AI dealmaking in North America appears to be entering a new phase.

After the enormous venture funding rounds and talent-driven ‘acquihires’ that dominated the last couple of years, recent deal flow suggests buyers are increasingly pursuing larger, conventional acquisitions aimed at securing technology, customers, data assets, and distribution.

SpaceX’s USD 54.75bn acquisition of Anysphere, the developer of AI coding agent Cursor, in June; Nvidia’s USD 12.93bn purchase of AI model repository Hugging Face in September; and AMD’s USD 8.2bn acquisition of spatial AI developer World Labs, also last month, all involved the purchase of established AI businesses.

That trio of deals marked a shift from the unprecedented capital raises by frontier AI models that defined the earlier wave of AI transactions. AI funding rounds were led by OpenAI’s USD 122bn raise in February 2026 and Anthropic’s USD 65bn in May, USD 30bn in February, and USD 15bn last November.

Alongside those mega financings came a succession of high-profile acquihire transactions, including Meta’s USD 14.3bn investment in ScaleAI, which brought founder Alexandr Wang onboard; and OpenAI’s USD 6.5bn acquisition of Jonny Ive’s IO Products, widely viewed as a talent and product-design driven transaction.

According to Mergermarket data, AI M&A and fundraising volume reached USD 450bn through the first nine months of 2026, triple the same period a year ago. Funding rounds accounted for 74% of that total. From June through September, however, traditional M&A accounted for 73% of total dollar volume.

While investors and advisors caution that AI remains an emerging industry and that buyers are still determining how to value AI businesses, they increasingly see acquisitions becoming an important complement to the funding rounds and acquihires that characterized the market’s earlier stages.

Three buckets

The growing prominence of large acquisitions suggests the AI market is in the early stages of maturing, said Ajay Agrawal, co-founder of Intrepid Growth Partners.

According to him, acquisitions now generally fall into three main categories.

In the first camp are purchases of novel algorithms and specialized talent, Agrawal said, pointing to AMD’s acquisition of World Labs. Founded by AI pioneer Fei-Fei Li, World Labs specializes in 3D ‘world models’ that are useful for robotics.

A second category includes acquisitions intended to gain access to customer bases and embedded workflow integrations.

The third type of acquisition involves strategics acquiring a “piece of the puzzle” for the creation of a broader end-to-end solution, Agrawal explained. Recent examples include Nvidia’s Hugging Face deal, as well as Salesforce’s roughly USD 2bn purchase of ListenLabs, which helps analyze customer interviews.

“People are starting to develop an intuition around what their full stack should look like,” he said.

Unlike previous technology cycles, AI improves through use, creating a feedback loop that strengthens products as they attract more users, Agrawal said. That dynamic is likely to accelerate AI dealmaking because companies may struggle to replicate years of accumulated learning once a rival establishes an early lead. That shifts the build-versus-buy calculus in favor of buying because time is more important than ever with AI.

“The more people that use it, the more it learns, and the better it continues to get. Because of that feature of AI, it’s really hard to catch up,” he said.

Despite the recent wave of large strategic deals, some believe the industry is still early in its ramp.

“It still feels like we are in the growth stage,” said Eric Issadore, a partner at Ropes & Gray. “I’m seeing a lot more of the minority investments, even if they’re not mega rounds, than traditional M&A.”

Mergermarket data supports that view: Funding rounds still account for about 60% of the total number of AI deals, a proportion that has remained constant over the last three years.

In other signs of a still-emerging industry, financial sponsors have yet to become major buyers of AI companies and acquirers continue to grapple with how to value AI businesses and their growth prospects, Issadore added.

But the emergence of those larger strategic acquisitions in recent months suggests the market is beginning a gradual maturation process, although it remains at an early stage, he said.From talent wars to applications

For Mark Kraynak, founding partner at Acrew Capital, dealmaking activity in AI is “all across the spectrum.”

“Way more than half” of Acrew’s portfolio companies are engaged in fundraising, acquisition discussions, acquihire conversations or broader strategic reviews, he said. “The market is very active.”

While hardware companies like Nvidia and AMD are making acquisitions to create ecosystems around AI, frontier AI labs are still frequently focused on talent.

“The labs are still incredibly acquisitive. I would say they are very hardcore about just buying talent,” he said.

But even among frontier model companies, priorities may be changing. As the capabilities of open-source models close the gap with the leading AI labs, questions are being raised over the long-term sustainability of their business models.

One concern is whether enough profit exists for infrastructure providers, foundation model developers, and application companies simultaneously. “Everybody needs margin in that picture,” Kraynak said.

Kraynak expects the leading model developers to want ownership of applications and customer relationships because those layers may provide stronger economics than foundation models alone. “They know there is likely more margin at the application layer,” he said.

Another source of future deal activity could come from the software sector. Issadore said many SaaS companies have spent the past two years figuring out how to incorporate AI into their products, either through internal development or acquisitions. Some that have succeeded and feel better insulated against the AI risk, may explore transactions or sale processes in the coming months.

If the first chapter of AI dealmaking was a race to fund the model builders, the next may be a race to own the businesses and applications that turn AI into enduring and profitable revenue streams.