Data and models

pays vetted experts to write training data

About $2B gross annualised run rate in June 2026, reported by The Information. Gross means before contractor pay, which Sacra says is 60 to 70 percent of it

data labelling

Meta paid $14.3B for 49%

high-end human data, self-funded

About $1.2B in 2024 (Sacra estimate)

Listed data company that collects and labels AI training data, with a fast-growing China business.

Group revenue A$119.9M in the half year to June 2026 (ASX filing)

student network turned expert supplier

About $1.1B annualised gross revenue in April 2026 (Sacra estimate). Net revenue about $450M after contractor pay

Runs human feedback and training data operations for frontier labs and large enterprises.

$134M in 2024, company figure reported by Sacra

Sells data labelling software and a data service for building AI training sets.

About $50M in 2024 (Latka estimate)

Supplies vetted domain experts, such as PhDs and senior engineers, to AI labs for training data.

More than $100M annual recurring revenue by December 2025 (TechCrunch). Sacra estimates about $500M annualised in August 2026 (estimate)

Builds finished training datasets and simulated environments for AI labs, enterprises and the US government.

Annualised run rate above $350M, company figure in September 2026. Implicator reports $375M

Supplies coding and engineering experts who train, test and check AI models for labs.

About $300M annual recurring revenue at its March 2025 round (TechCrunch)

ChatGPT and its API

About $13B revenue in 2025, against about $34B of costs (reported)

Claude and Claude Code. Strongest with businesses and coders

Annualised revenue run rate above $65B at the end of July 2026, reported from an investor update (Aug 2026)

Builds the Gemini models through Google DeepMind and sells them through Search, Cloud and the Gemini app.

25% of AI chip supply (Epoch AI)

Owns about 27 percent of OpenAI and sells Azure cloud and Copilot tools that run on its models.

20M+ users (2025)

Runs Meta Superintelligence Labs and builds its Muse models, with the Llama line now scaled back.

Sales more than tripled (H1 2025)

Invests in Anthropic and sells it cloud compute through AWS.

5% of AI chip supply (Epoch AI)

European lab

About $14B valuation (Sep 2025)

ex-OpenAI founders

Annualised run rate above $100M, self-reported and unaudited

Lab that aims to build frontier open-weight models in the US, backed by Nvidia.

$8B after a $2B round led by Nvidia in October 2025. Talks at $25B reported in 2026

Chinese research lab that releases its strongest models as open weights.

About 475M yuan (about $70.7M) in the first seven months of 2026 (The Information, reported)

Chinese lab that makes chat, voice and video models, with most app revenue from overseas subscriptions.

About $30.5M in 2024 (filings, reported). About $53.4M in the nine months to September 2025 (reported)

Chinese lab behind the Kimi chatbot and the Kimi open models.

Annual recurring revenue above $200M in April 2026 (reported)

Builds the Grok models and the X platform, and is now part of SpaceX.

About $500M annualised run rate in 2026 (reported)

Chinese lab that makes the GLM models, mostly sold to enterprises, and is now listed in Hong Kong.

About $44.4M in 2024 (filings, reported)

Research lab founded by Ilya Sutskever that aims to build safe superhuman AI and has no public product yet.

No revenue reported. No model, API or product released as of July 2026

Released the Llama open models, and now releases only some of its newer models as open weights.

Sales more than tripled (H1 2025)

releases open models and sells hosted access

About $14B valuation (Sep 2025)

hosts and fine-tunes open models

About $1B annualised (Sacra estimate, Feb 2026)

Lab that aims to release frontier open-weight models built in the US, backed by Nvidia.

$8B after a $2B round led by Nvidia in October 2025. Talks at $25B reported in 2026

Chinese research lab that releases its strongest models as open weights, including V4 in April 2026.

About 475M yuan (about $70.7M) in the first seven months of 2026 (The Information, reported)

the hub where open models live

Annualised revenue above $150M, from one unofficial report. The company does not disclose revenue

Chinese lab that releases open models for chat, voice and video, now listed in Hong Kong.

About $30.5M in 2024 (filings, reported). About $53.4M in the nine months to September 2025 (reported)

Chinese lab behind the Kimi open models and the Kimi chatbot.

Annual recurring revenue above $200M in April 2026 (reported)

Chinese lab that releases GLM open models and sells enterprise services, now listed in Hong Kong.

About $44.4M in 2024 (filings, reported)

Chinese tech group whose Qwen open models are among the most downloaded, and which runs Alibaba Cloud.

AI Cloud and Compute Services revenue $7.1B in the June 2026 quarter, up 45% (company)

Makes the open-source Hermes models and the Hermes Agent.

About $36M annualised by mid-September 2026 (WSJ, reported). Expects to pass $100M by end 2026

Non-profit research lab that releases fully open models, with their training data and code.

Sells expert human feedback and training scenarios to labs that tune their models.

About $2B gross annualised run rate in June 2026, reported by The Information. Gross means before contractor pay, which Sacra says is 60 to 70 percent of it

GPU cloud that bought Weights and Biases and OpenPipe to add model tracking and fine-tuning tools.

Quarterly revenue $2.6B, up 112%. $104B backlog. Still loss-making

Provides human feedback and evaluation for model tuning, now with a large Meta stake.

Meta paid $14.3B for 49%

Supplies human feedback and training scenarios to frontier labs for post-training.

About $1.2B in 2024 (Sacra estimate)

managed fine-tuning across many open models

Annualised run rate above $100M, self-reported and unaudited

fine-tuning plus hosting

About $1B annualised (Sacra estimate, Feb 2026)

pay-as-you-go GPUs for training jobs

About $300M annualised (CEO, 2026, reported)

Developer platform that makes it easier to fine-tune and deploy open language models.

Reported sale price of $100M to $500M to Rubrik in June 2025 (reported, not official)