The new AI superstars at Meta are grumbling about the company as a whole.

Inside Meta’s Growing Tensions Over Its Push for AI Superintelligence

When Meta CEO Mark Zuckerberg decided to overhaul the company’s AI strategy earlier this year, he brought in Alexandr Wang, a 28-year-old AI entrepreneur, to lead a new elite research group called TBD Lab. The team was placed in a special, isolated area next to Zuckerberg’s office to work independently from Meta’s usual corporate structure.

But after just five months, internal tensions have emerged between Wang’s team and Meta’s longtime leaders like Chris Cox (Chief Product Officer) and Andrew Bosworth (CTO).

Team Clash: AI Superintelligence vs. Social Media Business

Cox and Bosworth want the new AI models to be trained using Facebook and Instagram data so they can improve feeds, ads, and Meta’s core business.

But Wang disagrees. He argues the top priority should be catching up with OpenAI and Google by building a powerful “frontier” AI model — not focusing on social media products.

Members of TBD Lab believe many Meta executives care mainly about boosting the social media business, while they want to build superintelligent AI — something far more advanced.

Budgets, Power Struggles, and Departures

• Bosworth was reportedly asked to cut $2 billion from Reality Labs’ budget so more money could go to Wang’s team.

• Some teams complained that new data-centre computing power should support ranking algorithms (which make money), not AI models (which currently don’t).

• Meta denies any conflict or budget shifts.

Meanwhile, Meta’s restructuring led to:

• Dozens of senior AI researchers leaving for rivals

• Top executives including Yann LeCun, John Hegeman, and Clara Shih departing

• 600 layoffs in AI divisions (but none in TBD Lab)

Zuckerberg’s Big Gamble

Zuckerberg has invested huge sums into superintelligence:

• $14.3 billion in Wang’s startup

• A pledged $600 billion to build massive AI data centers

• High-priced recruitment of top researchers from OpenAI and Google

• Even personally delivering homemade soup to potential hires

TBD Lab’s task is to create a superintelligent AI model, while Meta gives away its models as open source and aims to make money by adding AI to Facebook, Instagram, WhatsApp, and smart glasses.

Product vs. Research: A Philosophical Divide

Cox and Bosworth want AI that improves Meta’s products — feeds, ads, recommendations.

Wang wants AI focused on pure research and superintelligence, without being tied down to business goals.

He has told colleagues that Cox only wants AI for product improvements, not breakthrough innovation.

Meta’s CFO recently said next year’s priority is improving social media algorithms using AI — signaling the debate is ongoing.

Current Status

Despite tension, TBD Lab appears stable:

• Only 2 researchers left after their first stock vesting deadline

• The team now has around 100 researchers

Wang is set to host his annual AI holiday party with top Silicon Valley names, though it’s unclear whether senior Meta executives were invited.