DarkMatter AIDC is an intelligence and knowledge-mapping platform built to investigate the global infrastructure supporting artificial intelligence. It records the organizations, facilities, services, technologies, ownership structures, energy dependencies, cloud relationships, and operational links that connect the modern AI ecosystem.
AI services appear simple at the surface: a prompt enters, a response returns. Beneath that interaction is a dense industrial network of data centers, accelerator clusters, cloud platforms, power systems, cooling infrastructure, fiber routes, software stacks, corporate ownership, contractors, and financial relationships. AIDC exists to make that hidden structure visible.
AIDC is a structured intelligence environment for collecting, classifying, connecting, and reviewing information about the AI data center ecosystem. It combines a human-operated research interface with machine-assisted collection, entity modeling, relationship mapping, evidence storage, and controlled publication.
AI infrastructure is distributed across companies, facilities, regions, technologies, utilities, and service providers. Public information is fragmented across corporate websites, technical documents, filings, reports, announcements, and infrastructure records. AIDC consolidates those fragments into a navigable intelligence model.
The project focuses on the physical and organizational systems that make large-scale AI possible: hyperscale data centers, GPU and accelerator deployments, cloud regions, networking, energy, cooling, ownership, hosting arrangements, and supporting technologies.
AIDC distinguishes between discovered information, proposed intelligence, and accepted knowledge. Research can be accelerated by automation, but permanent records remain governed by evidence, context, and human review.
The project is designed to answer not only where AI data centers exist, but who owns them, who operates them, what technologies they use, what services they provide, what infrastructure they depend on, and how they connect to the wider AI economy.
Discover organizations, facilities, services, technologies, projects, regions, and infrastructure components relevant to AI data center operations.
Assign consistent types, categories, roles, technologies, domains, and operational characteristics so information can be compared across the ecosystem.
Record ownership, hosting, supply, dependency, investment, technology, energy, network, service, and geographic relationships between entities.
Preserve source URLs, excerpts, confidence, timestamps, and review decisions so claims remain traceable to evidence rather than becoming detached assertions.
Track changes in infrastructure, ownership, services, technology adoption, expansion, project status, and public reporting over time.
Prepare structured information for human analysis, machine-readable outputs, future knowledge graphs, research tools, and connected DarkMatter systems.
AIDC organizes intelligence as a connected structure rather than a flat list. Every layer adds context, from broad classifications down to individual claims and the sources that support them.
High-level divisions such as organizations, infrastructure, technologies, services, energy, networking, ownership, regions, and projects.
Physical facilities, campuses, cloud regions, operational sites, announced developments, compute deployments, and infrastructure locations.
Companies, brands, operators, landlords, utilities, cloud providers, AI laboratories, vendors, investors, technologies, and services.
Ownership, operation, hosting, supply, dependency, partnership, investment, acquisition, technology use, network service, and geographic association.
Source documents, source URLs, excerpts, extracted claims, confidence scores, proposed changes, operator decisions, and activity history.
AIDC studies each data center as part of a broader network of physical, technical, commercial, and political dependencies.
AIDC combines human investigation, structured interfaces, and controlled automation. Each stage has a defined role so that collection never silently becomes publication.
Identify organizations, facilities, technologies, services, documents, projects, and candidate relationships.
Gather source material from official pages, public records, documentation, reports, filings, and operator-provided research targets.
Convert raw information into entities, attributes, classifications, relationships, evidence records, and proposed changes.
Inspect source quality, context, confidence, duplication, relevance, and compatibility with the existing intelligence model.
Commit approved information to the AIDC knowledge base while preserving evidence, review state, and operational history.
Navigate the resulting hierarchy and relationship network to understand the architecture behind AI infrastructure.
Harvester is the intelligence gathering crawler built for AIDC. It turns research targets into controlled jobs, discovers and retrieves sources, extracts structured claims, scores evidence, stages proposed database changes, and routes every publication decision through a human review gate.
Harvester extends AIDC without bypassing its governance model. It accelerates repetitive collection and extraction work while preserving source provenance, confidence, activity logs, proposal status, and operator authority.
The current project separates execution, extraction, review, and publication into distinct responsibilities. This keeps crawler success from being confused with verified knowledge and prevents endpoint data from leaking into generic relationship records.
worker.php · orchestrationClaims queued research jobs, enforces process locking and runtime limits, retrieves source documents, records execution events, invokes extraction, and leaves every proposed database change inside the review boundary.
extractor.php · structured interpretationSends bounded document context and the research goal to Ollama, requires schema-shaped JSON, validates canonical names, records the extraction engine and source endpoint, and rejects the former regex behavior that could turn sentence fragments into entities.
Publishing description, website, sector, company type, or a primary company proposal can resolve or create the canonical company first. Review order therefore does not orphan enrichment data, while duplicate-safe resolution protects existing records.
Endpoints publish to endpoints; direct company links publish to
company_endpoints; technologies and services use their dedicated relation
layers; ownership and other graph assertions remain in the relationship model.
AIDC is built around a deliberate distinction between collected material and accepted knowledge. The system can assist discovery, correlation, and proposal generation, but publication remains a controlled decision.
Automated tools may gather evidence and stage candidate changes, but they do not receive unrestricted authority to overwrite the primary intelligence model.
Intelligence should remain connected to its originating source, excerpt, retrieval record, confidence, and review history wherever possible.
Incomplete, conflicting, or uncertain information can be retained as research material without being falsely promoted to verified fact.
Changes, worker activity, review actions, errors, and publication decisions create a record that supports debugging, auditing, and future reassessment.
Artificial Intelligence Service
│
├── Model Provider
│ ├── Organization
│ ├── Cloud Platform
│ └── Compute Dependency
│
├── Data Center / Cloud Region
│ ├── Owner
│ ├── Operator
│ ├── Tenant
│ ├── Location
│ └── Project Status
│
├── Infrastructure
│ ├── GPU / Accelerator Platform
│ ├── Network / Interconnection
│ ├── Energy / Utility
│ ├── Cooling / Water
│ └── Security / Operations
│
└── Evidence
├── Source Document
├── Extracted Claim
├── Confidence
├── Review Decision
└── Publication Record
The 0.4.2 documentation set joins the original AIDC ownership-map canon with the current Harvester implementation, corrected endpoint publication model, provenance controls, and machine-readable data strategy.
AIDC exposes structured JSON API routes and database-backed entity records so companies, endpoints, technologies, services, brands, evidence, and relationships can be consumed without scraping presentation HTML. Human pages remain the visual intelligence console; machine outputs remain explicit data contracts.