Jensen Huang highlighted several key points:
The next-generation AI super chip is called Rubin, and it has already gone into production. Its performance is 5 times better than the current Blackwell, and inference costs have been cut to one-tenth.
This chip is not for personal computers; it is for data centers, factories, and robots.
Massive shipments in the second half of 2026, computing power will explode soon, and there will be no shortage of computing power in the future.
Open-sourced 3 large models, completely free for developers:
Cosmos World: Teaches AI to understand the physical world (gravity, physical collisions, etc.).
GROOT: Acts as a brain for robots, enabling them to move and grasp objects.
Alpamayo: Designed for autonomous driving, capable of reasoning through complex situations (more reliable than a thirty-year veteran driver).
This means: in the future, when making robots, self-driving cars, and industrial arms, there's no need to write code from scratch; just modify existing code to use (the threshold has been significantly lowered).
The Mercedes CLA will hit the roads in the US next year, using NVIDIA's full stack + Alpamayo to achieve autonomous driving on complex urban roads.
The biggest anxiety is not chips, but electricity.
He himself said that the currency of the future is watts; training models, running AI, and building data centers all consume electricity. The grid, transformers, and copper wires are all bottlenecking (electricity is still the strongest in Dongda, far ahead of others).
The core of Jensen Huang's speech is actually just one sentence:
From now on, the AI that simply chats will upgrade to a physical AI that can move, work, understand physical rules, and reason actions.
After watching, I believe the things truly worth investing in the future.
Electricity and the grid: AI + robots are power-hungry monsters, the grid, transformers, copper mines, and silver mines are the most desirable.
Copper: data centers, electric vehicles, robots, and photovoltaics are all scrambling for copper, with supply unable to keep up.
Gold: the more chaotic the world, the more desirable it becomes; central banks are already buying frantically.
Rare earths and lithium: AI chips, batteries, wind power, and solar energy are indispensable, with obvious supply bottlenecks.
In the future, whoever masters energy, computing power, and key resources.
Whoever can stand at the infrastructure layer of the AI world.