Job portal Handshake shifts to AI training data, growing from tens of millions to nearly $1 billion in one year

ME News. On April 14 (UTC+8), its AI training business’s annualized gross revenue has approached $1 billion. In January this year, it was $550 million, and a year ago it was only between $5 million and $10 million. After deducting payments made to contractors, the AI training business’s net revenue is close to $300 million. Handshake’s existing university recruiting software business has an additional annualized gross revenue of about $150 million. Handshake is at least the fourth data annotation company in recent years to surpass $1 billion in annualized gross revenue. Previously, Surge AI was the first to reach this scale in 2024, and Scale AI also reached it before Meta acquired nearly half of its shares.

The three-year-old AI outsourcing platform Mercor also broke through $1 billion earlier this year (last September it was $500 million). After deducting contractor expenses, its net income is $300 million to $400 million; it has already achieved free cash flow profitability. However, after a data breach at the end of March, Meta has indefinitely paused its cooperation with Mercor, and it remains unclear whether revenue has been affected.

Among other competitors, Turing had annualized revenue of $300 million as of the end of 2024. Last week, AfterQuery announced it had surpassed $100 million and a valuation of $300 million.

These companies’ customers include AI labs such as Anthropic and OpenAI. Their core business is to organize professionals such as lawyers, PhDs, and doctors as contract workers to score models and answer questions in their respective fields. Demand from AI labs for high-quality training data produced by domain experts continues to grow, but the moat in this business is limited: leading labs can easily switch between vendors, and some companies choose to build their own teams. For example, xAI directly hired more than 1,000 remote trainers.

(Source: BlockBeats)