Meta's New AI Unit Is in Chaos: What's Really Going Wrong Inside the Company

Meta's AI Ambitions Are Hitting a Wall From the Inside

Meta has made no secret of its intention to become a dominant force in the artificial intelligence race. With billions of dollars poured into research, infrastructure, and talent acquisition, the company has positioned itself alongside OpenAI, Google, and Microsoft as a key player in the global AI boom. But behind the polished product announcements and Mark Zuckerberg's confident public statements, something appears to be going very wrong — and it starts from within.

According to sources familiar with the matter and internal discussions reviewed by WIRED, Meta's newly formed AI unit is mired in dysfunction. Executives are confused about strategic priorities. Employees feel directionless. And the cultural tension between the old-guard Meta and the new AI-focused identity the company is trying to forge has created friction that is proving difficult to manage.

A Company Caught Between Two Identities

To understand what is happening inside Meta's AI operations, it helps to understand the company's broader transformation story. Meta spent the better part of two years and tens of billions of dollars pivoting toward the metaverse — a bet that largely failed to capture either consumer enthusiasm or investor confidence. Now, with generative AI reshaping the technology landscape, the company has made another sharp turn, redirecting resources and attention toward AI products and research at a breakneck pace.

The problem with pivoting quickly is that organizational culture rarely moves at the same speed. Meta was built as a social media and advertising company. Its DNA is rooted in growth hacking, data-driven product iteration, and moving fast. The kind of deep, long-horizon research culture that leading AI development requires is fundamentally different — and grafting one onto the other without careful management has proven harder than executives anticipated.

Sources suggest that this culture clash is playing out daily inside the new AI unit, with tensions running high between researchers who want to operate with academic-style freedom and product managers pushing for rapid deployment and commercial results.

Leadership Confusion and Internal Friction

Strong AI divisions require clear leadership, coherent vision, and effective communication between teams. By multiple accounts, Meta's AI unit is struggling on all three fronts. Internal discussions reportedly reflect a level of frustration among employees that goes well beyond typical workplace grumbling. One particularly striking exchange, referenced in reporting by WIRED, captures the tone: a directive to "tell him he's a piece of shit" — a phrase that, while colorful, symbolizes a broader breakdown in professional communication and organizational respect.

Whether such exchanges are isolated incidents or symptomatic of a deeper leadership culture problem is difficult to say from the outside. What is clear is that when language like this surfaces in internal communications at a company the size of Meta, it signals that something structural is not working. Healthy organizations — even fast-moving ones — do not typically communicate this way about colleagues or leaders.

Executives themselves appear uncertain about where the unit is headed. Multiple sources describe a lack of clarity around which AI projects should take priority, how teams should be structured, and what success actually looks like in this new division. That kind of ambiguity at the top tends to cascade downward, leaving employees without the direction they need to do their best work.

The Competitive Pressure Is Not Helping

Part of what makes this internal chaos so consequential is the competitive environment Meta is operating in. The AI industry is moving at a pace unlike anything seen in consumer technology in decades. OpenAI continues to release models that capture global attention. Google has embedded AI capabilities deeply into its search and productivity products. Microsoft has made a multi-billion-dollar bet on AI integration across its enterprise suite. Apple is quietly building its own AI layer into its hardware ecosystem.

In this environment, organizational dysfunction is not just an HR problem — it is a strategic liability. Every month spent navigating internal confusion is a month that competitors are shipping products, attracting talent, and establishing user habits that will be difficult to disrupt later.

Meta does have real assets in this race. Its open-source Llama family of models has earned genuine respect in the AI research community. Its access to vast amounts of social data gives it a potential training advantage in certain domains. And its scale means it can invest in infrastructure that smaller players simply cannot match. But raw resources cannot compensate for strategic incoherence. Without a clear direction and an organizational culture that supports execution, those advantages risk being squandered.

What Needs to Change for Meta AI to Succeed

If Meta is serious about competing at the frontier of artificial intelligence, several things need to change — and quickly.

• Leadership clarity: The AI unit needs executives who can articulate a coherent vision, communicate it consistently, and model the professional culture they want to build. Ambiguity at the top is the fastest way to demoralize talented people who have other options.
• Cultural alignment: Meta needs to decide what kind of AI company it wants to be and build the structures that support that identity. Forcing a research culture and a product culture to coexist without clear boundaries or mutual respect is a recipe for ongoing friction.
• Retention focus: Top AI talent is extraordinarily scarce and highly sought after. If word spreads — as it already has, through reporting like WIRED's — that Meta's AI unit is chaotic and poorly managed, recruiting and retaining the best people will become increasingly difficult.
• Strategic focus: Trying to do everything in AI at once often means doing nothing particularly well. Meta may benefit from narrowing its near-term focus and executing well on a smaller number of high-impact bets.

The Stakes Could Not Be Higher

Meta's struggles with its AI unit are not just an internal management story. They reflect a broader challenge facing every large technology company trying to retrofit itself for an AI-native world. Building an effective AI division inside a massive corporation is genuinely hard — harder, perhaps, than building one from scratch, because the weight of existing culture, processes, and priorities constantly pulls against the new direction.

The companies that figure this out will shape the next decade of technology. Those that don't will find themselves watching from the sidelines as the AI era matures without them. For Meta, with its resources and reach, failure is far from inevitable. But the window for getting the organizational foundations right is not unlimited — and right now, by all accounts, the clock is ticking.