Zuckerberg’s AI push has fueled rising costs but lags in short-term returns

Meta Platforms’ hefty AI spending comes into focus as CEO Mark Zuckerberg’s huge ambitions to win in the AI arms race are seemingly costing the social media giant more than it is delivering.

The Facebook parent company’s second-quarter earnings are around the corner, and the numbers are not exactly throwing a party. While Wall Street expects profit to rise by about 11.5% to $15 billion, it is still disappointing, as that is the slowest growth the company has seen in two years. And even with revenue going up (14.7% to roughly $44.8 billion), that is also still Meta’s weakest pace since early 2023.

Will Meta benefit from the hiring sprees and smart glasses dreams?

Meanwhile, expenses keep climbing. Nearly a 9% jump in operating costs has investors watching closely, wondering if Zuckerberg’s obsession with “superintelligence” is brilliance or bloat.

Recently, Meta dropped $14.3 billion into Scale AI in June, luring in its genius CEO, Alexandr Wang, along with a few other hires. That move sparked the launch of Meta’s new “Superintelligence Lab,” which sounds big on its own. Wang is now Meta’s chief AI officer, working with ex-GitHub boss Nat Friedman and serial founder Daniel Gross.

Meta also tried, unsuccessfully, to buy Safe Superintelligence, a startup co-founded by OpenAI co-founder Ilya Sutskever.

While Meta did not get him, the company took a bunch of top-tier researchers from OpenAI, Apple, and Google. Shengjia Zhao, who co-created ChatGPT, is now leading Meta’s AI lab as chief scientist; that’s not a small catch.

Zuckerberg is betting that if he builds the biggest, baddest AI team and gives them crazy amounts of compute power, they will crack the code to leapfrog the competition.

He even said Meta Superintelligence Labs will have “by far the greatest compute per researcher,” which sounds impressive. However, the market is wondering what this actually means for product and profit.

Analysts at Bank of America have indicated Zuckerberg’s comments are a sign of confidence, and while capex would increase, this also signals Meta as a “place of innovation.” The analysts wrote: “We expect AI investment to be a top focus area on the upcoming earnings call, and Meta likely needs to make a case for strong AI returns to drive multiple expansion.”

Meta’s Llama 4 missed the mark

Despite the efforts and huge investments into AI, Meta seems to have missed it with some products. Earlier this year, Meta tried to mimic Chinese startup DeepSeek’s new AI model and ended up stumbling.

The result of that initiative – Llama 4, its new open-source model, landed with a dull thud. Developers preferred the older Llama 3 because it was easier to work with. They have argued that the newer version was too rigid and too complicated.

The fallout was enough to trigger a full-on internal shake-up. Some at Meta even wanted to ditch the promised “Behemoth” version of Llama 4 altogether in favour of a closed-source, tightly-guarded model.

That would be a pretty stark U-turn from Meta’s open-source AI stance, though the company insists that position has not changed, at least not officially.

Behind the scenes, frustrations have been bubbling. Some engineers pushed to stick with the older dense AI model approach. Others pushed for the newer “mixture-of-experts” (MoE) style, more efficient, more flashy, more like what DeepSeek and OpenAI are doing. In the end, Meta went with MoE, but the release flopped.

However, it has not been all doom and gloom for the company. To be fair, Meta’s core ad business is still alive and kicking. That has helped soften investor nerves while Zuckerberg spends like a tech CEO in a sci-fi movie.

While analysts at Bank of America said his massive infrastructure investments are s sign of confidence about long-term revenue, confidence, however, does not pay dividends yet.

Some folks are getting flashbacks to the self-driving car frenzy of 2017, when tech giants handed out absurd salaries just to keep AI talent in-house. “Winner takes all” feels like the mood again. The only difference this time is that everyone is playing – OpenAI, Google, Anthropic—and now Meta is sprinting to catch up. Zuckerberg’s latest strategy shift shows he is willing to throw everything at the AI wall and see what sticks.

KEY Difference Wire helps crypto brands break through and dominate headlines fast