Special treatments — per-company audit view

Every human-authored rule governing extraction for the 13 tracked companies

13 companies · 26 rules · 0 human notes
0700

Tencent Holdings Limited

FYE: December

Category: chinese_hyperscaler · Currency: CNY · Coverage from 2015-01-01 · Approach: llm_primary

Quarterly convention: semi_annual

Company notes

  • HKEX-only filer. Reports in RMB/CNY. Cloud revenue is bundled
    into "FinTech and Business Services" segment which also includes
    payment services (WeChat Pay) — requires LLM judgment to isolate
    the cloud portion, or accept the segment as a proxy. Before ~2019,
    cloud was not separately identifiable in segment reporting.
    KNOWN LIMITATION: Tencent only files HKEX annual (HK-AR) and
    semi-annual interim (HK-IR) reports — no true quarterly
    disclosures. Chart will therefore show annual bars for 0700 with
    NO quarterly data; this is by design, not a coverage gap. If H1
    granularity is desired in the future, extract H1 from HK-IR and
    derive H2 = FY - H1.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
named_segment_fuzzy
Segment names
"金融科技及企業服務", "FinTech and Business Services"
Segment start
2019-01-01
Extraction
pdf_pdfplumber
Adjustment method
proxy_segment
Formula
cloud_proxy = FinTech_and_Business_Services_segment_revenue
Rationale
Tencent does not separately disclose cloud revenue. Their
"金融科技及企業服務" (FinTech and Business Services) segment
includes: (1) cloud services (Tencent Cloud), (2) payment
services (WeChat Pay — the MAJORITY of the segment), and
(3) enterprise SaaS. Using the full segment as a cloud proxy
overstates cloud revenue by approximately 2-3x. No subtraction
method is available because there is no separately listed
subsidiary or sub-segment to net out.
Caveats
  • OVERSTATES cloud by ~2-3x: includes WeChat Pay payment processing
  • Tencent Cloud is estimated at ~30-40% of the segment by industry analysts
  • Filing is in Chinese (Traditional); English version published separately
  • Values in RMB, converted to USD at period-end FX rate
  • Only HK-AR and HK-IR available (no 10-Q quarterly)
Notes
CNY. HKEX PDF extraction via pdfplumber. Annual + H1 only.

Human notes (0)

— no human notes yet —

AMZN

Amazon.com, Inc.

FYE: December

Category: hyperscaler · Currency: USD · Coverage from 2015-01-01 · Approach: xbrl_primary

Quarterly convention: three_month_column

Company notes

  • AWS has been reported as a separate segment since Q1 2015 (first
    quarterly segment disclosure in the 10-Q for Q1 2015). Clean
    segment data throughout the coverage period.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
named_segment
Segment names
"AWS", "Amazon Web Services"
Segment start
2015-01-01
Extraction
hybrid
Notes
Clean segment since Q1 2015

Human notes (0)

— no human notes yet —

APLD

Applied Digital Corporation

FYE: May

Category: neocloud_pureplay · Currency: USD · Coverage from 2022-04-01 · Approach: xbrl_primary

Quarterly convention: three_month_column

Company notes

  • Applied Digital pivoted from crypto mining to HPC/datacenter
    hosting around 2022. Pre-pivot revenue is not comparable. FYE May.
    Total revenue = datacenter revenue post-pivot.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
whole_company
Extraction
hybrid
Notes
Pure-play HPC/datacenter. Total revenue = DC revenue post-2022 pivot.

Human notes (0)

— no human notes yet —

BABA

Alibaba Group Holding Limited

FYE: March

Category: chinese_hyperscaler · Currency: CNY · Coverage from 2015-04-01 · Approach: llm_primary

Quarterly convention: standalone_quarterly

Company notes

  • Alibaba's fiscal year ends in March. Reports in RMB/CNY under
    IFRS. Files 20-F annually on SEC (preferred) and HK-AR on HKEX.
    Cloud segment renamed from "Cloud Computing" to "Cloud
    Intelligence Group" in FY2024. Quarterly data only available from
    6-K press releases or HKEX interim reports — limited granularity.
    FX conversion required: CNY → USD at period-end rate.
    IMPORTANT: 6-K quarterly values are STANDALONE (not cumulative
    YTD like 10-Q). Do NOT apply de-cumulation to 6-K data.
  • FISCAL-QUARTER CONVENTION (all BABA metrics):
    period_type=Q1 == 6-K period_of_report = Jun 30 (fiscal Apr-Jun)
    period_type=Q2 == 6-K period_of_report = Sep 30 (fiscal Jul-Sep)
    period_type=Q3 == 6-K period_of_report = Dec 31 (fiscal Oct-Dec)
    period_type=Q4 == reconcile-derived, anchored to 20-F period_of_report = Mar 31
    (fiscal Jan-Mar = calendar Q1 of the next calendar year)
  • BABA does NOT file a standalone 6-K for fiscal Q4; that value is
    always derived from the 20-F annual via the identity
    Q4 = FY - Q1 - Q2 - Q3 (equivalent: Q4 = FY - 9M)
    i.e. BABA's calendar-Q1 standalone revenue equals the full-year
    20-F value minus the three 6-K quarterly press releases filed in
    the preceding calendar year. The reconcile engine (`capex
    reconcile`) fires this rule automatically for revenue and
    cloud_segment_revenue. Before 2026-04-20 the cloud extractor
    emitted CALENDAR-quarter labels; a data migration relabeled the
    existing rows to the fiscal convention — see
    `scripts/relabel_baba_cloud_quarters.py` and the `audit_log`
    `period_type_relabeled` entries for the trail.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
named_segment
Segment names
"Cloud Intelligence Group", "Cloud Computing"
Segment start
2016-04-01
Extraction
hybrid
Notes
Renamed "Cloud Computing" → "Cloud Intelligence Group" in FY2024.
CNY reporting. Quarterly values come from 6-K press releases
(standalone, Apr-Jun / Jul-Sep / Oct-Dec); fiscal Q4 (=
calendar Q1 of the next calendar year) is not disclosed
quarterly and is reconcile-derived from the 20-F:
Q4 = FY − Q1 − Q2 − Q3
See top-level BABA company notes for the full fiscal-quarter
convention.

Human notes (0)

— no human notes yet —

BIDU

Baidu, Inc.

FYE: December

Category: chinese_hyperscaler · Currency: CNY · Coverage from 2015-01-01 · Approach: llm_primary

Quarterly convention: standalone_quarterly

Restatement policy prefer restated

  • Q1 2023 iQIYI deconsolidation (cloud = Total − Online − iQIYI formula changes)

Cloud revenue is derived (`Total − Online marketing − iQIYI`);
iQIYI deconsolidation in 2023 reshapes historical comparatives.

Company notes

  • Reports in RMB/CNY. Files 20-F on SEC. Cloud segment variously
    called "AI Cloud", "Baidu Cloud", "Intelligent Cloud". Segment
    reporting granularity has improved over time but was sparse in
    early years. FX conversion required.
    IMPORTANT: 6-K quarterly values are STANDALONE (not cumulative
    YTD like 10-Q). Do NOT apply de-cumulation to 6-K data.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
derived_subtraction
Segment names
"AI Cloud", "Baidu Cloud"
Segment start
2018-01-01
Extraction
hybrid
Adjustment method
subtraction
Formula
cloud_revenue = Others_revenue - iQIYI_standalone_revenue
Rationale
Baidu does not separately disclose cloud revenue. Their revenue
is split into "Online marketing services" and "Others". The
"Others" category includes: (1) cloud services, (2) iQIYI
video membership and advertising, and (3) other services.
iQIYI (NASDAQ: IQ) is a separately listed subsidiary that
reports standalone revenue. By subtracting iQIYI's standalone
revenue from Baidu's "Others" revenue, we approximate Baidu's
cloud + other non-iQIYI services revenue. This overstates
cloud slightly (includes some non-cloud items) but is the
closest available proxy without management commentary.
Caveats
  • Overstates cloud: includes non-cloud items in Others beyond iQIYI
  • iQIYI was deconsolidated in 2025; adjustment formula changes post-deconsolidation
  • Values in RMB, converted to USD at period-end FX rate
Notes
CNY. Annual only (20-F).

Human notes (0)

— no human notes yet —

CRWV

CoreWeave, Inc.

FYE: December

Category: neocloud_pureplay · Currency: USD · Coverage from 2025-03-28 · Approach: xbrl_primary

Quarterly convention: three_month_column

Company notes

  • Pure-play GPU cloud provider. Total revenue = cloud revenue.
    IPO'd March 2025 — very limited history. First 10-K filed
    March 2026 for FY2025.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
whole_company
Extraction
hybrid
Notes
Pure-play GPU cloud. Total revenue = cloud revenue. IPO Mar 2025.

Human notes (0)

— no human notes yet —

GDS

GDS Holdings Limited

FYE: December

Category: neocloud_pureplay · Currency: CNY · Coverage from 2016-11-02 · Approach: xbrl_primary

Quarterly convention: standalone_quarterly

Company notes

  • GDS is a pure-play datacenter operator. Total revenue IS the
    cloud/datacenter revenue — no segment extraction needed, just
    use the headline revenue figure. Reports in RMB/CNY. FX
    conversion required. IPO'd November 2016.
    IMPORTANT: 6-K quarterly values are STANDALONE (not cumulative
    YTD like 10-Q). Do NOT apply de-cumulation to 6-K data.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
whole_company
Extraction
hybrid
Notes
Pure-play datacenter. Total revenue = cloud/DC revenue. CNY.

Human notes (0)

— no human notes yet —

GOOGL

Alphabet Inc.

FYE: December

Category: hyperscaler · Currency: USD · Coverage from 2015-01-01 · Approach: xbrl_primary

Quarterly convention: three_month_column

Company notes

  • Google Cloud was first reported as a separate segment in Q4 2020
    (10-K for FY2020). Before that, cloud revenue was bundled into
    "Other Bets" or not separately disclosed. Coverage for cloud
    segment revenue starts Q4 2020; prior quarters show null.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
named_segment
Segment names
"Google Cloud"
Segment start
2020-10-01
Extraction
hybrid
Rationale
Google Cloud was first reported as a separate segment in Q4 2020
(FY2020 10-K). Before that, cloud revenue was part of the
undifferentiated "Google" segment. No subsidiary or line-item
subtraction is available to derive pre-2020 cloud revenue.
Google did disclose cloud annualized run-rate in some press
releases/earnings calls (~$8B run-rate in Q4 2019) but these
are not from 10-K filings and are not comparable to the
segment disclosure.
Caveats
  • Pre-Q4 2020 cloud revenue is null — not derivable from 10-K
  • Press release/earnings call disclosures could fill this gap but require a different source
Notes
Segment data from FY2020 onward only.

Human notes (0)

— no human notes yet —

IREN

IREN Limited

FYE: June

Category: neocloud_pureplay · Currency: USD · Coverage from 2022-11-17 · Approach: xbrl_primary

Quarterly convention: standalone_quarterly

Company notes

  • Australian company, NASDAQ-listed, reports in USD. Formerly Iris
    Energy. Primarily Bitcoin mining + HPC hosting. FYE June. Total
    revenue treated as datacenter revenue. Annual data only from 20-F.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
whole_company
Extraction
hybrid
Notes
Datacenter/mining. Total revenue. USD. Annual only.

Human notes (0)

— no human notes yet —

META

Meta Platforms, Inc.

FYE: December

Category: hyperscaler · Currency: USD · Coverage from 2015-01-01 · Approach: xbrl_primary

Quarterly convention: three_month_column

Company notes

  • Meta is an AI infrastructure BUYER, not a cloud vendor. They do
    not sell cloud services and have no "cloud segment revenue."
    Tracked for headline metrics (total capex, revenue) only.
    Excluded from the cloud_segment_revenue dataset.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue excluded

Meta is an AI infrastructure buyer, not a cloud vendor. They do not sell cloud services and have no cloud segment revenue. Tracked for headline metrics (capex, total revenue) only.

Human notes (0)

— no human notes yet —

MSFT

Microsoft Corporation

FYE: June

Category: hyperscaler · Currency: USD · Coverage from 2015-07-01 · Approach: xbrl_primary

Quarterly convention: three_month_column

Restatement policy prefer restated

  • FY2018 segment_reorg (Commercial Cloud → Intelligent Cloud)
  • FY2024 segment_reorg (Intelligent Cloud scope narrowed;
  • Office365 / Dynamics moved to Productivity & Business Processes)

Intelligent Cloud revenue is restated every time MSFT changes
its segment lines. The most recent 10-K's retrospective segment
table is authoritative — prior-year values in the current 10-K
supersede the originals. `capex audit --apply` reads the
latest 10-K's comparative columns and writes restated rows;
the selector's `filing_date DESC` tiebreaker auto-promotes
them over older extractions.

Company notes

  • Microsoft's fiscal year ends in June. FY2016 started July 2015.
    Segment reporting changed in FY2018 — "Commercial Cloud" became
    "Intelligent Cloud". Prior to FY2016, cloud revenue was not
    separately reported as a segment.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
named_segment
Segment names
"Intelligent Cloud", "Commercial Cloud"
Segment start
2015-07-01
Extraction
hybrid
Notes
Renamed from Commercial Cloud to Intelligent Cloud ~FY2018

Human notes (0)

— no human notes yet —

NBIS

Nebius Group N.V.

FYE: December

Category: neocloud_pureplay · Currency: USD · Coverage from 2024-10-21 · Approach: xbrl_primary

Quarterly convention: standalone_quarterly

Company notes

  • Dutch holding company spun out of Yandex's restructuring. AI
    infrastructure business. Very limited history — only FY2024 20-F
    available. Reports in USD. Total revenue = AI infra revenue.
    Historical quarterly (pre-restructuring Yandex) is in RUB via 6-K.
    IMPORTANT: 6-K quarterly values are STANDALONE (not cumulative
    YTD like 10-Q). Do NOT apply de-cumulation to 6-K data.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
whole_company
Extraction
hybrid
Notes
AI infra. Total revenue. Very limited history (FY2024 only).

Human notes (0)

— no human notes yet —

ORCL

Oracle Corporation

FYE: May

Category: hyperscaler · Currency: USD · Coverage from 2015-06-01 · Approach: xbrl_primary

Quarterly convention: three_month_column

Restatement policy prefer restated

  • FY2018 segment rename (Cloud SaaS/PaaS → Cloud services & license support)
  • FY2025 segment rename (→ 'Cloud')

ORCL has renamed and rescoped its cloud segment multiple times,
and each rename's retrospective comparatives differ from the
originals. Prefer the latest 10-K's segment-table comparatives.

Company notes

  • Oracle's fiscal year ends in May. Cloud segment naming has changed
    multiple times: "Cloud Software", "Cloud Services and License
    Support", "Cloud". The cloud-related segment may need LLM judgment
    to correctly identify across different reporting periods.
    Oracle acquired Cerner in 2022, which significantly affected
    revenue composition.

Dataset rules

CapEx, Revenue, Operating Cash Flow, D&A, PP&E
Treatment
xbrl_default
Extraction
xbrl_primary
Cloud / DC Revenue
Treatment
named_segment
Segment names
"Cloud Services and License Support", "Cloud Services", "Cloud"
Segment start
2015-06-01
Extraction
hybrid
Notes
Segment name changed multiple times. LLM judgment may be needed.

Human notes (0)

— no human notes yet —

Generated 2026-04-21 08:12 UTC from coverage.yaml + human_notes.yaml + audit_review_feedback table.