A data scientist builds a data business
Before founding Surge AI in 2020, Edwin Chen worked at Google, Facebook and Twitter. He told Forbes that obtaining high-quality human-labeled data repeatedly proved difficult. He left Twitter to build a company around that problem, using savings accumulated during his technology career.
Surge supplies training and evaluation data. Contributors compare model responses, write improved answers and develop evaluation criteria. Its commercial product places human judgment inside the development process for AI systems.
Distribution through an existing professional audience
Chen told Forbes that his data-science blog brought the first customers. He also approached data scientists directly. His prior technical work, existing readership and startup capital were part of the starting conditions rather than results created by a model alone.
Forbes reported $1.2 billion in revenue for 2024. Chen also described the company as profitable from its early days, but the article did not provide an independently verified profit amount.
Scale and three different financial measures
An April 2025 Bloomberg News report, citing a person familiar with the business, put Scale’s 2024 revenue at about $870 million. The same report placed its year-end annualized run rate at $1.5 billion and described $2 billion for 2025 as an expectation at that time.
The completed-year revenue, year-end annualized pace and forward forecast describe different periods. None is interchangeable with audited profit. The two reports show the commercial scale of AI data services while retaining those separate measurement boundaries.
Sources and product information
Sources: Bloomberg News, April 2, 2025, published by Bloomberg Law; Forbes Australia, Phoebe Liu, September 17, 2025. Evidence grade B: reported figures, not independently audited here.
APIToken（https://apitoken.company/） provides a unified model interface, independent project keys, model status and usage records. Public reporting does not establish that either company uses the service. Available models and conditions depend on the current site.