Meta Faces Internal Friction and Skepticism Over AI Strategy

Meta’s internal AI agent development is lagging behind executive expectations, even as the company pivots toward selling excess compute capacity. CEO Mark Zuckerberg admits the recent restructuring has not yet yielded the promised results.

AI & Culture Desk
Updated Nov 12, 2026 2 min read
Meta Faces Internal Friction and Skepticism Over AI Strategy

Meta is navigating a period of internal friction as its ambitious AI roadmap hits operational speed bumps. According to a report from TechCrunch, CEO Mark Zuckerberg told staff at a town hall on Thursday that the development of AI agents has not "accelerated in the way" executives previously expected. This admission follows a year of significant workforce upheaval, including the layoff of approximately 8,000 employees and the reassignment of 7,000 others into specialized AI groups like Agent Transformation. Zuckerberg reportedly conceded that these personnel changes were not as "clean" as intended, driven by leadership's anxiety over the company's ability to adapt quickly enough to the shifting tech landscape.

The transition has been difficult for the rank and file. While the company prepares to spend as much as $145 billion on AI infrastructure this year, the internal culture appears strained. According to TechCrunch, some engineers assigned to the new AI unit have described the environment as a "soul-crushing gulag." Zuckerberg noted that the perceived benefits of the new AI-focused structure have not "come to fruition yet," though he expressed optimism that the company would begin to see improvements from its heavy investments within the next three to six months.

Parallel to these internal struggles, Meta is exploring a new business line: selling excess AI computing capacity. This move, which recently caused the company's stock to jump 7%, aims to position Meta alongside established cloud providers like AWS, Microsoft Azure, and Google Cloud. However, the strategy has drawn skepticism from industry observers. Mark Douglas, CEO of the ad platform MNTN, characterized the plan as a tactical move rather than a major strategic pivot, suggesting it functions primarily to get attention rather than address core business needs.

The viability of building out massive compute infrastructure in the United States is also under scrutiny. Douglas argued that U.S. data center capacity may prove to be a poor long-term investment compared to alternatives in the Gulf, where sovereign-wealth-backed projects utilize cheaper energy sources. He noted that while Meta builds in the U.S., the Kingdom of Saudi Arabia is bringing online massive capacity at "significantly lower prices." Douglas questioned the logic of Meta, a company with billions of social media users, pivoting to a model that serves a small number of massive compute buyers.

Despite the skepticism surrounding the cloud infrastructure detour, Douglas remains bullish on Meta’s core AI strategy, specifically its open-source Llama models. He believes Meta’s ability to train models on vast amounts of data will ultimately benefit its advertising business, even as investors grow anxious about the immediate returns on AI spending. For now, the company remains caught between the high-stakes pressure to deliver on AI agents and the logistical challenges of managing a massive, capital-intensive infrastructure build-out that may not align with its traditional strengths.

Sources

  1. 01 Mark Zuckerberg tells staff that AI agents haven’t progressed as quickly as he’d hoped — TechCrunch — Insider
  2. 02 Meta’s cloud compute dreams: Why opt for U.S. AI data centers when Saudi Arabia has cheap oil and cheaper power? — Fortune