Meta’s AI push faces internal turbulence as Wang questions leadership style
Reports say Alexandr Wang believes Mark Zuckerberg’s micromanagement is stifling progress as Meta bets hundreds of billions on AI, amid leadership churn and a rocky rollout of Llama 4.
Meta’s internal struggle over its AI strategy intensified this year after reports that Alexandr Wang, the Scale AI founder hired to help steer the company’s AI reboot, privately views CEO Mark Zuckerberg’s hands-on approach as suffocating. Wang, 28, joined Meta earlier this year after the company spent more than $14 billion to acquire a 49% stake in Scale AI, and he has been positioned as the public face of Zuckerberg’s AI ambitions. The Financial Times, citing multiple people familiar with the matter, reported that Wang has expressed frustration with what insiders described as a tight control over the company’s AI push.
The disclosures come as Meta grapples with broader upheaval tied to its AI agenda, including rapid hiring, substantial spending, and high-profile departures that have unsettled investors and employees alike. Wang heads Meta’s secretive TBD Lab, which is tasked with developing a flagship model codenamed “Avocado.” Yet some current and former Meta staff have questioned whether Wang’s background—focused on AI data services rather than frontier model development—matches the scale of the company’s ambitions. The FT described internal concerns that tools and products had grown fragmented because multiple teams pursued their own roadmaps without a cohesive integration plan.
Internal tensions at Meta extend beyond Wang. The company has pressed forward with aggressive growth in AI leadership, including a hiring blitz across Silicon Valley and compensation packages reportedly reaching up to $100 million in some cases. One insider described a push to stay ahead of competitors, contributing to a sense of urgency that colleagues say may have overshadowed careful product integration. In this climate, Nat Friedman, the former GitHub CEO brought in to help knit Meta’s AI models into its product stack, faced mounting pressure to deliver quickly, with some staffers saying the pace risked rushing features to market. A prominent example cited by the FT was the accelerated rollout of “Vibes,” Meta’s AI-generated video feed, which insiders said was pushed out at breakneck speed to outpace rivals such as OpenAI’s Sora.
The rapid tempo and high spending have fed broader questions about Meta’s ability to absorb and lead in AI at scale. The company’s AI capital expenditures are expected to hit at least $70 billion this year, and Zuckerberg has signaled costs could surpass $100 billion annually. That level of investment has unsettled investors, especially as Meta’s stock price has faced downward pressure amid concerns over free cash flow and the long arc of the company’s AI push. Critics have pointed to the Llama 4 rollout as a concrete setback: despite ambitious ambitions, the model lagged rival offerings on benchmarks related to coding and complex reasoning. Meta also faced scrutiny over whether it attempted to “game” AI leaderboards by submitting a customized variant for ranking purposes.
As leadership strains mounted, several long-tenured executives exited or were reassigned. Chief Legal Officer Jennifer Newstead was recruited away by Apple, while John Hegeman, Meta’s chief revenue officer, announced his departure to launch a startup. Yann LeCun, the Turing Award-winning chief AI scientist, is leaving to pursue a new AI initiative after reportedly clashing with reporting structures and seeing his research priorities curtailed. Clara Shih, who was recruited from Salesforce to lead business AI, departed within a year. Even with these departures, Meta continued to trim its AI workforce, cutting about 600 roles within its AI teams as part of a restructuring aimed at accelerating decision-making.
The human capital shifts come amid a broader convergence of executives and researchers around Meta’s AI strategy. Nat Friedman’s role was to bridge a gap between rapid productization and more deliberate, foundational AI research, but insiders say the tension over pacing and priorities persisted. The company has defended its approach, arguing that it routinely experiments with different AI model variants and that public leaderboards can be misleading and easily gamed. Meta also asserted that recent chatbot policy concerns cited by external observers were based on examples that were erroneous or inconsistent with its rules.
Meta’s internal upheaval is unfolding against a backdrop of intense scrutiny from investors and competitors as AI remains a central battleground in the tech industry. The company has long portrayed its AI push as a core leverage for growth and differentiating products across its suite of services. Yet the combination of high-profile departures, questions about leadership alignment, and a series of high-cost bets has increased scrutiny of Meta’s governance and risk tolerance.
Despite the turmoil, Meta argues that experimentation and large-scale investment are necessary to remain competitive in a rapidly evolving field where rivals are also racing to deploy new capabilities at scale. The company has emphasized that its AI program remains in a phase of rapid iteration and expansion, with Avocado representing a longer-term, flagship objective rather than a single product release. In the near term, executives say, the company intends to continue expanding its AI infrastructure and talent pool while balancing product integration with foundational research.
As Meta navigates these challenges, observers will be watching closely how leadership restructures its AI organization, how Wang and his team align with product teams, and whether the company can maintain momentum after a year of significant upheaval. The outcome will likely influence Meta’s ability to translate ambitious AI visions into widely adopted technologies and services, as well as its appeal to investors wary of future spending and growth trajectories in the sector.