At Meta, will AI make people forget the Metaverse?

The social media giant's latest release has once again highlighted the scale of its spending on artificial intelligence. While investors initially welcomed solid results and what were seen as encouraging forecasts, Meta will have to justify its colossal investments quickly - notably through the next AI models Mark Zuckerberg has promised to launch in the coming months.

Rising results and optimistic prospects were enough to reassure the market, sending Meta shares higher (+10.4% on January 29). That confidence remains fragile, however. It could quickly be called into question if the Facebook parent fails to deliver on its technological promises. Another setback would remind investors that Meta is spending freely on artificial intelligence, with no guarantee of success. If it fails, AI could end up like the Metaverse once did: a financial sinkhole that did little more than drain cash.

A recent all-in on AI

Meta is planning nearly $122bn in capex in 2026. That would be a record CAPEX figure, surpassed only by Amazon for the current year. The effort is all the more colossal given that the group is still struggling to establish itself as a leading player in generative AI. Its open-source Llama model, meant to embody Meta's technological ambition, is lagging in an extremely crowded competitive landscape. 

Aware of this shortfall, Mark Zuckerberg has decided to change strategy. In the summer of 2025, Meta was already making headlines with the creation of "The List" a veritable super-team dedicated to AI. To attract these highly sought-after profiles, the Meta chief did not hesitate to offer spectacular pay packages, sometimes well beyond industry standards. The focus now shifts to the group's next two models: Avocado and Mango. Avocado is presented as a future rival to ChatGPT (OpenAI) and Gemini (Alphabet). Expected in the coming months, it will be dedicated to text. Mango, for its part, will handle image and video. Its launch is planned for this year, in a window close to that of the text model.

Behind this offensive lies a double objective. First, to launch a proprietary (and therefore non-open-source) model that is more powerful and genuinely competitive. Second, to position Meta as a key player in AGI, artificial general intelligence, capable of learning autonomously in a way comparable to human intelligence. Zuckerberg even invokes, ultimately, superintelligence - an AI that would exceed human capabilities. Meta's chief has clearly chosen a side in the controversy pitting AGI advocates against its opponents. 

A turning point now imminent

For now, this superintelligence narrative has won over investors. But recent history shows the market can turn very quickly. When third-quarter results were released, the announcement of another increase in investment triggered a sharp пад in the stock, on the order of 11% in a single session. Since then, the share price has returned to its previous levels and the analyst consensus appears more confident. Meta is up about 12% since the start of the year, while the Nasdaq is flat.

The challenge is now clear: deliver. Meta must convince the market it can become credible again on the AI front and, perhaps, set new all-time highs on the stock market. In that context, Avocado, the announced successor to Llama, is highly anticipated. The task nonetheless looks complex in the face of competition that has clearly intensified in recent months.

If the performance of this new model were to disappoint, investors could conclude that Meta has missed the AI turn. The stock could then suffer a fate comparable to that of Apple, whose artificial intelligence strategy also remains to be confirmed.

To end on a more positive note, Meta still has a major advantage: its ability to monetize AI at the heart of its advertising business. In the latest quarter, of $59.89bn in revenue, no less than $58.13bn came from advertising. A solid base the group intends to rely on to make its technology investments pay off and, perhaps, convince markets for the long term.