Meta Platforms is pivoting away from its open-source artificial intelligence (AI) strategy toward developing a proprietary model that it can monetize, Bloomberg reported Wednesday. The move, which includes using Alibaba's Qwen among other third-party models to train Meta's new system codenamed Avocado, sent the companies' shares in opposite directions.
AI Generated Image
Meta shares fell 1% Wednesday while American depositary receipts (ADRs) climbed 1.8%, outpacing the S&P 500's 0.7% gain. The shift marks Meta's biggest departure from the open-source strategy it has championed for years and comes after its Llama 4 model disappointed Silicon Valley earlier this year.
The new approach aligns Meta with rivals Google and OpenAI, who already operate closed models that generate revenue. Avocado is expected to debut next spring, according to Bloomberg, quoting people familiar with the matter.
The strategic shift follows intense pressure on CEO Mark Zuckerberg to show returns on Meta's massive AI spending. The company has pledged $600 billion for U.S. infrastructure over three years, mostly AI-related, while raising its 2025 capital expenditure (Capex) guidance to between $70 billion and $72 billion.
Llama 4 Disappointment Triggers Leadership Overhaul
The strategy change came after Llama 4's underwhelming reception earlier this year, Bloomberg reported. Zuckerberg sidelined some team members who worked on that project and launched an aggressive recruiting campaign, offering top AI researchers multiyear packages worth hundreds of millions of dollars.
The company's new Chief AI Officer, Alexandr Wang, who joined through a $14.3 billion investment deal with his startup Scale AI, advocates for closed models. Zuckerberg now spends much of his time working closely with new hires in a group called TBD Lab, which is developing Avocado.
The TBD team is using several third-party models including Google's Gemma, OpenAI's gpt-oss and Alibaba's Qwen in the training process, Bloomberg reported. This represents a notable shift for Zuckerberg, who raised concerns on Joe Rogan's podcast in January about Chinese models potentially being shaped by state censorship.
Internal Confusion and Development Challenges
The pivot has created internal confusion, CNBC reported Tuesday. Many within Meta had expected Avocado's release before year-end, but the model is now slated for first-quarter 2026 debut as it wrestles with training-related performance testing, according to people familiar with the matter.
"Our model training efforts are going according to plan and have had no meaningful timing changes," a Meta spokesperson told CNBC.
The company has deprioritized its open-source messaging. Some Meta employees were directed to stop talking publicly about open-source and Llama products after the Llama 4 launch while leadership recalibrated whether those efforts still made sense, Bloomberg reported.
Yann LeCun, known as one of the AI godfathers, recently left Meta after years leading its long-term AI research group, partly due to frustrations over resources, Bloomberg reported. Before his departure, some employees had been encouraged to keep LeCun, a big proponent of open-source technology, out of the spotlight.
Wall Street Questions Return on Investment
Meta's AI spending has become the focus of Wall Street investors who recently balked at Zuckerberg's pledge to continue investing heavily into 2026. While Meta maintains its AI investments strengthen its advertising business, some worry the expensive bet on data centers and infrastructure may not contribute to profit for years.
"In many ways, Meta has been the opposite of Alphabet, where it entered the year as an AI winner and now faces more questions around investment levels and ROI," KeyBanc Capital Markets analysts wrote in a November note.
The pressure has intensified as competitors advanced. Google's Gemini 3 drew solid reviews last month, while OpenAI announced GPT-5 updates and Anthropic debuted Claude Opus 4.5 in November. Nvidia Corp. CEO Jensen Huang highlighted OpenAI, Anthropic and xAI as major customers during the chipmaker's November earnings call but made no reference to Llama.
Success also depends on Zuckerberg's ability to sell his vision. The stated goal of achieving "superintelligence" has left people confused and concerned. Market research conducted with external consultants found the term triggered fears of AI's unchecked power, particularly in Europe where regulators have already targeted Meta's AI offerings, Bloomberg reported.
Alibaba's contribution to Avocado is meaningful not for potential revenue but because it bolsters Qwen's reputation as a powerful tool for AI developers, strengthening the Chinese company's position in the AI revolution.
Explore more exclusive insights at nextfin.ai.
