Scale AI Rival Invisible Technologies Valued at Over $2 Billion

(Bloomberg) — Artificial intelligence startup Invisible Technologies, a competitor of Scale AI Inc., is raising $100 million in a new funding round that underscores Silicon Valley investors’ interest in the building blocks of the AI boom.

The 10-year-old startup is valued at more than $2 billion in the deal, according to people familiar with the matter who asked not to be identified discussing private information. A new firm called Vanara Capital led the financing, marking its first public investment since spinning off from private equity firm TPG Inc. in August.

Invisible’s claim to fame is its technology’s role in helping train OpenAI’s original ChatGPT. It’s part of a growing cohort of companies that work in data labeling — the sometimes tedious task of streamlining and organizing the massive troves of information ingested by AI models so that the technology can use and learn from new data.

The practice of data labeling gained mainstream recognition in June, when Meta Platforms Inc. took a 49% stake in Scale, the best-known company in the space — and boosted investors’ interest in competitors. As part of the deal, which valued Scale at more than $29 billion, Chief Executive Officer Alexandr Wang joined Meta, though the startup still operates independently.

Invisible has made an effort to differentiate itself from competitors by offering labeling in more complicated areas. If the most basic labeling tasks require identifying a picture of a cat versus a muffin, advanced jobs can require significant expertise. Invisible offers what it calls an expert marketplace, which lets AI companies working on large language models select data labelers with relevant backgrounds — for example, advanced degrees in math or astrophysics.

Vanara was attracted to Invisible because of “the expertise that their network has within their labelers and annotators,” said Hayden Lekacz, a co-founder and managing partner, who is joining the company’s board. While competition is increasing in data labeling, “Invisible really does a much better job at delivering high-complexity, high-difficulty work products to their customers,” he said.

Invisible named Matthew Fitzpatrick, the former head of McKinsey’s AI software development group, as CEO in January. The startup now has 350 employees and has doubled the size of its engineering ranks this year. With customers including Cohere Inc., Microsoft Corp. and Amazon Web Services, Fitzpatrick said 2024 sales were $134 million, double the year before. He declined to give a forecast for this year.

In addition to data-labeling services, Invisible also offers a range of other products — including tools for fine-tuning models and measuring the breadth of a model’s available data. The company also has an enterprise business that’s targeting industries like food and beverage, insurance, asset management, and health care with products for managing supply chains and generating scripts for customer service agents in contact centers. It’s now working on customer-relationship software.