Meta Platforms Inc. (META) is doing something interesting with its AI strategy: borrowing from the competition. The social media giant has turned to an open-source AI model from Alibaba Group Holding Limited (BABA) to help build its next-generation system, a move that highlights both the collaborative nature of modern AI development and Meta's urgent need to catch up.
According to reports, Meta is using Alibaba's Qwen model, along with open-source systems from Alphabet Inc.'s (GOOGL) Google and OpenAI, to train a new AI model code-named Avocado. The company plans to release Avocado in the spring, though this version will only be available through an Application Programming Interface (API), suggesting a shift toward a more closed, revenue-focused approach rather than Meta's typical open-source philosophy.
The TBD Lab Experiment
Avocado is being built inside Meta's newly formed TBD Lab, which is staffed with researchers the company acquired at considerable expense. Leading the effort is new chief AI officer Alexandr Wang, whose previous company Meta purchased for a staggering $14.3 billion. That's a lot of confidence in one person's vision.
The lab is using what's known as "distilling" to train Avocado, a technique that has stirred controversy in AI circles. Essentially, developers use outputs from existing models to speed up the training process for new ones. It's efficient, but some question whether it's the best path to truly novel AI capabilities. In this case, TBD Lab is distilling Qwen and other open-source models to accelerate Avocado's development.
Metaverse Budget Gets the Axe
While Meta pushes forward on AI, it's pulling back elsewhere. The company is raising prices on its virtual reality devices and slowing down new hardware releases, focusing instead on premium products. More significantly, Meta is reportedly cutting up to 30% of its 2026 metaverse budget as it reallocates resources from Reality Labs to AI initiatives.
The company is tightening budgets across its Reality Labs division, delaying mixed-reality hardware projects, and reshaping leadership to prioritize high-return AI efforts. It's a significant strategic shift aimed at unlocking long-term value, even if it means scaling back Mark Zuckerberg's metaverse dreams.
Wall Street Approves
Analysts are generally supportive of Meta's newfound cost discipline. JPMorgan's Doug Anmuth and Bank of America Securities' Justin Post both view the metaverse cuts as a smart move that could unlock billions in savings and strengthen earnings per share growth. The reduction in Reality Labs losses creates breathing room for Meta's massive capital expenditure cycle focused on AI infrastructure and data centers.
Meta stock has gained over 11% year-to-date as investors digest the company's evolving strategy. At publication time on Thursday, shares were trading up 0.20% at $651.41, reflecting cautious optimism about the company's pivot toward AI and away from its more speculative metaverse bets.
The big question now is whether Avocado can deliver the kind of breakthrough Meta needs to compete with the likes of OpenAI and Google in the increasingly crowded AI landscape. Borrowing from Alibaba's playbook might be a pragmatic shortcut, but the proof will be in whether Meta can turn those open-source building blocks into something genuinely competitive.