Lisa Su and Fei-Fei Li Join Forces as AMD Acquires World Labs for $8.2 Billion

0 0

Just now, Silicon Valley saw an epic, high-stakes acquisition.

World Labs, the spatial intelligence startup founded by "AI godmother" Fei-Fei Li, is being acquired by AMD for a headline-grabbing $8.2 billion.

Lisa Su and Fei-Fei Li Join Forces as AMD Acquires World Labs for .2 Billion

Two Silicon Valley queens—Lisa Su and Fei-Fei Li—joining forces is a stunner.

A compute giant is diving into a native 3D ecosystem itself, marking AI's formal turn from language toward the physical world.

In her open letter *To Seek a Newer World*, Fei-Fei Li quoted Tennyson's *Ulysses*:

"Come, my friends, 'Tis not too late to seek a newer world."

A Historic Alliance: Chip Queen and AI Godmother Unite

Today AMD officially announced it has reached a definitive agreement to acquire World Labs, led by Fei-Fei Li.

After the deal closes, Li will become AMD Executive Vice President and Chief Scientist, reporting directly to AMD Chair and CEO Lisa Su.

This $8.2 billion deal is a handshake between two Chinese-American queens—far heavier than a routine commercial buyout.

On one side is "chip queen" Lisa Su. Taking over a money-losing AMD in 2014, she staged a turnaround from the brink and made it Nvidia's toughest rival.

On the other is "AI godmother" Fei-Fei Li. Once a Chengdu girl working odd jobs in New Jersey and minding the family dry cleaners, she later built ImageNet and helped open the door to the deep learning era.

Their connection goes back years.

When World Labs was founded, Su was an early investor and steadfast backer; starting last year the two sides also trained models together on AMD GPUs.

At CES this January, Su invited Li onto the keynote stage and announced that Marble, optimized on AMD's MI325X, had sped up by more than 4×. Li said on the spot she couldn't wait to see it improve further on the MI450.

Nine months later, Su simply bought the whole company. Li becomes AMD EVP and Chief Scientist, reporting directly to Su.

Why these two legendary queens bind now is spelled out in one line of Li's essay: "The universe is not made of words—it is made of real things."

As large language models push closer to a ceiling, tech giants increasingly believe that for AI to truly change the world, machines must understand, perceive, and reshape the real physical world.

Neither side can do that alone.

Su needs a top team to tell her years ahead what chips the next generation of models will require. For Li to make AI generate walkable 3D worlds in real time, the appetite for compute is enormous—she writes that without dedicated hardware investment, AI will stay trapped in the digital world.

Walking AI into the physical world is World Labs' ultimate mission—and for AMD, a critical bend in the road to overtake rivals.

### **"AI Is Personal—and Civilizational"

In the letter, Li rarely dissected her 26-year AI journey and wrote a line that sticks:

For me, AI is both personal and civilizational.

In 2000, entering academia as a graduate student, she was drawn to a question: how do we solve visual intelligence for machines and brains?

She didn't know then that those curiosity-driven steps would land on nearly every historic node of the AI path that followed.

In 2009, AI was still in a long "winter," and the industry was pessimistic about neural nets.

But Li and her team did something that looked "clumsy"—they released ImageNet. That massive dataset of tens of millions of human-labeled images fed algorithms with enormous "nourishment."

ImageNet, plus neural-net algorithms and their GPU implementations, together ignited the 2012 AlexNet miracle and directly birthed the deep learning era.

You can say: without ImageNet, there would be no AI boom today.

In the decade-plus that followed, Li's career looked almost unlocked:

Top research spanning computer vision, deep learning, robotics, and healthcare;

A jump into industry as founding Chief Scientist of Google Cloud's AI group;

A return to academia as founding director of Stanford HAI…

Every turn had a clear logic—"human-centered."

Her work has always been driven by a strong urge: not to show off technique, but to solve real-world problems and advance social good alongside people.

Now Li senses that academic research alone can no longer keep pace with the era's sprint.

"I knew the path with the deepest and broadest potential for positive impact was to start a company and tackle the boldest technical problems."

So in early 2024 she teamed up with world-class AI researchers Ben Mildenhall and Justin Johnson to found World Labs.

Her goal was crystal clear: build a frontier lab focused on AI's most important challenges—ones that, once solved, would sharply raise AI's ability to handle real-world problems.

In just two years, earlier this month World Labs delivered a shock to the industry: the multimodal world model Atlas.

The idea: from a few 2D images, predict the "next view." It is the first model of its kind trained from scratch, able to generate, reconstruct, and simulate interactive 3D environments directly from text, image, and video inputs.

AI has finally grown a true pair of "3D eyes" that can read the physical world.

Lisa Su Bets on AI's Next Fight

The person really playing this board is Lisa Su.

When she became AMD CEO in 2014, the company was worth under $3 billion with about $4 billion in annual revenue—Wall Street thought it was nearly dead.

Twelve years later, AMD's market cap once crossed $1 trillion, and the stock had risen 45×.

What did she get right? Two wild bets.

First, selling the fabs and outsourcing all chip manufacturing to TSMC. Everyone thought AMD was crazy to give up "making its own chips." But TSMC's advanced nodes later gave AMD the capital to fight Intel and Nvidia head-on.

Second, betting on chiplet architecture—breaking chips into small dies and packaging them. The risk was extreme then; today it's an industry standard.

In 2025, AMD signed a 6-gigawatt GPU supply deal with OpenAI, shaking Nvidia's "one company rules all" myth.

The MI355X beat Nvidia's B200 on multiple tests, delivering 40% more tokens per dollar.

But Su looks further: hardware alone isn't enough.

The next AI contest is a full-stack war of "chips + models + apps." So when Li founded World Labs, Su was among the first investors.

The two sides have trained models on AMD GPUs since last year. Buying World Labs gives AMD a world-class world-model team that knows early what future AI models will need from hardware.

Where compute is slow, where data won't move, which designs need to change—model and chip teams can find answers together.

Of course, $8.2 billion is only the starting point. The deal is expected to close by year-end and still needs regulatory approval; how much efficiency tech synergy brings will have to be proven by later results.

But Su's intent is already clear: wherever AI's next round goes, AMD wants a seat at the table earlier.

"Building the next generation of AI computing platforms requires a deep understanding of how models will evolve," Su said in the acquisition announcement—cutting to the chase.

World Labs' top team joining AMD means AMD will have one of the world's sharpest "algorithm brains" guiding future chip design.

That is the real power of the dual blades joining.

From here, Fei-Fei Li gets unlimited compute ammunition and freedom at the hardware layer to accelerate spatial intelligence.

And Lisa Su gets the edge that lets AMD's hardware grow directly onto the next generation of the most advanced AI models.