Data rates for training: Meta invests $ 14.3 billion in Scale AI-a strategic AI move

Published on: June 13, 2025 / Updated on: June 13, 2025 – Author: Konrad Wolfenstein

Training datasets: Meta invests $14.3 billion in Scale AI – A strategic AI move – Image: Xpert.Digital

Meta completes $14.3 billion investment in Scale AI - second largest acquisition in the company's history

The investment has already been finalized

Contrary to initial reports of ongoing negotiations, Meta Platforms has already completed its investment in Scale AI. The investment amounts to $14.3 billion and values the startup at over $29 billion. This makes it Meta's second-largest acquisition ever, after the $19 billion purchase of WhatsApp.

Alexandr Wang moves to Meta

A key aspect of the deal is the move of Scale AI founder Alexandr Wang to Meta. The 28-year-old CEO will lead a new “superintelligence” unit at Meta, aiming to achieve artificial general intelligence (AGI). Jason Droege, previously Chief Strategy Officer of Scale AI, will take over as interim CEO of the company.

Wang, who left MIT after just one year and founded Scale AI in 2016 at the age of 19, briefly became the world's youngest self-made billionaire in 2021. His technical expertise and business acumen have made Scale AI one of the leading providers of AI training data.

Scale AI: The backbone of the AI industry

Since its founding in 2016, Scale AI has become an indispensable service provider for the AI industry. The company specializes in data labeling and offers high-quality, annotated datasets for training machine learning models. Its clients include leading AI companies such as OpenAI, Google, Microsoft, and Meta itself.

Business development and financial figures

Scale AI is experiencing impressive growth: The company generated approximately $870 million in revenue in 2024 and reached an annualized revenue of $1.5 billion by the end of the year. For 2025, Scale AI plans to more than double its revenue to $2 billion.

Strategic importance for Meta

Zuckerberg's frustration with Meta's AI position

Mark Zuckerberg is increasingly frustrated with Meta's position in the AI race. Competitors like OpenAI appear to be ahead in both foundational AI models and consumer applications. The weak adoption of Meta's Llama 4 AI models in April exacerbated Zuckerberg's dissatisfaction.

Access to competitor information

The acquisition potentially gives Meta insights into its competitors' data priorities, as many AI labs outsource data services to Scale AI. This could give Meta a strategic advantage in the intense AI competition.

The market for AI training data is booming

Market size and growth forecasts

The global market for AI training datasets was valued at US$2.92 billion in 2024 and is projected to grow to US$17.04 billion by 2032, representing a compound annual growth rate (CAGR) of 24.9%. North America will dominate the market with a 47.95% share in 2024.

The broader market for data collection and labeling is set to grow even more strongly: from US$3.0 billion in 2023, it is expected to rise to US$29.2 billion by 2032, representing a CAGR of 28.54%.

competitive landscape

Scale AI competes with companies like Labelbox, Appen, Amazon Web Services (SageMaker Ground Truth), Google Cloud AutoML, and Microsoft Azure Machine Learning. Despite the competition, Scale AI has secured a leading position through its scalability, accuracy, and innovation.

Regulatory considerations

Meta may structure the deal as a 49% stake rather than a full takeover to avoid regulatory scrutiny. The company is already embroiled in legal disputes with the Federal Trade Commission over antitrust issues related to previous acquisitions.

A turning point for Meta

With this strategic investment, Meta signals its serious intention to catch up in the AI race. The combination of financial investment and the recruitment of Alexandr Wang demonstrates Meta's willingness to dedicate significant resources to its AI ambitions. Whether this strategy will be successful remains to be seen, but it underscores the crucial importance of high-quality training data for the future of artificial intelligence.

Suitable for: