AI Infrastructure Companies in 2026
AI infrastructure companies build the software layers between raw compute and AI applications. That includes data platforms and analytical databases, training-data and evaluation pipelines, and vector databases for retrieval. GPU clouds, neoclouds, pure inference APIs and model hosting have their own directories. This category covers the data, tooling and storage layers teams use to build, train and ship AI systems.
AI Infrastructure Companies to know
We picked companies by how notable they were as of October 2026, based on reported funding and valuation, revenue or adoption figures, and press coverage. Each one had to be independent and operating, so companies acquired in 2025-2026 (such as Modular, bought by Qualcomm) were left out. Each also had to fit the broad AI infrastructure scope (data platforms, training-data and evaluation infrastructure, vector databases) better than a sibling directory. We excluded companies that mainly rent GPUs, sell inference APIs, host models or offer LLM observability. The list is ordered roughly by overall scale and visibility.
Databricks
Data and AI platform built around the lakehouse architecture
Databricks sells a lakehouse platform that combines data warehouse and data lake functions, with governance (Unity Catalog), SQL analytics and tools for building AI agents (Agent Bricks, Lakebase). It was founded by the original creators of Apache Spark from UC Berkeley's AMPLab. It reported a $7 billion revenue run-rate in Q2 2026, up more than 80% year over year.
- Headquarters
- San Francisco, California, USA
Scale AI
Training data, RLHF and evaluation infrastructure for AI models
Scale AI provides data annotation, RLHF data, model evaluation and red-teaming to AI labs, enterprises and governments, and runs the Remotasks and Outlier annotation platforms. In June 2025 Meta took a 49% non-voting stake and founder Alexandr Wang left for Meta. Scale launched the Scale Labs research division in March 2026 and named Francis deSouza CEO in July 2026.
- Headquarters
- San Francisco, California, USA
Surge AI
Human-data and RLHF annotation provider for frontier AI labs
Surge AI provides RLHF data, language data annotation and reinforcement-learning environments, mainly for frontier model developers. Reported clients include Anthropic, OpenAI, Google, Microsoft, Meta and the U.S. Army. The company was bootstrapped as of mid-2025 and reported about $1.2 billion in revenue for 2024.
- Headquarters
- San Francisco, California, USA
Snorkel AI
Data development platform and expert data for frontier and enterprise AI
Snorkel AI started at the Stanford AI Lab. It builds tools and services for programmatic data labeling, expert-curated training data and evaluation environments used to train frontier and agentic models. In September 2026 it reported more than $375M in annual recurring revenue.
- Headquarters
- San Francisco, California, USA
ClickHouse
Open-source columnar analytics database with managed cloud and AI tooling
ClickHouse develops the open-source columnar OLAP database of the same name, which began at Yandex in 2009, and sells a managed cloud service. In January 2026 it acquired the LLM engineering platform Langfuse and launched a managed PostgreSQL service for transactional and analytical workloads in AI applications. The company reports more than 3,000 customers.
- Headquarters
- Palo Alto, California, USA
Pinecone
Managed vector database for retrieval and AI applications
Pinecone offers a fully managed vector database for semantic search and retrieval-augmented generation. In 2026 it made Pinecone Nexus (a knowledge layer for agents), full-text search and bring-your-own-cloud deployment generally available. Edo Liberty founded the company, Ash Ashutosh is CEO, and it says it has more than 10,000 customers.
- Headquarters
- San Francisco, California, USA
Qdrant
Open-source vector search engine and database written in Rust
Qdrant develops an open-source vector similarity search engine written in Rust for RAG, recommendations and semantic search, and offers managed and hybrid cloud deployments. It began as a GitHub project by co-founder Andrey Vasnetsov and reports more than 250 million downloads. Named customers include Canva, HubSpot, Roche and Bosch.
- Headquarters
- Berlin, Germany
Weaviate
Open-source AI-native vector database
Weaviate builds an open-source vector database with built-in embedding generation, a Query Agent and Engram, a personalization feature. It can run as a shared or dedicated cloud service or be self-hosted. The company is remote-first and reports more than 20 million open-source downloads.
- Headquarters
- Amsterdam, Netherlands
Labelbox
Training-data, evaluation and post-training platform for AI labs
Labelbox started as a platform for building machine-learning training data. It has since added Horizon, which provides RL environments and evaluations for frontier labs, and Recursion, a platform for running AI agents across enterprise systems. The company says it has raised $189 million, is profitable, and works with more than 90% of leading US AI labs.
- Headquarters
- San Francisco, California, USA
Zilliz
Company behind the Milvus open-source vector database and Zilliz Cloud
Zilliz created Milvus, an open-source distributed vector database licensed under Apache 2.0 and now a graduated project of the LF AI & Data Foundation. It sells Zilliz Cloud, a managed version of Milvus. The company was founded in 2017 and moved its headquarters to the San Francisco Bay Area in 2022.
- Headquarters
- Redwood City, California, USA
AI Infrastructure Companies at a glance
| Company | Headquarters | Focus |
|---|---|---|
| Databricks | San Francisco, California, USA | data platform, lakehouse |
| Scale AI | San Francisco, California, USA | training data, RLHF |
| Surge AI | San Francisco, California, USA | training data, RLHF |
| Snorkel AI | San Francisco, California, USA | training data, data labeling |
| ClickHouse | Palo Alto, California, USA | analytical database, data platform |
| Pinecone | San Francisco, California, USA | vector database, RAG |
| Qdrant | Berlin, Germany | vector database, open source |
| Weaviate | Amsterdam, Netherlands | vector database, open source |
| Labelbox | San Francisco, California, USA | training data, data labeling |
| Zilliz | Redwood City, California, USA | vector database, open source |
Frequently asked questions
What is an AI infrastructure company?
An AI infrastructure company builds the software layers AI teams rely on. Examples: data platforms and analytical databases (Databricks' lakehouse, ClickHouse), training-data and evaluation providers (Scale AI, Surge AI, Snorkel AI, Labelbox), and vector databases (Pinecone, Weaviate, Qdrant, Zilliz's Milvus). GPU clouds, neoclouds, inference APIs and model hosting have separate directories.
Which AI infrastructure companies have raised the most recently?
Databricks closed a $5B round at a $190B valuation in August 2026, led by Coatue. Snorkel AI raised a $350M Series E at a $3.5B valuation in September 2026. ClickHouse raised a $400M Series D led by Dragoneer in January 2026. Qdrant raised a $50M Series B led by AVP in March 2026.
Which companies provide vector databases for AI?
Four notable vector database companies. Pinecone is fully managed and made Nexus, full-text search and BYOC generally available in 2026. Weaviate is open source with built-in embeddings and a Query Agent. Qdrant is open source, written in Rust and based in Berlin. Zilliz created the open-source Milvus and sells the managed Zilliz Cloud.
How do training-data companies like Scale AI, Surge AI and Snorkel AI differ?
Scale AI offers annotation, RLHF, evaluation and red-teaming, and Meta owns a 49% non-voting stake. Surge AI focuses on RLHF and RL-environment data for frontier labs, was bootstrapped as of mid-2025, and reported about $1.2B in 2024 revenue. Snorkel AI grew out of Stanford research on programmatic labeling, now sells expert data and evaluation environments, and reported more than $375M in ARR in 2026.
How is AI infrastructure different from AI GPU clouds or AI inference companies?
GPU clouds and neoclouds rent compute, and inference companies sell fast model-serving APIs or chips. AI infrastructure companies work on top of or around that compute. They manage and analyze data (Databricks, ClickHouse), produce training and evaluation data (Scale AI, Snorkel AI, Labelbox), or store embeddings for retrieval (Pinecone, Qdrant, Weaviate, Zilliz).
More AI company directories
- Healthcare AI CompaniesOpenEvidence, Abridge, Tempus AI and more
- AI Coding CompaniesCursor (Anysphere), Cognition, GitHub and more
- AI GPU Cloud CompaniesCoreWeave, Lambda, Modal and more
- AI Model Hosting CompaniesHugging Face, Baseten, Modal and more
- Healthcare Intelligence CompaniesIQVIA, Definitive Healthcare, Komodo Health and more
- AI Observability CompaniesArize AI, Braintrust, Langfuse and more
About this directory: OpenCurious curates this list for researchers, founders, operators and buyers. Company descriptions and headquarters are based on publicly reported information as of and change often; figures a company reports about itself are its own claims. This is an editorial selection, not a paid ranking: no company paid to be listed, and companies appear in no particular order. Company names and logos are trademarks of their respective owners and are used only to identify each company. See our editorial policy. To suggest a company or a correction, email hello@opencurious.com.
The OpenCurious Newsletter
Stay curious. Get the best of the open web.
New directories, researcher tools, and open-source finds — straight to your inbox. No spam, ever.
- New directories
- Researcher tools
- Open-source finds
Occasional emails from OpenCurious. Unsubscribe anytime. Privacy Policy