The Top AI Themes of 2025 and What We’re Watching For Next Year

Meta Platform’s announcement on Monday that it has acquired Chinese agent startup Manus represents a big win for Manus’ backers, including Benchmark, ZhenFund and HongShan, who last invested in Manus in an April funding round that valued it at $500 million. Meta is paying more than $2 billion, my colleague Juro Osawa has reported (as have other outlets).

That’s a huge gain on an eight month old investment. The deal also continues Meta’s acquisition spree as it looks to revamp its AI organization. Whether this deal will be the one to help Meta make a comeback, though, is the question.

Onto today’s column…

This year, the AI party continued, with VC funding, research talent and circular compute deals flowing freely. Still, we’re seeing signs that partygoers are getting a little weary.

For instance, there’s been one or two reports of data center delays lately and we expect the new year will bring even more, even while AI firms race to build billions of dollars worth of infrastructure to power their AI models. Meanwhile, the lead in the AI chatbot market that OpenAI once enjoyed has narrowed considerably. At this point, the three major AI developers—OpenAI, Anthropic and Google—are pretty much neck-and-neck when it comes to the performance of their models, reviving questions about the commodification of large language models. 

It doesn’t help that major users of AI models like Salesforce and Microsoft are having trouble selling AI-powered products. And all the while, talk of an AI bubble is hanging around like that wet-blanket partygoer who won’t leave. 

In the meantime, let’s walk through the top 10 most unforgettable AI themes from the past year.