Meta’s Billion-Dollar Bet on AI Talent: An Empty Treasury of Innovation
Mark Zuckerberg aimed to position Meta as a leader in artificial intelligence, akin to Florentino Pérez’s strategy in football. Last summer, the company made headlines for securing top-tier talent, notably AI prodigy Alexandr Wang, who took charge of the “Superintelligence” division. Yet, as time passes, Meta has little to show for its vast investments—a concerning reality in a fast-paced tech landscape.
Delays in Delivery
Despite pouring billions into an ambitious AI restructuring, the company struggles to meet its scheduled deadlines. Sources within Meta suggest that the organization is facing significant obstacles in rolling out its projects. As generative AI continues to evolve rapidly, internal pressure is mounting, and employees are increasingly on edge regarding the missed milestones.
The Dreaded Avocado Model
Internally dubbed “Avocado,” Meta’s newest foundational AI model has been in development for several months, but initial results indicate it is underperforming. Although it beats the previous Llama 4 and the outdated Gemini 2.5 in internal tests, it fails to match the capabilities of Gemini 3.0 and the relatively newer Gemini 3.1. This shortfall raises questions about the efficacy of Meta’s strategic investments.
Strategic Delays
Rather than rush an inferior product to market, Meta has opted to delay Avocado’s release, pushing it back to a possible launch in May at the earliest. The decision reflects the company’s awareness of the competitive landscape, where releasing a subpar model could damage its reputation and market positioning.
Considering Alternatives
The urgency of the situation is underscored by reports that Meta’s leadership may consider licensing Google’s Gemini for use in its products—an option that reflects the inadequacy of their own model to support major applications like WhatsApp, Instagram, and Threads. Such a move would be unprecedented and indicative of a deeper struggle within the company.
Investment vs. Results
Meta’s fiscal commitment to AI innovation is staggering, with intentions to invest $600 billion in building data centers alone. With a projected capital expenditure of $135 billion focused on AI, nearly double last year’s figure, the gap between spending and outcome raises critical questions. Despite these financial commitments, Meta remains on the sidelines as its rivals advance rapidly in AI development.
Internal Tensions Brewing
The atmosphere within Meta is increasingly fraught, with reports of internal strife. The “TBD Lab,” under Wang’s leadership, faces immense pressure to produce viable AI models. Tensions have escalated between Wang’s team and seasoned executives like Chris Cox and Andrew Bossworth, who prioritize integration with Meta’s core advertising business—an area where Wang appears less proficient.
A Shift Toward Closed Models
In the early stages of the AI race, Meta was recognized for its open models, especially with Llama becoming a standard. However, recent developments suggest a pivot towards closed models, similar to those of competitors like OpenAI and Google. This strategic shift may limit collaboration, but it offers Meta greater control over its technology.
The Fruits of Investment
Despite a monumental infusion of resources, Meta’s achievements to date are underwhelming. The only notable release from its AI investments is “Vibes,” a Sora-like app that failed to gain traction. Additionally, the initial wave of talent acquisitions is reversing, with many AI researchers opting to leave for competitive opportunities or to embark on their ventures.
In conclusion, Meta’s ambitious foray into AI has met roadblocks that threaten to undermine its substantial investments. With delays, internal divisions, and a lack of competitive offerings becoming evident, the company’s future in the sector hangs in a precarious balance.