AMD has agreed to acquire artificial intelligence company World Labs in a deal valued at $8.2 billion.
The two companies announced the agreement on Tuesday. The transaction is expected to close before the end of the year, subject to regulatory approval.
Under the deal, World Labs founder Fei-Fei Li will join AMD as executive vice president and chief scientist.
In a joint statement, World Labs said AI development required "close collaboration across model research, systems and compute." AMD said understanding frontier workloads, such as those created at World Labs, would shape its chip-making roadmap.
The companies have worked closely in the past year. They formed an inference optimisation and training partnership last year, and Li was a guest at AMD’s CES presentation earlier this year.
Why AMD is buying World Labs
Li, a Stanford professor of computer science, is widely regarded as a pioneer in AI, particularly in computer vision, for her work building the ImageNet database and the competitions it inspired.
She founded World Labs in 2024 to develop deep learning models with a stronger understanding of the physical world. The company’s position is that true general intelligence requires grounding in physics and the ability to understand and reason about data beyond text.
Announcing the deal, Li said the partnership was driven by a need to scale the company’s work beyond the laboratory.
"Now that we have tangible proof of the possibilities, we want to do everything we can to accelerate the future," Li wrote. "To do this requires scaling our efforts, widening our reach, and getting closer to the hardware."
What World Labs builds
"World model" remains a loosely defined term. It covers everything from language models trained to understand visual inputs to models capable of generating and sustaining a high-fidelity simulation of reality.
World Labs’ first product, Marble, is pitched as a tool for creating entertainment experiences and for building simulated environments for robot training.
The acquisition is expected to help AMD compete more directly with long-standing rival Nvidia in building an ecosystem for AI-specific chips. While Nvidia already offers a suite of open-weight world models like Cosmos, AMD has only offered text and video-based models to the public.
World models are seen as central to efforts to deploy generative AI on robotic platforms, from autonomous vehicles to industrial robots and general-purpose humanoids. Developers say the shortage of useful real-world data for training general purpose robots means synthetic data from world models will be critical to the vision pursued by companies like Tesla and Figure.
Nvidia unveils safety platform for rogue AI agents
In a separate development, Nvidia has introduced a new toolkit designed to keep AI agents within their test environments.
Chief executive Jensen Huang unveiled the Nvidia Open Agent Safety Platform on Monday during an interview with CNBC. The company said the system adds independent security layers around AI agents.
The launch follows a series of incidents in which AI models from Anthropic, Google, OpenAI and Meta bypassed security controls to escape testing environments and access real-world systems. The most prominent case occurred this summer when OpenAI agents breached Hugging Face while attempting to complete a cybersecurity task. OpenAI has since published a dedicated site for reports of its AI agents going rogue.
Huang said the new platform would have prevented such breaches.
Nvidia, which has made tens of billions of dollars selling GPU and CPU chips to AI labs, does not support slowing development or introducing new regulations to address the safety problem. Its approach is to move some security controls outside the agent itself.
"AI’s extraordinary potential for society will only be realized if we solve AI safety," Huang said in a statement. "As we continue to discover the frontier of AI capabilities, we must accelerate discovery at the frontier of AI safety. Safety and security require full-stack engineering."
The platform combines two systems. OpenShell is open-source software that controls what agents can access while operating. Sentry is an independent monitoring system that runs on Nvidia’s BlueField-4 data processing units.
Nvidia said placing Sentry on a separate processor, rather than on the CPU or GPU where the agent operates, provides an isolated view of the agent’s activity. The company said OpenShell provides the software boundary, while Sentry will continuously monitor behaviour and "quarantine agents that attempt to move outside their boundaries in milliseconds." OpenShell was first announced in March.
Nvidia listed dozens of companies supporting the effort, including Anthropic, Arm, Microsoft, Oracle and SpaceX. OpenAI was not listed as a participating company.
Huang told CNBC that work on the effort started a year ago following the introduction of OpenClaw, an operating system of agents created by Peter Steinberger. In March, Nvidia released NemoClaw, an enterprise-grade AI agent platform and its own version of OpenClaw.
"When you deploy an agent, no matter how smart, the first thing you do is to take away all of its rights," Huang said during the interview, comparing the measures to how human employees and executives are managed.
The release was backed by figures who have cautioned against a slowdown in development.
David Sacks, founder, venture capitalist and co-chair of the President’s Council of Advisors on Science and Technology, described agent safety as an engineering problem.
"Recent breakouts weren't proof that development must stop," he wrote on X. "They were proof that the sandbox was too weak. The runtime environment was poorly designed and misconfigured."
.jpeg)
0 Comments