Skip to content

Zuckerberg is excited about the future of AI at Meta

Facebook and Instagram remain cash cows. Their advertising profits fund Meta, Mark Zuckerberg's company, in billions of investments in Artificial Intelligence.

Meta's AI ambitions, among other things, are reflected in high costs for data centers.
Meta's AI ambitions, among other things, are reflected in high costs for data centers.

- Zuckerberg is excited about the future of AI at Meta

Meta, the Facebook corporation, is raking in billions from its advertising business and plowing a significant portion of it into expensive expansions of data centers for artificial intelligence. Founder and CEO Mark Zuckerberg sees Meta AI's chatbot on track to become the world's most used AI assistant by the end of the year.

In presenting the latest quarterly results, Zuckerberg painted a picture of a future where influencers could have personalized AI chatbots mimic them for fans, and advertisers could hand over their campaigns on Meta platforms entirely to software.

But do people really engage in genuine conversations with AI chatbots? Zuckerberg claimed that users already use Meta AI for "rehearsing difficult conversations before having them with a human" or searching for information. However, this example was not well-chosen, as just hours earlier, Meta admitted that its chatbot had told some users that the attack on former President Donald Trump was fiction. The company blamed this on the known issue of "hallucinations," where AI software fabricates information.

And Meta's AI visions come at a high cost. The company's expenses rose by seven percent to $24.22 billion in the last quarter. For this year, Meta now expects costs between $37 and $40 billion, and it's preparing investors for significant growth in 2025. Especially, the computing power for training AI models is expensive.

Zuckerberg estimated that the next in-house AI model, Llama-4, would require about ten times more computing power for training than the current version. Meta is already a major customer of chipmaker Nvidia, whose systems dominate AI training, and units can cost several thousand dollars each.

The fear of not having enough computing power

Zuckerberg acknowledged that predicting the need for computing power is difficult, but he wants to be on the safe side: "At this point, I'd rather risk building capacity before it's needed than be late." Investors seem to agree, as the stock rose more than seven percent in after-hours US trading.

Meanwhile, Chief Financial Officer Susan Li emphasized that the AI infrastructure could be used for various purposes, from training AI to personalizing video selection for individual users, which could help in competition with platforms like TikTok.

And Meta has the financial cushion thanks to its booming advertising business. In the last quarter, the company's revenue jumped 22 percent year-over-year to around $39 billion, while profit surged 73 percent to nearly $13.5 billion (€12.44 billion).

About 3.2 billion users accessed at least one of Meta's apps daily. The business of applications like Facebook, Instagram, and WhatsApp generated an operating result of $19.3 billion in the last quarter. And Threads, Meta's still ad-free alternative to Elon Musk's online platform X, is on track to reach 200 million monthly users.

Besides AI, Meta has another expensive bet on the future: Reality Labs, where the company bundles its businesses in virtual worlds and computer glasses, continues to swallow up significant amounts of money. The segment reported an operating loss of nearly $4.9 billion, compared to a loss of $3.74 billion a year ago.

Over the years, Meta has already burned through around 50 billion dollars in this niche business. However, Zuckerberg points to the popularity of the camera glasses developed with Ray-Ban, which could provide important contextual data to Meta's AI about users' surroundings. He argues that this would allow an AI assistant to see what the user sees.

After discussing the potential of AI chatbots, Zuckerberg suggested that Instagram, with its large user base, could be an ideal platform for influencers to interact with their fans using AI-driven bots.

Despite the high costs of training AI models, Meta is actively investing in expanding its data centers, recognizing the importance of computing power in advancing its AI capabilities, including on Instagram.

Read also:

Comments

Latest