Mark Zuckerberg spent the past year rebuilding Meta’s place in the artificial intelligence race. Now the Meta CEO faces a tougher assignment. He must convince investors, developers, and users that his $14.3 billion wager on Alexandr Wang can grow into a real business. That challenge sits at the center of the Meta AI bet.
Meta pulled Wang, the Scale AI founder, into its push during June 2025. Zuckerberg took a 49% stake in Scale AI for $14.3 billion, then handed Wang the keys to a new lab. Meta refused to keep trailing OpenAI, Anthropic, and Google in the fiercest technology market on the planet.
One year later, the company carries both momentum and doubt.
A model arrives, the pressure builds
Meta Superintelligence Labs, the team Wang runs, shipped Muse Spark on April 8, 2026. The company called it the first model from the new group, then integrated it into the Meta AI app (meta.ai), Facebook, Instagram, WhatsApp, Messenger, and its Ray-Ban Meta glasses.
That gave Zuckerberg a product story. It did not hand him a profit story. So the Meta AI bet still waits for proof.
Meta’s stock tells part of the tale. Shares have slipped roughly 18% over the past year, even as rivals climbed. Investors want returns, not pledges.
Investors want dollars, not demos
Meta still earns almost every dollar from advertising. That strength doubles as a weakness.
The company posted $56.31 billion in first-quarter 2026 revenue, up 33% from a year earlier. Advertising delivered $55.02 billion of that total. Clearly, the core apps still print cash.
But the same numbers reveal how much Meta leans on a single engine. Zuckerberg has used machine learning for years to sharpen ad targeting, ranking, and recommendations. Wall Street understands that playbook. Now it wants paid AI tools that reach beyond ads, which is the heart of the Meta AI bet.
William Blair analyst Ralph Schackart framed the pressure plainly.
“Meta needs to provide more proof points of both adoption and commercialization,” he said.
A strategy shift with real risk
Muse Spark also flipped Meta’s approach. For years, the company leaned on open-weight Llama models. That move won fans among researchers and developers. But Llama 4 underwhelmed the market, so Zuckerberg changed lanes.
Muse Spark runs as a proprietary, closed model built for Meta’s own apps first. That design gives Meta an edge that rivals cannot copy easily. OpenAI, Anthropic, and Google fight hard to pay users. Meta already owns platforms where billions of people spend hours each week.
Still, the strategy carries danger, and it shapes the Meta AI bet. Meta has delayed the Muse Spark developer interface since April, though it says a public launch remains on track for June. If outside builders lose patience, Meta could look like a closed product shop guarding ads rather than a true platform.
Rob May, CEO of the startup Neurometric, put the developer problem bluntly.
“I think the AI community largely ignores Meta at this point,” he said.
That skepticism matters. Developers often decide which models gain cultural and commercial weight.
A trust test for every launch
The Meta AI bet also arrives during a wary moment across business. Companies know AI can cut costs, speed work, and lift customer service. They also know it can trigger expensive errors.
Air Canada learned that when a chatbot gave a passenger wrong bereavement-fare advice, and a tribunal ordered a refund. Zillow learned it when its home-buying algorithm misread the market, forcing the shutdown of Zillow Offers. Samsung faced a scare after workers reportedly fed secrets into outside chatbots. CNET bruised reader trust after errors surfaced in dozens of AI-written stories. IBM’s Watson Health proved that hype alone cannot save a weak result.
So Zuckerberg must sell reliability, safety, and financial discipline, not just speed. The Meta AI bet depends on Muse Spark working inside products that billions of people touch.
Big advantages, bigger expectations
Zuckerberg hardly lacks firepower. Meta averaged 3.56 billion daily active users across its apps in March 2026. It owns deep cash flow, vast data, and one of tech’s strongest ad machines.
Those assets buy room to experiment. Meta can drop AI into search bars, chats, glasses, shopping tools, and creator features. It can test subscriptions and paid business tools. It can make AI feel native to social media. That path could let Meta dodge a direct brawl with enterprise-first rivals and turn AI into a daily consumer layer, which marks the long game inside the Meta AI bet.
Yet the spending stings. Meta now expects 2026 capital expenditures of $125 billion to $145 billion. It cut roughly 8,000 jobs in May to offset the load. Reality Labs losses already strained investor patience, so the Meta AI bet revives an old worry. Zuckerberg once asked Wall Street to trust the metaverse. Now he asks for faith in another costly future.
The next launches decide it
The pressure now shifts to cadence. Muse Spark gave Wang and Zuckerberg a start, not a finish. Meta must ship stronger models, open real access for developers, and prove subscriptions can grow. Above all, the Meta AI bet must stand on its own, away from the ad machine.
Meta has muscled back into the AI conversation. But Zuckerberg still has to close the business case, and a promise alone will not carry the Meta AI bet this time.
What do you think about Meta AI bet? Can Meta turn Muse Spark into a real business, or will Zuckerberg’s wager stumble like the metaverse before it? Please share your views below.