Meta Takes a Big Step Into AI With Scale Investment
Meta just dropped $14.3 billion on a 49% stake in Scale AI, pushing the startup’s valuation past $29 billion. That makes it the company’s second-largest deal ever—right behind the $19 billion WhatsApp buy in 2014. Not exactly small change.
Scale’s founder, Alexandr Wang, broke the news to employees in a tweet, calling it a “major milestone” and a nod to their work so far. He’ll stay on the board, but the day-to-day CEO role shifts to Jason Droege, the firm’s Chief Strategy Officer.
For Meta, this isn’t just about throwing money around. They’ve been scrambling to keep up in the AI race, rolling out new models and even a standalone AI app recently. The company says the deal helps “deepen” their work on AI training data—which, let’s be honest, is the backbone of everything in this space.
The Push and Pull of Centralized AI
But here’s where it gets tricky. Big Tech’s AI dominance isn’t sitting well with everyone. Renz Chong, who runs the decentralized platform Sovrun, points out that open-source models are now competing head-to-head with the closed systems from companies like Meta.
“State-of-the-art doesn’t have to mean centralized,” Chong told *Decrypt*. The problem? A lot of projects still rely on off-chain APIs or centralized endpoints, which puts smaller players at a disadvantage. Early decentralized infrastructure, he argues, could change that—but it’s an uphill battle.
Meta’s investment might dodge some regulatory headaches, too. Since they’re not taking full control, it could fly under the radar of antitrust watchdogs. That’s probably a smart move, given the Senate’s recent questions about Meta’s ties to China’s AI efforts.
What Scale AI Actually Does
If you’re not familiar with Scale, here’s the gist: they handle data labeling, the tedious but crucial work of tagging data so AI models can learn from it. Think of it as teaching a machine what a cat looks like by showing it thousands of labeled cat photos. Their clients include heavyweights like Google and OpenAI, and much of the labor happens overseas.
Meanwhile, the decentralized crowd isn’t sitting still. Sovrun teamed up with Virtuals Protocol to launch ReadyGamer, a project that drops AI-driven NPCs into games. Revenue dipped earlier this year, but activity’s picking back up—slowly.
Chong’s take? The real shift isn’t just about better tech. It’s about “who gets to shape it.” Whether that’s enough to counterbalance Meta’s latest power move, though, is still up for debate.
