Wiresock leverages AI to meet data needs for learning, successfully attracting $23 million in investment

Artificial intelligence learning-data startup Wirestock has successfully attracted $23 million (approximately 345.3 billion KRW) in new investment. As demand keeps rising for “world models” that go beyond generative AI and for computer usage models, image- and video-based data supply companies are seeing a more pronounced trend of capital inflows.

According to TechCrunch, this round was led by early-stage investment firm Nava Ventures, with participation from SBVP, Formula VC, and I2BF Ventures. Wirestock’s total external investment has increased to $26 million (approximately 390.3 billion KRW).

Business Shift and Data Supply Expansion

Founded in 2018, Wirestock originally offered a service to help photographers and creators distribute content on stock photo platforms. However, starting this year, the company has shifted its focus to providing training data for AI development companies. It is reported that the company currently has around 700,000 users who create a catalog of more than 50 million images and videos.

Wirestock not only sells general datasets, but also sells customized training data. Its standout advantage lies in producing videos that show users how to operate applications. For example, a startup developing an AI-based web design tool can commission the creation of data that includes scenes of editable interface elements within Figma.

Performance Growth and World Model Demand

This expansion in its business also appears to have contributed to performance growth. TechCrunch reports that prior to this funding round, Wirestock’s annualized sales had already exceeded $40 million (approximately 600.4 billion KRW). The backdrop is a surge of capital pouring into the world model space over the past year. A world model is an AI that learns visual data to understand and simulate real-world environments, and its application potential in robot training and automation has attracted significant attention.

In practice, related funding is being concentrated among a small number of promising companies. World Labs, led by Fei-Fei Li, raised $1 billion in February from a consortium that includes Nvidia ($NVDA). Meanwhile, competitor AMI Labs also completed an investment totaling $1.03 billion, and has begun developing world models for hardware engineering automation.

Computer Usage Models and Future Expansion

The visual data provided by Wirestock can also be used to train computer usage models. Standard Intelligence, which received $75 million in funding last month, used videos of users performing application operations to build the “FDM-1” algorithm. The company says this AI can automate multi-step tasks using fewer hardware resources than existing models.

Wirestock plans to use this investment to expand its technical staff, such as AI researchers and engineers. New hires will work on developing software products that enable enterprise-grade AI labs to jointly build and collaborate on datasets. The company is also considering expanding its business scope beyond images and videos to other types of training data in the future.

AI Data Market Trends

This funding comes after large-scale fundraising by other AI data providers, which also makes it noteworthy. Afterquery, headquartered in San Francisco, recently raised $30 million in investment. The company sells curated datasets that include AI prompts, user responses, and the step-by-step thought process used to arrive at those responses. This kind of data helps improve an AI’s “reasoning ability.”

Ultimately, market attention seems to be shifting from simply bigger models to more fine-grained training data. Wirestock’s fundraising once again shows that the core of AI competitiveness lies not only in semiconductors and model performance, but also in acquiring “high-quality data” that reflects real-world application scenarios.

TP AI Note and Disclaimer: This article was summarized using a language model based on TokenPost.ai. The main content may be omitted or may not be consistent with facts.