According to CNBC, Meta’s AI strategy is causing internal confusion as it shifts away from its open-source Llama models. CEO Mark Zuckerberg, who touted Llama as future industry leaders last year, barely mentioned it in the October earnings call. Instead, the company spent $14.3 billion in June 2024 to hire Scale AI founder Alexandr Wang and his team, raising its 2025 capital expenditure forecast to between $70 billion and $72 billion. Internally, many expected a new frontier model codenamed “Avocado” to launch by the end of 2025, but its release is now reportedly pushed to the first quarter of 2026 due to performance testing. A Meta spokesperson, however, stated that model training is on plan with “no meaningful timing changes.”
Meta’s AI Strategy Whiplash
Here’s the thing: this looks like a classic case of strategic whiplash. Last year, Meta was all-in on the open-source evangelism, arguing it was the path forward for AI. It was a clever angle, positioning them as the altruistic underdog against the closed “walled gardens” of OpenAI and Google. But that narrative has clearly cooled. The silence on Llama in recent earnings is deafening. It seems the reality of being in third or fourth place in a brutally fast race has sunk in. So the playbook has flipped: if you can’t beat ’em with your homegrown talent, buy the best in the world. The $14.3 billion Scale AI deal isn’t just a hire; it’s a massive, desperate bet.
The Avocado Delay and Wall Street Worries
And then there’s “Avocado.” A delay from late 2025 to Q1 2026 might not sound huge, but in AI time, that’s an eternity. It feeds directly into the perception, noted by analysts, that Meta is falling further behind. When your rivals are shipping iterative model improvements every few months, a slipped timeline screams internal struggles. Is the model not performing? Is the new, expensive talent trying to re-architect everything? The company line is that everything’s fine, but Wall Street isn’t buying the calm act. The stock’s underperformance tells the real story. Investors poured money into Meta’s “Year of Efficiency,” but now they’re staring down up to $72 billion in capital expenditures with little to show for it in the AI consumer market. Where’s the ROI?
The Brutal Competitive Landscape
So what’s the fallout? The winners here are clearly OpenAI, Google, and maybe Anthropic. Their models are the benchmarks, and they’re cementing enterprise and developer relationships while Meta figures out its next move. Meta’s pivot essentially validates their closed, product-focused approach. The loser, for now, is Meta’s credibility as an AI leader. They’re spending Alphabet-level money but without Alphabet’s (or even Microsoft’s) clear product pipeline or integration. They’re trying to build frontier models, support open source, and integrate AI into all their apps all at once. It’s scattershot. And in a field where focus is everything, that’s a dangerous way to operate. Can “Avocado” really catch up to what GPT-5 or Gemini Ultra will be by 2026? It’s a tall, tall order.
