After the DeepSeek whiplash, Meta Platforms has become the brightest star in the Magnificent Seven.
Its share price is up 20% so far this year, pushing it to a record valuation of around US$1.8-trillion, with $500-billion added in just the past 12 months.
At a time of frayed nerves over the return on investment for artificial intelligence, Wall Street is buying what CEO Mark Zuckerberg is selling — the stock is trading at 27x blended forward earnings, the highest since September 2020.
There are several reasons for this optimism — and just as many reasons to question it.
Driving the rally is the sentiment that Meta’s AI strategy has been validated by the emergence of DeepSeek, the open-source AI model out of China. While the initial market panic was in response to fears that China was wiping the floor with Silicon Valley, what it showed — as far as Meta was concerned — was that cheaper and more open AI models might be the way ahead. If that’s the case, as the leading US creator of such a thing, Meta is in a strong position.
Then, Meta came out with a positive earnings report that suggested its advertising business was performing terrifically. A part of this was thanks to the fact that it was applying its AI chops to help advertisers create ads and target users more effectively. A collaboration with Nvidia generated an 8% increase in “quality” ads being shown to users. Four million clients are now using Meta’s AI tools to make better ads, a fourfold increase from just six months ago.
Zuckerberg reiterated to investors that his company stands to be the first to have a billion users of its chatbot. Meta’s core strength, the numerous huge platforms it runs, means there are 3.35 billion people it can prompt to use AI (regulations in some regions notwithstanding). This will be a “durable long-term advantage”, Zuckerberg said.
Facelift
He’s also promised that the original Facebook app, long neglected on the innovation front, will get something of a facelift. Zuckerberg wouldn’t detail his plans precisely, but he said a focus in the coming year would be to make the app more “culturally influential”.
That he thinks he can make Facebook relevant again underscores the swagger from Zuckerberg that is seeping into investors’ outlook. They see the gold chain, curly hair and closer relationship with President Donald Trump and the right and think this can all be good for Meta’s prospects. The decision to do away with fact checkers and loosen moderation policy hasn’t hurt ad sales, the company noted. In return for his rightward lurch, he might enjoy retreating regulatory fights at home and an ally in fights against the EU. “We now have a US administration that is proud of our leading companies,” Zuckerberg said.
Read: DeepSeek is a big test for Meta and its embrace of open-source AI
But a company can’t get by on “masculine energy” alone. The question that hangs over Meta is how its AI forays will meaningfully drive revenue in the future. Unlike those of its competitors, which are building AI systems and data centres to sell that computing power, Meta’s AI ambitions are markedly more inward — a client base of one.
Its open-source Llama model, impressive as it is, is being given away to anyone who wants it in the hope that it will become smarter more quickly. Then, the plan is to apply that intelligence to its own products in the form of chatbots and other as-yet undetermined enhancements. But to what end? Meta’s AI may well gain a billion users, but how this achievement will make money is up in the air. It may be ads. Or businesses might pay to integrate with it to reach Meta users. That’s not a given — once upon a time Meta thought businesses would flock to integrate with its Messenger app; they didn’t.
Building Meta AI into more than a distraction or a gimmick will be difficult. Its utility is less impressive than the “agentic” capabilities of an assistant that can tap into your calendar and mapping software, like Google’s, or integrate into your business’s day-to-day work tools, like Microsoft’s. Nor will it be at users’ immediate fingertips, like Google Assistant or Apple’s Siri (assuming Apple can solve its own problems).
Even on writing code, one of the most promising early uses of AI, Zuckerberg conceded his “good mid-level engineer” isn’t going to be available externally “anytime soon”.
I’m a fan of virtual reality, but the suggestion that AI could enhance Meta’s “metaverse” segment feels flimsy. Adoption of headsets and sunglasses is still poor. Despite new models and an advertising blitz, Meta managed year-on-year revenue growth of just 1% in the last quarter. Meta lost $17.7-billion chasing the metaverse in 2024.
Read: Trump AI gaffe has Meta scrambling
Zuckerberg is right to champion the gains from AI that Meta has seen so far. But they mostly relate to improving its existing business of flogging ads to doom-scrollers rather than building an exciting new venture in AI. Grander ambitions, and potential rewards, can be found elsewhere. — Dave Lee, (c) 2025 Bloomberg LP
Get breaking news from TechCentral on WhatsApp. Sign up here.
Don’t miss:
Microsoft, Meta defend hefty AI spending after DeepSeek stuns tech world