GRIDINT RESEARCH · DATA HUB

Capital spend & filings.

The AI build-out as sworn to the SEC — annual capital expenditure from the filings themselves, every point carrying its accession number, tag differences named rather than blended.

published 2026-08-27 · data cutoff 2026-08-26 · edition 1 — updated only by a new dated edition

THE REPORTThe Paper TrailRead it →

Five trails, one river.

1002003002020Q22020Q42021Q22021Q42022Q22022Q42023Q22023Q42024Q22024Q42025Q22025Q42026Q2FILINGS PER QUARTER · STACKED AREAS BY FORM FAMILYMSFTGOOGLMETAAMZNNVDA1002003002020Q22020Q42021Q22021Q42022Q22022Q42023Q22023Q42024Q22024Q42025Q22025Q42026Q2FILINGS PER QUARTER · STACKED AREAS BY FORM FAMILYMSFTGOOGLMETAAMZNNVDA
Filings per quarter, stacked areas by form family — window entries markedFilled areas: amber = insider forms (3/4/5, 144), cyan = current reports (8-K), green = annual + quarterly reports, violet = offerings and registrations, gray = everything else. The dashed verticals mark where each company’s fixed-count window begins — the rise is windows joining, NOT filing growth, and the caption says so. The chips isolate companies.source SEC EDGAR submissions API · the ~1,000 most recent filings per company · windows differ and are drawn — method form families: insider = 3/4/5 + 144 · reports = 10-K/10-Q · current = 8-K · offerings = S-*/424* · counts per calendar quarterRead the report →

Plot it on any two axes.

The shape of the records.

Records by company

MSFTMSFT: 8 of 37 (21.6%)8 · 22%GOOGLGOOGL: 8 of 37 (21.6%)8 · 22%METAMETA: 8 of 37 (21.6%)8 · 22%AMZNAMZN: 8 of 37 (21.6%)8 · 22%NVDANVDA: 5 of 37 (13.5%)5 · 14%37 records · counted from the records at build · source SEC EDGAR XBRL company facts — filed actuals only

SOURCE · SEC EDGAR XBRL company facts — filed actuals only · data cutoff 2026-08-26 · counted from the records

How this hub was made.

Figures are filed actuals from SEC XBRL company facts; guidance is never in a filing and never on this hub. Amazon and Nvidia file under their own concept (PaymentsToAcquireProductiveAssets) — the tag rides every row. Each accession links to the filing index at sec.gov. Downloads are verification extracts.

The capital spend & filings exhibit page → · The institute's method →

Every edition, dated.

  • Edition 1 published 2026-08-27 · data cutoff 2026-08-26 · first edition

A hub changes only as a new dated edition, on the institute's word — never on a schedule. Collection continues underneath; publication is deliberate.

Verify this hub.

Verification extracts — the working records with their receipts, attribution first, pointing at the origin. We never provide data; the source is the authority, this hub is the lens.

Cite this hub

GRIDINT Research (2026). Capital spend & filings data hub, edition 1 (data cutoff 2026-08-26). https://gridint.ai/data/filings
@misc{gridint_filings_2026,
  title = {Capital spend & filings data hub, edition 1},
  author = {{GRIDINT Research}},
  year = {2026},
  note = {Data cutoff 2026-08-26; source: SEC EDGAR XBRL company facts — filed actuals only},
  url = {https://gridint.ai/data/filings}
}

In pandas

import pandas as pd
D = pd.read_json("https://gridint.ai/demo/filings/explorer-records.json")
records = pd.DataFrame(D["rows"].tolist(), columns=["ticker", "fiscal_year", "capex_usd_b", "xbrl_tag", "accession_path"])

The first row of that extract, as it stands: ["MSFT", 2019, 13.9, "PaymentsToAcquirePropertyPlantAndEquipment", "789019/000156459021039151/0001564590-21-039151-index.htm"] — the values are the receipt, whatever a column is called.

Every hub →

RELATED BY SOURCE

SEC EDGARAI economics data hubthe same recordsSEC EDGARCapital spend & filings — built with GRIDINTa report from themSEC EDGARCompanies — AI economics · built with GRIDINTa report from themSEC EDGARThe AI Economya report from them