7 segments · 122 companies · as of Sep 18, 2026
Data & AI Platforms
The governed data, expert feedback and licensed knowledge behind AI systems
Where it sits
Fed by: Cloud & AI CapitalFeeds: Applications & Agents · Physical AI
Models need useful data at training time and reliable enterprise context at inference time. This layer covers analytical stores, operational retrieval, data movement, expert feedback and licensed information. Buyers pay for freshness, permissions, accuracy and rights to use the content.
Why now
Consolidation puts Confluent inside IBM and Informatica inside Salesforce. Meta's Scale stake shifts expert-data purchasing, while publishers negotiate AI access. Retrieval integrations and training licenses increasingly require separate commercial treatment.
Data Clouds & Lakehouses
Analytical data platforms unite storage, SQL, machine learning and governance. Enterprise buyers value existing data gravity and controlled access; native AI tools expand consumption without requiring every dataset to move into a separate model vendor's system.
AI Data Cloud pairs governed tables with Cortex AI, Search and Analyst for enterprise data customers
leaderLakehouse, Unity Catalog and Mosaic AI connect analytics, model development and governed enterprise agents
leaderFabric and OneLake unify analytical data with Power BI and Azure model services for Microsoft enterprises
majorSageMaker Lakehouse and Redshift connect S3 data, analytics and AI development for AWS customers
majorBigQuery and BigLake connect enterprise data to Vertex AI models and governed Google Cloud analytics
majorAutonomous AI Lakehouse and AI Database 26ai expose enterprise data and Iceberg tables across major clouds
majorwatsonx.data combines hybrid lakehouse analytics with Confluent streams and governed model context
majorMaxCompute and Hologres connect enterprise analytics with Qwen and Model Studio on Alibaba Cloud
majorVantageCloud and ClearScape Analytics bring model development to established enterprise data warehouses
majorHybrid data platform combines lakehouse storage and Cloudera AI for regulated cloud and on-premises customers
majorFoundry ontology and AIP connect operational enterprise data to governed model and agent workflows
majorBusiness Data Cloud preserves SAP semantics; acquired Dremio adds an open lakehouse and federated analytics
majorDSS unifies data preparation, ML, LLM pipelines and governance across existing enterprise data platforms
challengerData 360 unifies customer records and zero-copy access to Snowflake and Databricks for Agentforce context
challengerGalaxy and Enterprise use Trino to query distributed lakehouse data for enterprise analytics and AI
challengerManaged Hudi lakehouse and ingestion pipelines keep cloud data fresh for model training and analytics
challengerAnalytical cloud engine serves low-latency SQL over large datasets for interactive apps and agent queries
challengerDataArts Studio and LakeFormation govern lakehouse data feeding ModelArts and Pangu enterprise deployments
challengerData Lake Compute provides serverless lakehouse queries over COS data for enterprise analytics and AI
challenger
Sources: news.sap.com, databricks.com, blogs.oracle.com, newsroom.ibm.com, qlik.com, snowflake.com
Databases & Vector Search
Operational and specialist databases retrieve fresh context for AI applications. Vector, keyword and graph retrieval increasingly coexist with transactions. Existing databases compete with dedicated vector engines on relevance, latency, filtering and the cost of synchronizing data.
Atlas hybrid search and Voyage AI embeddings/reranking ground enterprise agents in operational documents
leaderAI Database 26ai combines vector retrieval, SQL and transaction data for existing enterprise customers
majorHybrid Elasticsearch retrieval combines lexical search, vectors and reranking for enterprise RAG applications
leaderManaged vector search supplies retrieval infrastructure for model-based applications and agent memory
majorOpen vector database and managed cloud combine hybrid search, filtering and model integrations
challengerRust vector engine and managed cloud target filtered, production-scale retrieval for AI developers
challengerMilvus and Zilliz Cloud serve large vector indexes for enterprise RAG and multimodal similarity search
majorReal-time database adds vector retrieval and semantic caching for agent state and low-latency responses
majorHaveli-owned Capella combines JSON documents, vector search and mobile synchronization for enterprise AI
majorGraph database and GraphRAG capture relationships that complement vector retrieval in enterprise agents
majorReal-time SQL combines transactions, analytics and vector search for data-intensive AI applications
challengerTiDB distributed SQL adds vector retrieval while serving operational data for AI applications
challengerColumnar database combines fast analytical retrieval with vector indexes for AI and telemetry workloads
majorOpen retrieval database supplies embeddings, hybrid search and context storage for AI developers
challengerNeon-derived Lakebase supplies serverless Postgres for AI apps alongside Mosaic AI Vector Search
majorCrunchy Data technology underpins Snowflake Postgres; Cortex Search retrieves governed enterprise context
majorAzure AI Search and Cosmos DB ground agents using enterprise search, vectors and operational data
majorOpenSearch, Aurora pgvector and Bedrock Knowledge Bases combine retrieval with AWS operational stores
majorAlloyDB, Spanner and Vertex AI Vector Search provide managed retrieval for Google Cloud AI applications
majorDataStax Astra DB and Cassandra add operational vector data to the watsonx enterprise AI stack
majorPostgres pgvector and managed database APIs store embeddings and retrieve application context for AI agents
challengerHANA Cloud vector engine combines similarity search with transactional business data and enterprise permissions
challengerAnalyticDB and Tair vector search retrieve enterprise embeddings beside Alibaba Cloud operational data
challengerLance columnar storage and vector indexes retrieve multimodal data for local and managed AI applications
challenger
Sources: mongodb.com, cossmology.com, databricks.com, finance.yahoo.com, bixel.com
Integration & Governance
Streaming, transformation and content connectors make enterprise data usable by agents. Customers buy freshness, lineage, access controls and business semantics. Consolidated owners are shown once, with acquired products named explicitly rather than treated as independent stocks.
Confluent Kafka, Flink and Tableflow feed real-time enterprise data into watsonx following the March 17 acquisition close
leaderInformatica governance and MuleSoft connect trusted enterprise data to Data 360 and Agentforce
majorMerged Fivetran + dbt Labs combines ingestion, transformation and semantic context for enterprise AI
majorTalend integration, data quality and lineage prepare governed data products for enterprise RAG and agents
majorLakeflow, Unity Catalog and acquired Tecton feature technology prepare fresh, governed model inputs
majorDatavolo ingestion and Openflow move structured and unstructured data into Snowflake AI workflows
majorFabric Data Factory and Purview connect pipeline orchestration, cataloging and policy for enterprise AI
majorBusiness Data Cloud and Datasphere preserve business meaning and permissions for enterprise model context
majorPermissioned Box AI applies document metadata and access controls to enterprise extraction, search and agent access
majorContent Aviator and enterprise content platforms ground AI in permissioned business documents
majorDesigner and AI-assisted preparation turn repeatable analytics workflows into enterprise-ready model inputs
majorIntegration, data preparation and governance expose business datasets to Domo agents and AI services
challengerMarkLogic 12.1 and DataDirect connect governed enterprise data and vector retrieval to AI assistants
nicheBA Insight and RightAnswers connect enterprise knowledge repositories to search and generative AI
nicheData fabric unifies records without migration and preserves permissions for process-linked AI agents
nicheWorkflow Data Fabric links operational records and external sources to governed AI agents
majorAI Governance and Data Intelligence Cloud catalog model inputs, lineage, ownership and enterprise policies
majorActive metadata and a governed data catalog expose lineage and business context to enterprise AI teams
majorData Intelligence Platform catalogs trusted datasets and policies for AI grounding and enterprise data discovery
majorProcess Intelligence Graph supplies operational context and process constraints for enterprise agents
majorKafka-compatible streaming, Connect and Iceberg integration deliver fresh events to agent and model pipelines
majorAstro manages Airflow pipelines that schedule data preparation, model workflows and production agent dependencies
majorEnterprise integration and Agentic MCP expose authenticated business actions and governed data to agents
majorIntegration, DataHub and Agentstudio connect enterprise systems, governed records and tool-using AI agents
major
Sources: newsroom.ibm.com, salesforce.com, fivetran.com, qlik.com, databricks.com
Expert Data & RL Feedback
Specialist providers recruit experts, produce preference data and build tasks that teach or evaluate models. Neutrality matters after Meta's Scale investment; buyers also demand verifiable coding, scientific and professional tasks instead of undifferentiated low-cost annotation.
Data Engine and expert feedback train enterprise and frontier models; Meta owns ~49% non-voting, not control
leaderExpert language, coding and preference-data teams supply RLHF and model evaluations for frontier labs
leaderAPEX and expert task data support reinforcement learning; acquired Deeptune adds simulated software environments
majorCoding experts and verifiable tasks supply post-training data and evaluations to frontier model developers
majorAI expert network turns professional and academic knowledge into training and evaluation data for frontier labs
majorExpert operations build enterprise task data and evaluations; Cohere uses the service for agentic workflows
majorZara recruiting and vetted expert networks supply specialist model-training tasks and human evaluations
challengerAlignerr expert network and labeling platform supply multimodal training data and model evaluations
majorFlow and expert-data services combine programmatic labeling with benchmarks and model evaluation
majorExpert annotation and evaluations serve model developers; Nebius retains majority economics without voting control
challengerHuman-in-the-loop labeling and quality controls supply visual and multimodal training datasets
nicheVerifiers and the Environment Hub supply reusable RL tasks and scoring environments for agent training
challengerJapanese data platform combines human annotation, collection and model evaluation; Sony and NTT are financial backers
nicheDatumo Eval and Dataset Store combine Korean expert data with model safety evaluation for enterprise buyers
nicheSynapse combines annotation, dataset management and ML workflows; Korean and Ghanaian operations supply training data
nicheMultimodal annotation and expert workflows create preference data and evaluations for post-training
nicheIndex, Annotate and Active curate visual datasets, manage labeling and diagnose training-data quality
nicheAcquired iMerit supplies Ango Hub and domain experts for multimodal annotation, evaluation and reinforcement learning
nicheVetted participant and expert panels supply human preferences, reasoning judgments and model evaluations
nicheMultilingual speech, text and image datasets plus managed annotation support model training and evaluation
nicheExpert teams create domain-specific reasoning tasks, rubrics and preference data for frontier-model post-training
nicheDiagnosUs aggregates medical-expert judgments to label clinical images, audio and other health data
niche
Sources: hokai.io, teahose.com, mercor.com, forgeglobal.com, ca.investing.com, primeintellect.ai
Listed Training-Data Suppliers
Public-market suppliers expose investors to data collection, annotation, language coverage and expert evaluation. Customer identities are often confidential. AI-data revenue is separated conceptually from larger telecom, outsourcing or mobility businesses that own these operations.
GenAI Platform and expert data teams support fine-tuning datasets, response annotation, RLHF evaluation and red teaming
leaderVia TELUS Digital, supplies multilingual data, RLHF and model evaluation after the October 2025 full acquisition
majorCrowdGen and global language operations supply training, safety testing and evaluation to AI developers
majoriMerit unit supplies expert training, evaluation and RL data after EXL completes its acquisition in August 2026
majorAI Solutions uLabel covers text, speech, video and sensor data; expert tasks extend into RLHF and agent evaluation
challengerAI Data Solutions combines multilingual annotators, safety reviewers and model evaluators for enterprise buyers
majorAI data annotation teams label text, images and speech and evaluate model outputs for enterprise training projects
majorTrainAI supplies language-specialist annotation, evaluation and red teaming for multilingual model customers
nicheDataOcean AI, also branded Speechocean, licenses speech, text and visual training datasets; one Shanghai-listed issuer
nicheManaged expert-data projects and its Korean crowd platform label and evaluate datasets for local foundation-model teams
nicheArcade language-data collection and human validation produce multilingual text and speech datasets for model training
nicheWNS Product Data Operations supplies speech, text and image annotation; the parent completed its purchase in October 2025
majorAI training and annotation services supply multilingual labeling, RLHF and human evaluation for enterprise models
niche
Sources: trans-cosmos.co.jp, wns.com, cogitotech.com, uber.com, kr.investing.com, roic.ai
Licensed Knowledge & Financial Data
News, scientific publishers and financial-data owners sell access to scarce information. Training licenses differ from retrieval connectors and subscription entitlements. Deal amounts are included only when reported; planned integrations are not presented as completed transfers of unrestricted data rights.
Licensed Data API gives OpenAI access to forum content; Google renewal terms remain unverified after July 2026 talks
leaderOpenAI news license reportedly spans $250M over five years; the March 2026 Meta license adds up to $50M annually
majorAmazon licenses news, Cooking and Athletic content under its 2025 pact; reported annual fees are $20–25M
majorWestlaw and Practical Law legal content ground CoCounsel research and drafting for professional customers
majorElsevier licenses clinical research to OpenEvidence for ClinicalKey AI under its 2023 partnership; fees undisclosed
majorMarch 2026 OpenEvidence license covers 400+ journals and Cochrane content; commercial fees remain undisclosed
majorTaylor & Francis–Microsoft 2024–27 deal includes >$10M initial access fee and recurring 2025–27 payments
majorGoogle 2025 partnership combines K-12 content with Vertex AI/Gemini; terms do not disclose a standalone license fee
nicheLSEG News is included in ChatGPT Financial Services in 2026; broader data entitlement integration is planned
majorCapital IQ entitlement integration with ChatGPT is planned in September 2026; existing licensed rights still apply
majorSeptember 2026 ChatGPT plan links existing Moody's data entitlements; it does not grant blanket training rights
majorSeptember 2026 ChatGPT integration plans use licensed user entitlements for investment datasets; fees undisclosed
major2026 ChatGPT and Claude integrations expose licensed financial data and regulated workflows; fees undisclosed
majorStack Overflow licenses developer knowledge through 2024 Google and OpenAI OverflowAPI deals; fees undisclosed
nichePeople Inc. licenses publishing content under its May 2024 OpenAI deal; AI licensing contributes recurring revenue
nicheARC3 licenses peer-reviewed research; the August 2026 OpenEvidence agreement supplies clinical retrieval and citations
majorWeb of Science Research Assistant grounds scholarly answers in indexed papers, citations and licensed research data
majorUSA TODAY news and local archives earn AI licensing revenue through agreements with Meta, Microsoft and Amazon
majorOpenAI licenses news from Business Insider, POLITICO and other publications for attributed AI answers
majorOpenAI agreement licenses Nordic news from titles including Aftonbladet and VG for attributed AI answers
majorEnterprise MCP connects entitled users and agents to licensed financial datasets while preserving access controls
major
Sources: springernature.com, bloomberg.com, openai.com, llmpulse.ai, wsj.com, newsroom.wiley.com
Media Rights & Crawl Access
Visual and music catalogs supply licensed generation while web intermediaries meter access to publisher content. Settlements, investments and crawl payments convey different rights. Getty and Shutterstock remain separate after abandoning their merger, and Disney's proposed Sora pact has ended.
Six-year OpenAI license signed in 2023 covers image, video and music data; the Getty merger is abandoned
majorNVIDIA partnership launched in 2023 trains commercial image generation on licensed Getty content; fees undisclosed
major2025 Suno and Udio settlements add licensed music models and artist opt-in terms; Suno v6 launches in 2026
majorOctober 2025 Udio settlement licenses recorded music and publishing for a new 2026 platform; fees undisclosed
majorMusic unit invests in Stability AI in August 2026; this does not settle separate Suno/Udio copyright claims
majorMidjourney suit and cease-and-desist letters set character-licensing terms; Google and Character.AI pulled its IP
majorHuman Native adds licensed training-data access; Pay Per Crawl meters publisher-approved AI retrieval
leaderPublisher licenses and metered retrieval charge AI agents per use; rates and permitted uses remain publisher-specific
challengerGist uses licensed publisher content and shares 50% of revenue with partners; agreements include UK news publishers
challengerStock imagery licenses support Firefly training; contributor compensation and commercial-use terms govern model inputs
nicheBot protection identifies AI crawlers and agents so publishers can block, permit or govern automated content access
nicheHuman Defense and agent verification distinguish automated visitors for publisher content-access policies
nicheBook-rights marketplace secures author and publisher permission for AI training, referencing and display licenses
nicheAttribution technology tracks training-source contributions and supports music licensing and rights-holder payments
nichePublisher tools set AI content-license terms, meter retrieval and block unauthorized crawlers
niche
Sources: investor.shutterstock.com, investors.gettyimages.com, universalmusic.com, wmg.com, gov.uk, blog.cloudflare.com
Glossary
- Lakehouse
- A data architecture combining low-cost object storage with database-style tables, governance and analytics.
- Vector search
- Finds semantically similar items using numerical representations rather than exact word matches.
- RAG
- Retrieval-augmented generation supplies a model with relevant documents or data at answer time.
- RLHF
- Reinforcement learning from human feedback uses expert preferences to improve model behavior.
- Data rights
- Contractual permission to train on, retrieve, display or redistribute content; each right can differ.
Heat, spend, exposure and shares are editorial estimates. How to read the map ↗