‘The era of data-labeling companies is over,’ says the CEO of a $2.2 billion AI training firm

• Simple data labeling is becoming obsolete as AI models require more complex training data, says Turing's CEO.
• AI training companies need to be a "proactive research partner" for major labs, Jonathan Siddharth said.
• AI data-labeling startups have been garnering massive valuations over the past year.

Basic data-labeling work — the kind built on tagging images or sorting text — is becoming obsolete, said the CEO of a $2.2 billion AI training firm.

Jonathan Siddharth, the CEO of Turing, said on an episode of the "20VC" podcast published Monday that "the era of data-labeling companies is over."

"Data needs have significantly changed," said Siddharth. Early models relied on annotators tagging images, classifying text, or performing simple tasks that could be outsourced at scale. Today's systems, like agentic models and reinforcement-learning architectures, require more complex data, he added.

"It's more real-world data, data that touches how real humans do knowledge work," Siddharth said, adding that major labs want to work with AI training companies that can be a "proactive research partner for them."

"It's now the era of research accelerators," he said.

Siddharth said AI training companies need to focus on building a reinforcement-learning environment — simulated mini-worlds — that replicate human workflows across different industries.

To do that, AI training companies must recruit human experts in various domains, Siddharth said.

Turing announced in June that it had raised $111 million in Series E funding at a valuation of $2.2 billion. Earlier this year, the AI training firm said its annual revenue run rate reached $300 million in 2024 — nearly triple the year before.

Rise of AI data-labeling companies

AI data-labeling startups have been garnering massive valuations over the past year.

In June, Meta acquired a 49% stake in Scale AI, valuing the company at over $29 billion. Mercor said in October that it closed a funding deal valuing the startup at $10 billion.

The AI training boom has also fueled a fast-growing freelance workforce. Business Insider reported in September that several freelancers and contractors said that their AI training work has provided them with thousands of dollars a month, although it can be disturbing and unpredictable. Business Insider spoke with more than 60 data labelers about their work experiences.

The demand for AI training has also created an underground market for access to these platforms. Business Insider reported on Monday that it uncovered more than 100 Facebook groups selling unauthorized access to real and fake contractor accounts. Although reselling accounts is prohibited by AI-training companies, scammers and opportunistic gig-seekers are cashing in on the growing demand for AI-training gigs.