Meta Delays Key AI Model on Performance Issues; Orders Broadcom ASICs, Lacks New Flagship
summarizeSummary
Meta Platforms has delayed its "Avocado" AI model from March to at least May, citing weaker performance in reasoning, coding, and writing compared to rivals. The company also confirmed it has not launched a major flagship AI model since Llama 3 in 2024, with Llama 4 (2025) being limited to smaller models. Concurrently, Meta is ordering custom ASICs from Broadcom for its ranking and recommendation systems, planning large-scale production and installation of multiple gigawatts of these chips. This news provides specific details regarding Meta's AI development progress, following recent announcements of a new Applied AI Engineering organization and the acquisition of AI agent social network Moltbook. The significant Broadcom ASIC order aligns with Meta's previously announced substantial increase in 2026 capital spending plans, targeting $115 billion to $135 billion. The delay of a key AI model due to performance issues and the absence of a new flagship model are significant negative signals for Meta's competitive standing in the rapidly evolving AI landscape. While the Broadcom ASIC order demonstrates a commitment to custom hardware for AI efficiency, the overall picture suggests potential challenges in their AI model development, which is critical for future growth and innovation. Investors will be closely watching for updates on the Avocado model's performance and release, as well as any announcements regarding new flagship AI models. The execution of the Broadcom ASIC deployment and its impact on operational efficiency will also be key.
At the time of this announcement, META was trading at $612.80 on NASDAQ in the Technology sector, with a market capitalization of approximately $1.6T. The 52-week trading range was $479.80 to $796.25. This news item was assessed with negative market sentiment and an importance score of 8 out of 10. Source: Wiseek News.