What To Know
- In early 2025, in an effort to quickly compete with OpenAI, Meta rushed to copy DeepSeek, a Chinese AI lab that had stunned the community by open-sourcing an advanced mixture-of-experts (MoE) model called R1.
- While the expenses involved in this AI overhaul are massive—estimated total spending for 2025 is between $113 billion and $118 billion—analysts at Cantor say it may still be worth the price if Meta can claw back lost ground.
AI News: Wall Street Braces as Meta Bets Big on Artificial Intelligence
Meta’s latest quarterly earnings announcement is drawing immense attention—not just for its financials, but for the massive strategic upheaval underway behind the scenes. The company, once laser-focused on building a metaverse empire, has now redirected its energy and billions in cash towards conquering the AI frontier. What spurred the sudden change? An ill-fated decision to imitate a rising Chinese AI player has backfired, forcing CEO Mark Zuckerberg to overhaul the company’s entire AI trajectory. In recent months, Meta has responded with an unprecedented AI hiring spree, sweeping up leading talent from rival tech giants and reorganizing itself into a more competitive force.
Meta CEO Mark Zuckerberg bets big on AI with a bold strategic overhaul following Llama 4’s disappointing debut
Image Credit: Mark Zuckerberg
This AI News report explores the whirlwind developments that are reshaping the direction of one of the world’s most influential tech firms.
From Metaverse to Machine Learning
Meta’s change in direction has been rapid. For years, the company faced skepticism from investors over its obsessive investment into the metaverse—a concept still struggling to find a consumer foothold. But this time around, the market seems more forgiving. AI, unlike the metaverse, has immediate and visible utility across industries. That makes the pivot far more palatable to investors who watched Meta burn billions in virtual real estate.
But this was not a calculated transition. In early 2025, in an effort to quickly compete with OpenAI, Meta rushed to copy DeepSeek, a Chinese AI lab that had stunned the community by open-sourcing an advanced mixture-of-experts (MoE) model called R1. Meta’s answer was Llama 4, an upgrade to its Llama AI model family.
Instead of being a triumphant follow-up to the popular Llama 3, however, Llama 4 disappointed. Developers found it harder to integrate and customize. Criticism poured in from developers and insiders alike, some alleging Meta gamed benchmark tests to boost the model’s reputation. Zuckerberg reportedly lost confidence in the existing AI leadership and decided to burn the blueprint.
Enter Meta Superintelligence Labs
In response to the Llama 4 flop, Meta shifted to offense. In June, the company invested a staggering $14.3 billion into Scale AI. That move brought Scale AI CEO Alexandr Wang into the Meta fold as its new Chief AI Officer. He now leads a newly formed entity—Meta Superintelligence Labs—alongside former GitHub CEO Nat Friedman and AI entrepreneur Daniel Gross.
This elite team was further bolstered by hires from OpenAI, Google, and Apple. Most notably, Shengjia Zhao, co-creator of ChatGPT, was tapped as chief scientist of the new lab.
While the expenses involved in this AI overhaul are massive—estimated total spending for 2025 is between $113 billion and $118 billion—analysts at Cantor say it may still be worth the price if Meta can claw back lost ground. However, they note the investment is likely pushing Meta’s budget above its initial forecasts.
Why Llama 4 Failed to Impress
Llama 3 was a favorite among developers. Its dense architecture made it easy to work with and fine-tune. But in its haste to embrace the MoE architecture inspired by DeepSeek, Meta released two smaller versions of Llama 4, promising a much larger “Behemoth” version to follow.
Unfortunately, the MoE-based Llama 4 didn’t deliver the dramatic improvements that developers expected. It wasn’t necessarily more powerful than other open-source models out of China and failed to meaningfully surpass Llama 3. Internal friction also played a role, with some teams advocating to stick with the simpler dense model structure. Ultimately, the lure of overtaking OpenAI pushed the decision toward a new architecture that didn’t resonate with its core user base.
Now, questions are swirling within Meta’s AI ranks. Executives are reportedly considering ditching plans to release the “Behemoth” version altogether and instead focusing on building a powerful, proprietary model that’s no longer fully open source.
Open Source Strategy in Doubt
Historically, Meta has maintained a strong commitment to open-source AI tools. Its PyTorch platform and earlier Llama models were widely embraced by the community. But with mounting internal doubts and external pressure, the company may rethink that stance.
While a Meta spokesperson insists that “our position on open source AI is unchanged,” insiders reveal that a closed-model future is actively being discussed. The New York Times was first to report on these internal deliberations.
Meta’s Core Business Remains Resilient
Despite AI setbacks, Meta’s core business—advertising—remains strong. Analysts estimate the company’s Q2 revenue grew 15%, although that’s down from 22% in the same period last year. It’s the slowest pace of growth since early 2023, and forecasts suggest more modest numbers ahead.
However, Zuckerberg isn’t retreating. In a July post, he boldly declared that Meta would invest “hundreds of billions of dollars” to scale up its AI infrastructure. He claimed Meta Superintelligence Labs would boast “industry-leading compute” and “the greatest compute per researcher.”
Those ambitions have not gone unnoticed. Bank of America analysts described Zuckerberg’s statement as a sign of both confidence and aggressive intent. They also noted that the announcement served to attract elite AI researchers, signaling Meta’s desire to become an AI powerhouse.
AI Talent Wars Heating Up
The AI arms race is becoming increasingly fierce. Meta’s massive recruitment drive mirrors the 2017 self-driving car craze, where companies like Uber and Google threw absurd compensation packages at engineers. That playbook is now back in style—with AI researchers commanding premium salaries and signing bonuses.
Market analysts compared today’s dynamics to a winner-takes-all battlefield. Meta’s actions seem to confirm that view. The company is determined not to fall behind OpenAI, Google, or Anthropic, all of whom are also ramping up their AI investments and recruiting efforts.
Even some of Meta’s critics concede that its current strategy feels more grounded than the speculative metaverse era. Uday Cheruvu of Harding Loevner, which holds Meta shares, said the mood around AI is more euphoric—and more universal—than anything seen in tech for years.
Where It All Leads
Meta is at a defining crossroads. Its failed Llama 4 rollout proved that even tech giants can falter when trying to emulate rather than innovate. But the company’s rapid reaction—building a new AI empire from scratch, hiring some of the best minds in the business, and committing billions in funding—could turn the tide.
The stakes are high. Meta isn’t just trying to compete; it wants to lead the next generation of AI platforms. Whether its latest hires and strategic shift will deliver on that ambition remains to be seen. But one thing is clear: Zuckerberg is all-in, and Meta’s future now hinges on the success of its AI gambit.
If Meta manages to convert its spending spree into cutting-edge products that reshape the AI landscape, it may reemerge as the most influential tech innovator of the decade. If not, it risks becoming a cautionary tale of a company that kept pivoting but never landed the big win.
For the latest developments in the AI Industry, keep on logging to Thailand AI News.
