Scale AI Deal with Meta Is Already Showing Strain

Memet Deniz Yucekaya

August 31, 2025

Meta’s $14.3 billion partnership with Scale AI was meant to anchor its AI ambitions. Just months later, cracks are already appearing. Executive exits, internal friction, and a pivot toward rival vendors now raise questions about how solid this deal really is.

Early departure raises eyebrows

One major red flag: Ruben Mayer, a top exec from Scale’s team, left Meta after only two months. He said his role was to help launch the lab, but insiders claimed he wasn’t central to its core AI team. Whether he was sidelined or simply moved on, it sent a signal.

Awe Dropping Apple Event to Unveil iPhone 17 Air and Six More New Products

Meta’s researchers go elsewhere

Despite the multibillion-dollar deal, Meta’s core AI team is now sourcing data from Mercor and Surge. According to five people close to the project, researchers are favoring these vendors over Scale due to better-quality datasets and clearer annotation standards.

Legacy methods, modern problems

Scale’s crowdsourcing model once made sense for early AI tasks. Today’s demands are different. As language models become more complex, they need data labeled by domain experts, not temp workers. Surge and Mercor built their businesses around that standard.

Scale has attempted to shift direction with its Outlier platform, but its competitors already had a head start, and it shows.

What’s going wrong behind the scenes

Here are the most visible warning signs surrounding the Meta–Scale relationship:

• Ruben Mayer exited Meta after a short tenure
• Meta researchers prefer competing data sources
• OpenAI and Google ended their ties with the company
• Scale reduced staff by 200 in July
• Several AI leads at Meta have resigned
• Internal teams question the long-term fit of the partnership

Meta’s AI effort needs clarity, not chaos

Mark Zuckerberg hoped this investment would fast-track innovation. But instead, MSL has become harder to manage. New hires from DeepMind and OpenAI are reportedly struggling to navigate Meta’s internal systems, and original team members have left in frustration.

Big plans, uncertain execution

With another flagship AI model on the way by late 2025, Meta doesn’t have room for stalled momentum. Whether this deal will steady or collapse further could shape its AI future.