Meta has had a tough few weeks.

The parent company of Facebook and Instagram has lost two pivotal court cases, laid off hundreds of people and taken steps that appear to amount to a U-turn on the Metaverse, a virtual reality project that CEO Mark Zuckerberg once heralded as the future.

Analysts say Meta's hard pivot to artificial intelligence, underscored by tens of billions of dollars in investment, has been a mixed bag, and if the hope is to compete with the leading players in the space — like OpenAI, Anthropic and Google — it's getting harder by the day.

On the legal front, a New Mexico jury found that Meta failed to protect young users from child sexual exploitation on its platforms.

And late last month, it was found liable by a jury in Los Angeles for the depression and anxiety of a woman who used social media as a small child. Google, which owns YouTube, was also found liable.

Frances Haugen, who blew the whistle on the company in 2021 over platform safety, and wrote a memoir about working at Facebook, says the verdicts validated concerns she had raised earlier.

"They explored a lot of different ways that they could have built their product. They weighed pros and cons of those different methods," she said. "And in the end, time and time again, they chose options that were more profitable over options that were more safe."

Meta has said it disagrees with the court outcomes, and will appeal — as will Google. In a statement emailed to NPR, a Meta spokesperson said teen mental health is complex "and cannot be linked to a single app," and noted that the company has safeguards in the form of parental controls and teen accounts.

But the verdicts put a dent in a defense that social media companies have used in the past, that the problem is the content — which they do not control — rather than the delivery method. Experts say the trial outcomes open the door to a flood of similar cases that will likely embroil Meta, and others, for years.

The company also announced that it was laying off some 700 people, many of them part of Reality Labs, the division that runs the company's Metaverse products.

When it launched its Metaverse concept in 2021, Zuckerberg changed the name of the company from Facebook to Meta, a way to emphasize how central virtual reality was to his vision for the future. In a video, he described an online experience that would allow people "to feel present like we're right there with people no matter how far apart we actually are. We'll be able to express ourselves in new, joyful, completely immersive ways."

Now, the company is "right-sizing" (i.e. shrinking) its investment in Reality Labs, although Samantha Ryan, its VP of content, wrote in the announcement: "We aren't going anywhere. We're in it for the long haul."

But the Metaverse has struggled to take off. "I think of it as, yeah, as absolutely one of the largest fails," said Megan Duncan, an associate professor in the school of communication at Virginia Tech who studies social media. For users, she said, "the Metaverse itself was not capturing their imagination, or being able to see a practical reason and practical use for it."

Zuckerberg's focus on the Metaverse had another consequence, she said: "He went all in on it. And he missed the boat on AI."

Meta last year invested more than $70 billion in artificial intelligence, and expects to nearly double that this year. On a recent earnings call, Zuckerberg said the company is on a "rapid trajectory" and would "start shipping new models and products" in the coming months.

And the company points to successes like rising sales of its AI-enabled glasses.

But some critics say competing with leaders in the field is getting harder by the day in a field dominated by OpenAI's ChatGPT, Anthropic's Claude and Google's Gemini.

"I don't think that Meta is going to be able to build a best-in-class generalist model, because from a resource standpoint, both with respect to GPUs [graphics processing units] and human talent, it's just not going to be easy to do," said Arnal Dayaratna, a research vice president at the tech consultancy IDC.

He points to the eye-popping sums that companies like OpenAI, Anthropic, Google and Nvidia are plowing into their own initiatives to hoover up GPUs that are crucial for AI data centers, and to recruit top engineers.

"I think they're too far behind," he said of Meta. He says they might have better luck picking a niche, like AI video or image creation, rather than trying to compete with all-purpose models like ChatGPT.

Steven Levy, an editor-at-large at WIRED who wrote the book Facebook: The Inside Story, said Meta's AI efforts haven't been a total flop.

"They've actually done a great job in using AI to power their feeds and their ad networks as well," he said — advertisement networks which are crucial revenue engines for the company. "So far they haven't really shown that they can build a tremendous frontier model."

But he points to successful pivots by Meta in the past — like when Facebook went from being a college-focused website to one that anyone could access, or when the company successfully adapted to smartphones, or when it shifted to being "a broadcast network" feeding users a wide range of content rather than just a place to check in on friends.

The company also mints money. Last year it reported revenue of more than $200 billion.

The company, with founder Zuckerberg at the helm, is being tested — and seems to know it. Last week it unveiled massive pay incentives that will reward top executives if they can innovate in a way that pushes Meta's share price to hit soaring targets. It's similar to the stratospheric pay structure Tesla shareholders voted in for CEO Elon Musk last year.

Dayaratna says this is a good thing.

"I'm certainly glad to see it," he said. "It means that they're still alive in some way, like, sort of not giving up yet."