By Antonio Di Giacomo, Senior Market Analyst at XS.com
Meta Platforms saw its shares fall by nearly 1.2% after news emerged of a significant shift in its artificial intelligence strategy: the development of a new closed and
monetizable model called Avocado, expected to debut in the spring. The announcement also boosted Alibaba’s stock, which rose around 2%, as its Qwen model is reportedly being used in the training process for Meta’s new system. This move marks a significant departure from the open-source strategy previously
championed by Mark Zuckerberg.
The new Avocado model represents a break from the approach that gave rise to the Llama family, whose models were released openly to the community. The company is now seeking to replicate the commercial strategy of other leading players in the sector, developing a closed system that is more controlled and has clear
monetization potential through paid access for corporate clients. This shift aims to maximize recurring revenue and compete more directly in the advanced AI services market.
One of the reasons for this reorientation is the underperformance observed during the development of Llama 4, whose progress was reportedly slower than expected
and yielded insufficient results on complex reasoning tasks. The combination of delays, technical limitations, and internal concerns about the model’s viability
prompted Meta to pursue a more ambitious next-generation system for large-scale commercial applications.
In building Avocado, Meta is not relying solely on its own technology; it has reportedly incorporated external models, including Alibaba’s Qwen. The integration
of third-party systems reflects the growing interdependence among tech giants in AI development, a field where the pace of innovation requires massive computing
resources, strategic alliances, and access to highly specialized architectures.
The strategic shift also coincides with a period of internal tensions at Meta. The new head of AI, Alexandr Wang, is said to have expressed discomfort with Zuckerberg’s micromanagement, as the CEO remains directly involved in key technical decisions.
This dynamic has created friction within the team, especially at a time when the company is centralizing decision-making and reducing the flexibility that characterized its earlier, more open-innovation-driven stages.
The departure of Yann LeCun, a historic figure in Meta’s AI research and a pioneer of the discipline, adds further pressure to the internal landscape. LeCun chose to
leave the company to launch a new independent project focused on advanced intelligence models, symbolizing the end of an era marked by academic research
within Meta and the beginning of a phase much more oriented toward profitable, business-driven products.
Budget cuts in non-priority areas, such as the metaverse, and increased investment in AI-powered products, including wearables and generative assistants, accompany this strategic pivot. Meta seeks to accelerate the time between research and monetization, sending a clear signal to the market that it intends to compete more aggressively in the global AI ecosystem.
In conclusion, the Avocado project embodies a crucial transition for Meta: from a vision built on openness and collaboration to a closed, proprietary, and directly
monetizable model. The use of external technology, internal tensions, and the departure of key figures pose significant challenges but also reflect the urgency of
positioning the company at the forefront of AI. If Meta succeeds in making Avocado a global reference model, the shift will be seen as a strategic success; otherwise, it
could represent a costly detour in an industry where speed and precision determine the winner.