These 8 big deals defined the year in artificial intelligence

Nvidia ended the year doing the same thing that turned heads all year: extracting founders from their startups and leaving other employees to figure it out.

On Christmas Eve, Nvidia announced it had entered into a non-exclusive licensing agreement with Groq. The financial terms were undisclosed.

Under the agreement, Jonathan Ross, Groq's founder and CEO, as well as the startup's president and other team members, will join Nvidia, the world's most valuable company with a market capitalization of over $4.5 trillion.

Groq will continue to operate as an independent company, with CFO Simon Edwards assuming the CEO role.

Disney struck a three-year licensing agreement with OpenAI in December, becoming the first major content partner for Sora, OpenAI's video generator.

The deal isn't just about rights. Disney will invest $1 billion in OpenAI and receive warrants to buy additional equity — a structure that ties content, distribution, and upside together.

Business Insider reporter Lucia Moses argues the logic can be summed up in one word: engagement. Time spent on Disney+ and other major streaming services has remained essentially flat in recent years, she writes, while YouTube and social video continue to grow.

The OpenAI partnership could help Disney meet audiences where they already are, by allowing fans to use tools like ChatGPT and Sora to generate and remix content featuring Disney characters.

Developer coding assistant Windsurf was a startup Cinderella story. Then it became a cautionary tale.

The company was in talks to sell to OpenAI for $3 billion this past spring. The deal fell apart at the last minute.

Google emerged as a new suitor. Instead of an outright acquisition, Google said it would pay $2.4 billion to hire its CEO and top talent, as well as license the company's intellectual property.

The deal effectively split Windsurf into two pieces, with the hundreds of remaining employees ending up at Cognition, another AI coding startup. That left some with a bitter taste.

Part of the social contract in startups is that early employees work long hours at lower pay in the hope that they will share in the riches should their startup eventually get acquired or go public.

Now, traditional acquisitions have become rarer because of the time and uncertainty of getting regulatory approval. As a result, companies like Google, Microsoft, and Meta have been getting creative, using licensing deals to skirt regulators and snap up key talent.

SoftBank snapped up a private equity firm with the ultimate goal of controlling more of the physical infrastructure behind AI.

On Monday, SoftBank said it has entered into an agreement to acquire DigitalBridge, an alternative asset manager that invests in data centers, cell towers, fiber networks, and power that make the AI boom possible.

The transaction is expected to close in the second half of 2026, subject to regulatory approvals.

The deal comes as SoftBank reshuffles its portfolio. In November, the company said it had sold nearly $6 billion of Nvidia stock to free up cash for OpenAI. SoftBank is also doubling down on "physical AI," building a robotics beachhead with a dedicated investment arm and a pending deal to buy ABB's robotics business.

The Trump administration agreed to take a 9.9% stake in Intel in August. It was a rare moment when Washington wasn't just regulating an industry or subsidizing it, but buying into it.

Under the terms Intel disclosed, the government will invest $8.9 billion for 433.3 million shares, and the stake is to be passive, with no board seat.

The investment aims to bolster domestic capacity for designing and manufacturing advanced chips. Intel has seen foreign competitors, such as Taiwan Semiconductor Manufacturing Company, cut into its market share of advanced chips. The White House has said it is a national security risk if industries continue to import the vast majority of advanced chips from overseas.

This also likely won't be the last deal of its kind. President Donald Trump and top aides have floated the idea of doing similar deals with American defense contractors and have considered a sovereign wealth fund.

One of the year's biggest deals sparked the question: Was Meta really spending $14 billion to hire a 28-year-old?

Alexandr Wang cofounded and ran Scale AI, the data-labeling company that helps enterprises train and deploy AI models by supplying high-quality training data and tooling. In June, Meta agreed to pay $14.3 billion for a 49% stake in Scale AI — and, just as importantly, to bring Wang in to lead its next act in AI.

Wang became Meta's chief AI officer and the public face of its superintelligence push, with a mandate that goes well beyond building better chatbots. His job is to convince elite researchers that Meta is the place to do their life's work — and to turn the company's key advantage, its compute budget, into a lasting edge.

When Google agreed to buy Israeli-American cloud security firm Wiz earlier this year, it threw a $32 billion bucket of water on the idea that the mega-deal drought would persist.

In March, Google agreed to acquire the startup at an all-cash price roughly equal to Iceland's gross domestic product last year, pending regulatory sign-off. This marks the search giant's largest acquisition.

The acquisition served as an early litmus test of the Trump administration's willingness to greenlight pricey Big Tech mergers and acquisitions. The deal passed an antitrust review by the Department of Justice in November, clearing a key hurdle to closing.

Trump kicked off his second term by unveiling a joint venture with OpenAI, Oracle, and SoftBank to build a chain of data centers across the US — a project called Stargate that aims to invest $500 billion in domestic infrastructure by 2029.

In Abilene, Texas, a flagship data center has already opened. Construction crews have broken ground on Stargate sites in Lordstown, Ohio, and Port Washington, Wisconsin, with several more campuses to come.

To make the ongoing buildout a success, OpenAI CEO Sam Altman has estimated it could take one-fifth of the nation's existing skilled trade workforce.