Meta's Pivot to Proprietary AI Model 'Avocado' Amplifies Strategic Uncertainty Amid Heavy Capex and RL Losses

What happened

Meta is reportedly developing a new frontier AI model, codenamed Avocado, which may shift from open-source to proprietary, signaling a strategic realignment to compete with rivals like OpenAI and Google. This follows a $14.3 billion investment to recruit top AI talent, including the founder of Scale AI, as part of Meta's aggressive AI push. However, internal confusion over this shift raises concerns about operational efficiency and strategic clarity during a period of already heightened capital intensity. According to the DeepValue report, Meta's AI and infrastructure capex is ramping to $70-72 billion in 2025, while Reality Labs incurs over $17 billion in annual losses, straining free cash flow despite strong core ad profits. Investors must now evaluate whether this proprietary AI gamble can justify the escalating risks and spending, especially with the stock trading at a premium valuation.

Implication

The shift to a proprietary AI model could enhance Meta's competitive moat by offering differentiated products, but it also heightens dependency on successful execution and market adoption in a crowded field. High capex commitments, already a key concern per the report, may rise further if Avocado demands additional infrastructure, squeezing free cash flow and amplifying capital allocation risks. Internal confusion suggests potential management or cultural inefficiencies, which could delay product launches or lead to wasteful spending amid ongoing Reality Labs losses. This adds another speculative layer to Meta's investment profile, requiring investors to monitor AI strategy updates and capex efficiency closely for signs of payoff. In the near term, any missteps could trigger valuation compression given the stock's premium multiple and modest margin of safety, reinforcing the need for disciplined capital discipline.

Thesis delta

The investment thesis for Meta previously balanced a dominant core ad franchise against high-risk bets in AI and Reality Labs, with open-source models like Llama offering ecosystem benefits. This news signals a shift towards a more proprietary AI strategy, potentially increasing monetization opportunities but also raising execution and competitive risks, which could alter the risk-reward profile. It underscores the urgency for clearer evidence that AI spending translates into durable growth, as continued capital intensity without tangible returns may erode the margin of safety.