G42 and Analog Back Yann LeCun's AMI Labs for Physical World Models
Abu Dhabi's AI giants are betting on Yann LeCun's vision of world models over chatbots to give AI physical common sense.
Models · Source: Mid-East.info
What happened
Abu Dhabi technology group G42 and its portfolio company Analog signed non-binding agreements with Advanced Machine Intelligence. AMI Labs is a Paris-based frontier AI startup co-founded by Turing Award laureate Yann LeCun. The companies will collaborate on AI research, computing infrastructure, and sovereign-enabled cloud services. They want to build world models that understand physical environments instead of just predicting the next word in a text prompt.
The agreements were signed during the UAE-France Business Dialogue in Abu Dhabi. French President Emmanuel Macron and Abu Dhabi Crown Prince Sheikh Khaled bin Mohamed bin Zayed Al Nahyan attended the high-profile event. The proposed partnership combines the deep research capabilities of AMI Labs with the massive computing power of G42. It also integrates the physical intelligence platform built by Analog. Analog is led by Alex Kipman, the inventor of Microsoft HoloLens and Kinect, who moved to Abu Dhabi to launch the company.
This deal deepens the growing technology ties between the United Arab Emirates and France. G42 already operates a dedicated AI data center in Grenoble. The Abu Dhabi group previously formed a strategic partnership with French AI startup Mistral in May 2025. The current agreements with AMI Labs provide a framework to exchange information and define shared objectives. Future commercial terms, scope, and specific industrial projects will require separate definitive contracts.
Key facts
- May 2025 — When G42 formalized its strategic partnership with Mistral AI
- Grenoble — Location of G42's operational AI data center in France
- Alex Kipman — Analog founder and CEO who invented Microsoft HoloLens
- Alexandre Lebrun — Co-founder and CEO of Advanced Machine Intelligence
Why it matters
The industry is slowly shifting focus from pure language models to physical intelligence. Chatbots lack common sense because they only learn from static text on the internet. World models learn by observing physical environments to predict outcomes, reason, and plan actions in real time. Builders focusing on robotics, industrial automation, and smart cities now have a clear signal that sovereign wealth is funding the infrastructure for physical AI. This moves the frontier of AI away from browser windows and into the physical world.
This partnership merges massive computing scale with top-tier research talent across borders. The UAE has the capital, energy resources, and ambition to build massive data centers. France has deep pools of machine learning researchers and a growing ecosystem of AI startups. As these two nations lock in strategic partnerships, startups building physical AI applications will likely need to align with these sovereign cloud ecosystems. Companies that fail to secure access to this localized compute power will struggle to train data-heavy world models.
For builders
Industrial automation applications
The MOUs specifically target industrial applications for world models. Founders building software for manufacturing, logistics, or autonomous infrastructure can leverage these incoming physical intelligence platforms. Heavy industries will pay a premium for AI that actually understands physical constraints and reduces operational downtime.
Sovereign cloud deployment
G42 is pushing its sovereign-enabled cloud infrastructure through its Core42 subsidiary. Startups handling sensitive physical data for governments or critical infrastructure will need to deploy on these localized clouds. Companies relying solely on US-based hyperscalers might lose out on lucrative Middle Eastern and European government contracts.
Hardware and vision sensors
World models require massive amounts of real-world observation data to learn physics and common sense. Hardware engineers building better vision systems and edge sensors have a massive opportunity here. The companies funding these models will pay heavily for high-quality physical data pipelines that feed their learning algorithms.
My take
I have said it before and I will say it again. You cannot reach artificial general intelligence just by feeding the internet into a text box. LeCun is right to bet on world models that learn physics, and Abu Dhabi is smart to bankroll the compute required to build them.
Original reporting: Mid-East.info. This is my rewrite and opinion.