Ontology

The first public equities ontology

Built from source data and grounded in SEC filings. Every company, subsidiary, filer, and relationship Kepler knows about traces back to the documents that disclose it.

Request Demo
Relationships / NVIDIA Corporation
NVIDIA Corporation
Company · 29,329 mentions · 139 filers
GraphList
MMellanoxcompanyNINVIDIA Intl. HoldingscompanySanta Clara, CAplaceJHJensen HuangpersonTSTSMCcompanyMPMeta PlatformscompanyBlackRockfundFidelity ContrafundfundAMAMDcompanyINIntelcompanyMSMicrosoftcompanyCWCoreWeavecompanyinvestorcustomerNVNVIDIA
StructurePeopleOperationsMoneyOtherstatedinferred
Every edge cites the filing that states it · 2,389 entities connected to NVIDIA

Every edge is one filer's word,
with the sentence attached

The unit of the ontology is a claim: one thing one filer said, with the verbatim sentence attached. Every claim with the same subject, relationship, and object rolls up into a single edge, and every claim stays underneath it. Click a relationship and you land on the filing and the quote.

Microsoftcustomer ofCoreWeave~67% of revenue · Q3 2025
Verbatim quote

“We recognized an aggregate of approximately 67% and 65% of our revenue for the three months ended September 30, 2025 and 2024, from our largest customer, Microsoft.”

Filer CoreWeave (CRWV)Form 10-Q, filed Nov 13, 2025Assertion StatedGrounded

From filings
to a graph you can audit

01

Read the filings

10-K, 10-Q, 8-K, S-1, DEF 14A, and 20-F filings, from the company itself and from every other filer that names it. Insider and ownership forms (3, 4, 5, 144, D) are parsed directly as structured data.

02

Extract claims, with their quotes

Each relationship is pulled out with the verbatim sentence it came from, and every quote is checked against its source before it counts.

03

Check which way it points

A second pass verifies direction, so a customer is never recorded as a supplier and a parent never as a subsidiary.

04

Resolve the entities

Names are matched against SEC company records and their aliases, so NVDA, NVIDIA, and NVIDIA CORP become one node. Entities without an SEC filer identity still get a node.

Six kinds of entities, dozens of typed relationships

Relationship names follow FIBO, the financial industry's standard ontology: we borrow its distinctions and its naming.

Entities
Companies Funds People Places Facilities Regulatory dockets
Relationship families
Structuresubsidiary of, ultimate parent of, affiliate of, joint venture partner
Moneycustomer of, lends to, guarantees, holds equity in
Operationscontracts with, leases to, operates, located in
Regulatoryregulated by, holds authorization from, litigation against
Peopleofficer of, founder of, advises, employs

See how much to trust
every connection

The graph shows how each relationship is known and how many independent filers back it, so a single passing mention never looks like a fact.

How it's known

A filing said it outright, it follows from what a filing said, or it was computed from other claims.

How many filers say it

Lines thicken with corroboration: the number of independent filers asserting the same edge.

Whether filers agree
  • Filers agree
  • Filers differ in scope
  • Filers contradict

When filers describe the same relationship differently, Kepler shows the disagreement instead of picking a side.

Who is connected to whom, on whose word, in which sentence

Kepler's agents query the graph directly while they work: they search entities, expand connections, and read the underlying claims with their verbatim quotes and the filer who wrote them. As an agent answers, it moves the graph you are looking at, so you can follow its reasoning.

See it on your coverage