With the merger and acquisition completed, Li Feifei officially joins AMD with his team, and Lisa Su personally extends her warm welcome.
This is truly a showdown between two titans!
In the office in San Francisco, "Queen of Semiconductors" Lisa Su and "Godmother of AI" Li Feifei sat side by side on high stools, facing a full room of applauding World Labs employees.
On local time November 16, AMD officially announced the completion of its acquisition of World Labs, the spatial intelligence company co-founded by Li Feifei, which has now been fully integrated into AMD.
In the official announcement video, Lisa Su wears a black hoodie, gesturing and talking cheerfully, while Li Feifei standing next to her is already dressed in a black jacket with the AMD logo printed on the left chest.
At the bottom of the frame of this celebration video, the white subtitle even modified AMD's long-used slogan to "together we advance_worlds".
To acquire this team that was founded only two years ago, AMD offered an all-stock transaction consideration of approximately 8.2 billion US dollars.
After the closing of the deal, Li Feifei officially serves as AMD's Executive Vice President and Chief Scientist, reporting directly to Lisa Su.
From signing the acquisition agreement on September 28 to completing all closing procedures on November 13, the entire transaction took only 46 days.
14.98 Million Shares Secured, AMD's Second Largest Acquisition in History Closed
According to the Form 8-K/A supplementary document submitted by AMD to the U.S. Securities and Exchange Commission on November 16, the acquisition was fully closed on November 13.
This speed is much faster than the original expectation of both parties to close the deal before the end of the year.
This closing adopts the method of all-stock private placement, AMD has privately issued 14.978 million common shares to the original shareholders of World Labs, with a total valuation of approximately 8.2 billion US dollars (about 54.6 billion RMB).
In AMD's acquisition history, the volume of this transaction is second only to the Xilinx acquisition that closed in February 2022.
In 2022, Lisa Su led AMD to acquire FPGA leader Xilinx through an all-stock transaction, with a total value of up to approximately 49 billion US dollars at the time of closing.
This acquisition set a record for M&A in the chip industry at one stroke, and directly supported AMD's subsequent embedded and adaptive computing business landscape.
Four years later, Lisa Su once again spent 8.2 billion US dollars in AMD's stock to acquire an AI team that specializes in 3D world models.
After the transaction is finalized, Li Feifei, co-founder and CEO of World Labs, officially joins AMD's core management team, serves as Executive Vice President and Chief Scientist, and reports directly to Lisa Su.
The other two co-founders of World Labs, Justin Johnson and Ben Mildenhall, will also remain in their positions.
They will continue to promote the research and development of foundation models as an independent cutting-edge research organization within AMD, while working collaboratively with AMD's hardware, software and system engineers.
In the computer vision community, these two are also well-known figures with impressive resumes.
Justin Johnson is a PhD student supervised by Li Feifei at Stanford University. He participated in the creation of the CS231n course that almost all AI beginners have watched, and is also the core promoter of real-time style transfer and the PyTorch3D library;
Ben Mildenhall, who comes from Google Research, is the co-inventor of the NeRF paper that ignited the entire 3D vision reconstruction wave in 2020.
△
In the celebration posts released simultaneously by the two companies, World Labs also quoted the line from Tennyson's *Ulysses*: "Friends, let us go forth to seek a newer world", echoing the title of Li Feifei's previous open letter.
In sharp contrast to the enthusiasm between the two parties, the reaction of the capital market that day was somewhat calm.
After the market opened on Monday, November 16, affected by the pullback of the entire semiconductor sector and AI infrastructure concept stocks, AMD's share price did not rise, but once fell by about 5%.
Wall Street is obviously still waiting for this newly formed team to deliver its first answer of hardware-software collaborative optimization.
From 230 Million to 8.2 Billion, What Have They Built
In 2024, Li Feifei co-founded World Labs, whose main direction avoided the crowded large language model track at that time, and directly cut into spatial intelligence that can understand and simulate the 3D physical world.
When the company came out of stealth mode that year, it secured 230 million US dollars in financing, with its valuation directly reaching 1 billion US dollars.
Last November, World Labs launched its first multimodal world model Marble.
Users only need to input text, images or rough 3D layouts to generate 3D scenes that can be freely explored, edited and exported.
In February this year, World Labs completed a new financing of up to 1 billion US dollars. The list of investors included Autodesk (which contributed 200 million US dollars), NVIDIA, and AMD was also among the investors at that time.
However, what really made the industry see the technical strength of this company was the new all-around world model Atlas released on September 1 this year.
Li Feifei specifically analyzed Atlas's technical base at the CoRL 2026 conference a few days ago.
At a time when large language models generally believe in "Next Token Prediction", Atlas's core idea is "Viewpoint Is All You Need", which predicts the next perspective through precise camera pose condition control.
During the training phase, Atlas ingests text, images, videos, depth signals and 3D structure signals at the same time, and can switch between three modes of dense reconstruction, sparse reconstruction and pure text generation under the same autoregressive diffusion Transformer architecture.
In a blind test initiated by the third-party organization RapidData, the test side collected about 500,000 human evaluations to compare the camera-controlled generation capabilities of major video and world models.
Eventually Atlas ranked first, capable of generating coherent perspective videos of more than 60 seconds with a resolution of 1440p.
More importantly, it has achieved landing implementation in the robotics field.
World Labs also acquired Scenix, a robotics company that specializes in the real-scene-to-simulation and simulation-to-reality pipeline, this year.
Li Feifei demonstrated on site that the robot is fully trained in the virtual simulation environment built by Atlas, and then directly deployed to the real physical world in a zero-shot manner.
On November 12, the day before the acquisition closed, World Labs announced the full opening of Atlas to all the public and developers.
To Compete with NVIDIA, AMD Decides to Develop Its Own Models
Few days before the closing was announced, Lisa Su and Li Feifei had their first on-stage dialogue at the CoRL 2026 conference.
At the beginning of the conversation, Lisa Su judged that physical AI, especially the robotics track, will be a multi-trillion-dollar level opportunity in the coming years.
She also revealed that AMD already has a fairly large robotics research team internally, but the overall layout is still in the early stage.
Li Feifei put forward the judgment of "AI is computing, and computing is AI" on the stage.
This perfectly matches what Lisa Su emphasized in the closing announcement that "computing platforms and AI models must be jointly optimized".
When previously asked why World Labs did not start developing Atlas right away, Li Feifei once half-jokingly said bluntly, "Because Justin's team needs a huge amount of chips".
To truly make the world model bigger and stronger, the support of massive GPU computing power is indispensable.
Previously, AMD and World Labs were just partners, and the two parties have been jointly tuning model training and inference on AMD GPUs since 2025.
However, external cooperation is always separated by commercial boundaries, making it difficult to adjust the underlying pain points before the chip is taped out and finalized.
Now that the two companies have become one family, Li Feifei's team can directly bring Atlas and subsequent cutting-edge models to participate in the architectural design of AMD's future Instinct series chips and ROCm software stack, and use real spatial intelligence workloads for stress testing even before the hardware drawings are finalized.
AMD has taken a series of intensive actions to complete its business landscape recently. In addition to spending 8.2 billion US dollars to acquire World Labs and announcing the acquisition of inference chip company Taalas, it also recently planned to take a stake in quantum computing company PsiQuantum, and signed a data center computing power lease agreement with Core Scientific.
On the other side, old rival NVIDIA also spent 12.9 billion US dollars to acquire the open-source model platform Hugging Face in early September, and launched the Cosmos world foundation model, Omniverse and GR00T in the robotics and physical AI track.
Facing the hardware-software integrated high wall built by NVIDIA, AMD has been emphasizing "open AI ecosystem" repeatedly from signing the agreement in September to closing the deal in November.
AMD wants to provide developers with another set of options through open hardware, open software and self-developed cutting-edge models.
The competition between chip giants can no longer be decided only by the parameter sheet of a single graphics card.
This article is from the WeChat official account "QbitAI", written by Kressy, authorized for release by 36Kr.