firm capex forecast (Amazon, Microsoft, Google, Meta, Oracle, CoreWeave, SpaceX)~$933B projected 2026 capex (JPMorgan), up from $443B in 2025n/d unified revenue benchmark exists at this aggregate levelMore than doubling capex forecast has no equivalently scaled revenue benchmark yet
Sources
  • Supercycle capex tracking, Sept 13-16 2026
  • August 2026 hyperscaler earnings roundup
  • Ramp AI Index, Sept 9 2026
  • Rhodium via Sept 17-18 2026 revenue report
  • JPMorgan capex forecast coverage, Sept 2026