Inside Meta’s AI Struggle: Tensions Rise Between Zuckerberg and Billionaire Hire Alexandr Wang

A new report reveals growing internal tensions over Meta‘s aggressive artificial intelligence strategy. According to the Financial Times, senior leaders are facing friction over management and direction. The central figure in this reported conflict is Scale AI founder Alexandr Wang, who joined Meta earlier this year.Wang’s arrival was part of a massive $14 billion deal where Meta took a major stake in his company. Publicly, he was hailed as a key architect for Meta’s AI future. Privately, however, he has expressed frustration with CEO Mark Zuckerberg’s hands-on style, describing it as restrictive to associates.

Pressure Mounts Amid Costly AI Push and Performance Questions

The reported leadership strain comes during a period of intense pressure and spending at Meta. The company is projected to invest at least $70 billion in AI capital expenditure this year. Zuckerberg has signaled these costs could go even higher, a move that has concerned some investors.Internally, the push has not been without setbacks. The rollout of Meta’s Llama 4 large language model faced challenges. It reportedly lagged behind key competitors in certain benchmarks, like coding and reasoning. Employees cited issues with training data and limited testing time as factors.Some teams have faced pressure to ship AI features rapidly to compete with rivals like OpenAI. This has sometimes come at the expense of internal alignment. The situation is compounded by a stream of senior leadership departures from the company.

Can Meta’s Costly Bet Overcome Organizational Friction?

The massive financial bet is being led by a reorganized team. Wang now heads a confidential internal unit developing a next-generation AI system codenamed “Avocado.” However, some employees have quietly questioned if his data-services background is the right fit for leading frontier research at Meta’s scale.Other senior AI figures are also reportedly reassessing their roles. Veteran scientist Yann LeCun is expected to step away to pursue independent research. This follows disagreements over reporting structures and research priorities within the AI division.Meta has disputed aspects of the Financial Times report. The company stated it routinely tests multiple AI model variants and warned that public benchmarks can be misleading. It also pushed back on suggestions of widespread internal dysfunction.

La Liga Standings: Barcelona Holds Commanding Lead Over Real Madrid at Winter Break

The ultimate success of Meta’s enormous AI investment may hinge on resolving these internal tensions. The company’s ability to streamline leadership and execution could be just as critical as its technology in the global AI race.

Info at your fingertips

Who is Alexandr Wang and why did Meta hire him?

Alexandr Wang is the billionaire founder of Scale AI, a data-labeling company. Meta hired him after investing over $14 billion for a 49% stake in his firm, positioning him as a central leader in its AI strategy reset.

What problems did Meta face with its Llama 4 AI model?

According to the report, the Llama 4 model lagged behind competitors in benchmarks for coding and reasoning. Internal factors included gaps in training data, limited testing time, and organizational silos between teams.

Why are investors concerned about Meta’s AI spending?

Meta’s AI capital expenditure is projected to hit at least $70 billion this year, with costs expected to rise further. This enormous, open-ended spending has unsettled some investors worried about the company’s long-term cash flow and profitability.

What is the “Avocado” project at Meta?

“Avocado” is the internal codename for Meta’s next-generation AI system. The confidential project is being led by Alexandr Wang’s unit and represents the company’s ambitious effort to build frontier AI technology.

How has Meta responded to the report of internal tension?

Meta has disputed parts of the Financial Times report. The company stated it tests many AI model variants and cautioned that public benchmarks can be misleading. It also pushed back on characterizations of widespread dysfunction.