180h
Avg Lead Time
Recorded, AI infrastructure domain
4
Figures Watched
Huang, Altman, Amodei, Shotwell
—
Signal Status
Updated continuously

The Structural Question

The Stargate initiative committed $500 billion to hyperscale AI data centers. The implicit assumption: frontier AI requires massive centralized compute indefinitely.

DeepSeek's January 2025 release reported a lower training-and-inference compute footprint than prior frontier models. Edge AI inference, on-device models, and distributed architectures are in deployment. The open questions are the timeline of edge adoption and the disposition of the $500B in committed datacenter capital.

The prediction-market series and the behavioral and economic indicators below are the material this page presents; the reader determines what they imply.

Observable Market Events Tracked

Nvidia earnings guidance shift
GPU demand mix moving from datacenter to edge/inference chips
AI capex guidance reduction
Any Magnificent Seven signals slowing datacenter commitment
SpaceX IPO — completed
424B4 priced 2026-06-12; NASDAQ: SPCX. The combined entity discloses Starlink, X, and xAI/Grok as one reporting company — xAI is not broken out separately. FY2025: revenue $18.67B, net loss $(4.94)B, AI-segment capex $12.73B.
Stargate contract status
Oracle's 2025-03-10 earnings release named OpenAI as a signed cloud-agreement customer (aggregate RPO +63% to over $130B) but stated the Stargate contract itself was not yet signed as of that date.
AI chip export control markets
Policy markets affecting GPU allocation between US and China

Behavioral Watchlist

Jensen Huang
Nvidia CEO
GPU allocation language
Sam Altman
OpenAI CEO
Compute requirements commentary
Dario Amodei
Anthropic CEO
Deployment architecture positioning
Gwynne Shotwell
SpaceX President & COO
Operational signal — post-IPO reporting cadence (NASDAQ: SPCX)

Public statements are recorded weekly against a fixed, pre-disclosed set of behavioral indicators. A convergence entry is logged when the behavioral deviation reaches 2.0 and a correlated market moves in the same window; the deviation and the movement are shown, not named.

Logged — 2026-07-24
Jensen Huang's first-ever post on X backed a 25-signatory letter ("Open Weights and American AI Leadership"), joined by Satya Nadella (Microsoft) and Mark Zuckerberg (Meta); Elon Musk amplified it ("Jensen is right"). Nadella: "we are outlining a path for open-weight models to strengthen American competitiveness... while protecting national security." Zuckerberg: "Open source is a positive and important force for both empowering people and preventing centralization." OpenAI and Anthropic did not sign.

Disclosed Financial Ties

OpenAI and Anthropic carry no SEC filings of their own. What follows are facts their public counterparties disclosed about the connection, in the counterparty's own filing — cited by form and date. Nothing here is scored or ranked.

CoreWeave discloses OpenAI committed to pay up to $6.5B through May 2031, and Meta up to $21B
10-Q, filed 2026-05-08
AMD issued OpenAI and Meta each a warrant for 160M shares at $0.01, vesting on GPU-purchase milestones
10-Q, filed 2026-05-06
Amazon disclosed a $12.3B fair-value step-up on its Anthropic preferred stake in Q1 2026
10-Q, filed 2026-04-30
Meta discloses an unconsolidated data-center joint venture: $2.37B carrying value, $45.99B maximum exposure to loss
10-Q, filed 2026-04-30
Alphabet discloses data-center credit-derivative backstops growing from $16.9B to $43.8B in two quarters, plus a further $24.1B pending
10-Q, filed 2026-07-23
Constellation Energy's 20-year Microsoft PPA underlies the Crane (Three Mile Island) restart; a FERC order separately cites "data center load expansion" as a driver of PJM capacity price pressure
10-Q, filed 2026-05-11

Off-Balance-Sheet Financing — Private Credit

A Research Report tier — sourced to a third-party investigation, not an SEC filing. Presented alongside, not merged with, the disclosed-filing facts above. The $10B Blue Owl commitment and Meta's $45.99B VIE exposure in the section above were both found independently from primary filings before this report was read; they corroborate this reporting rather than depend on it.

Alphabet, Amazon, Meta, Microsoft, and Oracle carry a reported $1.65 trillion in AI-related debt that does not appear on their balance sheets, on top of roughly $1.35 trillion they report directly — via long-term lease agreements with data-center operators rather than owning the debt themselves
Nikkei Asia investigation, via Prof G Markets, 2026-07-27
The data-center operators taking on that debt are reportedly funded by private-credit providers — Blue Owl Capital, Apollo, and Blackstone are named
Prof G Markets, 2026-07-27
California State Teachers' Retirement System (CalSTRS) is reported as one of the largest investors in Blue Owl's public fund — a named channel from pension capital into this financing chain
Prof G Markets, 2026-07-27
A group of senators wrote to Treasury Secretary Scott Bessent in January 2026 urging scrutiny of the private-credit ecosystem, warning AI investment "appears to drastically exceed realistic assumptions of business and consumer demand" and could trigger "a broader financial crisis"
Prof G Markets, 2026-07-27
The Office of Financial Research — created after the 2008 crisis specifically to monitor this kind of systemic risk — has had its staff cut by half and its budget cut by more than 40%
Prof G Markets, 2026-07-27
Blue Owl Capital reportedly agreed to invest up to $10 billion in future AI data-center deals after a single 15-minute meeting with the developer
Prof G Markets, 2026-07-27

Private credit is not regulated by the SEC, the Federal Reserve, or other bank regulators, which is precisely why the $1.65T figure above is a third-party estimate rather than a line item in any single filing. Apollo's and Blackstone's own SEC filings have not yet been searched directly for a primary-source figure to corroborate their portion of this — unlike Blue Owl, whose own 10-K was searched directly (see above).

Market Reaction

A third evidentiary tier, distinct from the SEC filings above: real-time analyst and market response to the disclosed facts. Sourced to Forbes/Morgan Stanley, not a filing.

SpaceX (NASDAQ: SPCX) closed July 24, 2026 at $115.07 — down from a $160.95 IPO price and a $201.80 peak (June 16), a ~$1 trillion peak-market-cap decline
Forbes, 2026-07-26
Q1 2026 divisional results: Starlink +$1.19B operating income; Space Launch -$662M; AI division -$2.47B loss against $818M revenue
Forbes, 2026-07-26
Morgan Stanley (Adam Jonas): at a $100 share price, the market would be ascribing zero value to the AI division
Forbes, 2026-07-26

Walk-Forward Backtest — Capex Intensity

A year-by-year, no-lookahead capex-to-revenue signal, validated against a known infrastructure-buildout failure, then applied to the population — not just the famous names above — to find who shows the same pattern before it's a headline. Every candidate below was checked against public coverage first; native facts, no score.

Core Scientific: capex/revenue was 0.60 the year it filed Chapter 11 (Dec 2022, aggressive debt-financed Bitcoin-mining buildout). Post-bankruptcy it collapsed to near-zero, then re-escalated during its AI/HPC pivot to 2.29x revenue in FY2025 — nearly 4x its pre-bankruptcy intensity
10-K history, 2020–2025
Applying the same signal across the population (1,808 filers, filtered to remove REITs/shippers/utilities unrelated to AI): the top real candidates are former crypto miners pivoted to AI/HPC hosting — Nebius (7.67x), TeraWulf (6.29x), WhiteFiber (3.39x), Bit Digital (2.52x), IREN (1.14x)
population scan, CY2023–2025
Self-check result: unlike the insurer version of this backtest, every one of these is already the subject of dedicated financial-media coverage — Nebius' capex is a named Yahoo Finance headline; TeraWulf is covered specifically for its Anthropic-linked data-center lease; IREN carries a "Strong Buy" consensus across 10 analysts
self-check, 2026-07

Zero of five candidates counted as a discovery — a real, honest negative. The signal works (it reproduces the pattern that preceded a real bankruptcy), but AI/crypto- adjacent stocks currently sit under intense public and analyst attention, unlike regional insurers, where the same method found genuine under-the-radar findings. The buried-disclosure opportunity appears to be sector-dependent, not universal.

Economic Indicators Monitored

FRED: PCUOMFGOMFG
PPI Semiconductors
AI chip demand pressure
FRED: IPG3344S
Industrial Production: Chips
Supply vs demand signal
Live Intelligence
Loading…
Active signal series and convergence records visible to subscribers