Wirestock closed a $23 million Series A to scale its creator-sourced training-data platform, the company said in a release dated May 15. The round was led by Nava Ventures with participation from SBVP — the fund co-founded by Sheryl Sandberg — alongside Formula VC and I2BF Global Ventures, bringing total capital raised to about $26 million.
The company says it now sits above $40 million in annual run-rate revenue, with creator payouts growing 20x year over year since it pivoted from a marketplace into a licensed data supplier in 2023.
What the platform actually sells
Wirestock aggregates content from a network of more than 700,000 contributors — photographers, videographers, graphic designers, 3D artists, filmmakers and musicians — and licenses curated datasets to AI labs training image, video and multimodal models. Categories include stock photo, video, design files, 3D models, gaming assets and audio.
The company pitches itself as the rights-cleared alternative to scraped web corpora: creators sign explicit licensing terms, are paid monthly, and contribute structured metadata that labs can use to assemble bespoke training sets.
Why the round lands now
Frontier labs are exhausting their easy pools of high-quality visual data just as multimodal training demand spikes. Video generation, world models and embodied-AI training pipelines need orders of magnitude more annotated frames, depth maps and 3D scenes than text-only LLMs ever did — and most of that material is still locked inside stock libraries, studios and individual creator hard drives.
Licensing pressure is also escalating. Disney and Universal sued Midjourney in June 2025 over alleged mass IP infringement in training data, and Getty's long-running case against Stability AI yielded a landmark UK ruling on November 4, 2025 — largely in Stability's favor on copyright, with Getty's secondary copyright claim rejected outright after it had already abandoned its primary copyright and database-right claims before closing submissions; Stability lost only on limited trademark findings. The ruling nonetheless underscored how exposed unlicensed pipelines have become. Reports of frontier labs being pressured to retroactively scrub training datasets after rights complaints continue to circulate. A clean upstream supplier with documented chain-of-custody is starting to look less like a nice-to-have and more like procurement table stakes.
Where the money goes
Wirestock says the new capital will fund custom-dataset capacity for labs that want bespoke collections, deeper integrations with research partners, and expanded support for emerging modalities — specifically calling out motion design, 3D modeling and music as next-priority verticals. The company will also invest in creator tools, payout infrastructure and earning pathways.
For AI builders, the read-through is simple: the supply side of multimodal training data is becoming a venture-funded category, not a side hustle. Labs that depended on murky scraping at GPT-3 scale will need real procurement relationships at GPT-6 scale, and the vendor list is being established right now. Expect more $20M–$50M Series A rounds in this lane through 2026.
