Two years ago Fei-Fei Li and three fellow researchers started World Labs with $230 million and one bet: the next wave of AI would need models that predict the physical world.
Yesterday AMD agreed to buy that company for $8.2 billion.
Li becomes AMD's Executive Vice President and Chief Scientist, and her co-founders Justin Johnson and Ben Mildenhall keep leading the World Labs team inside AMD.
The price tracks what world models consume.
Marble, the first public tool out of that team, generates small 3D spaces from Gaussian splats that export straight into film and game pipelines.
Underneath the demos sit models trained on vast amounts of video.
AMD keeps pointing at robotics: synthetic data generated in simulation instead of gathered in the real world.
Here is what matters if you run a smaller team.
The research is public.
The barrier left is GPU time.
A robotics or vision team testing a new simulation pipeline does not need frontier lab money.
It needs a few hundred GPU hours and a weekend.
That gap is why we built BHK Cloud.
RTX 3090 compute at $0.15/hr, cloud storage at $2.49/TB per month, and Buy Now Pay Later so you start with zero upfront and pay after one month.
If you had a weekend of GPU time for a simulation project, what would you point it at first?
https://ai.bhkcloud.com/?utm_source=linkedin&utm_medium=social&utm_campaign=bhk_social_2026w40&utm_content=morning
Originally posted on linkedin_personal