NVIDIA’s GTC event, particularly Jensen Huang’s keynote, should be mandatory watching for anyone interested in working for, buying or building tech. The event is as well orchestrated as an old Steve Jobs event, and while Huang isn’t as charismatic as Jobs was (few are), the content is less about pitching products and more about pitching a very different future.
At these events over time, I’ve seen the future of drones, robots, autonomous cars, AI simulation (up to world scale), gaming and video conferencing, though this last, which used photorealistic digital background built from components, was unintended.
This month’s event was no exception. I’ll focus on what was interesting.
The Birth of Next Generation AI
The keynote opens up how AI can be used to create predictive models, views into the future, turn knowledge into what looks to me like magic, improve our knowledge of the human body to make us healthier, protect critically endangered wildlife, turn ideas into positive scientific and social revolutions, help make people better both physically and mentally, and take that leap to where no one has gone before.
As a background, Huang used his headquarters building in simulation, and as many know, that building was first fully simulated before it was built (something everyone should do). NVIDIA has been so successful with AI and artificial intelligence it can’t build its graphics cards fast enough, and its Blackwell-based 5090 cards are sold out worldwide.
This event has grown from one of the smallest of its type in the industry to one of the largest, arguably growing beyond the capabilities of the resources in Silicon Valley. And this isn’t the only thing pushing resources. According to Huang, a year ago we were 100x off in terms of the resources needed to run AI at scale largely because a year ago we weren’t as focused on quality as we are now. DeepSeek was one of the big examples of this shift, and it caught the world sleeping.
NVIDIA’s Blackwell processor would have seemed like an insane product to create two years ago because the performance requirements back then seemed more than adequate. But even Blackwell may not be enough to meet the emergent needs of AI users, and no one else is close to providing a Blackwell solution let alone something stronger. Part of this is because AI has moved from something people create to something that is created by AI. This is already forcing a massive change in datacenter design. That is the basis of the announcement at the event of Blackwell Ultra and advances in photonics that increase Blackwell’s capabilities to address these new and growing needs.
Huang spoke to targeted AI from manufacturing to automobiles.
NVIDIA Halos
This is basically a logical shield for automobiles, creating a protective layer between the car and the environment around it. When it comes to autonomous driving, NVIDIA is well ahead of everyone else in terms of safe performance. With Omniverse being the safest and more comprehensive simulation, to Cosmos, a learning model where a teacher is created, and then subsequent tuning automatically optimizes the results. This could be particularly useful for fringe events like extreme weather shifts, black ice or occluded visibility which have all been problematic to this technology.
I think future cars should tell you what tech is in them so you can better trust the result with your family.
Robotics
This is one of the segments that I personally find the most interesting. NVIDIA has been at the forefront of this. It’s ironic that the company that first argued that robots would eventually outstrip both PC and smartphones was Dell Technologies which still has made no move toward this massive opportunity. NVIDIA has been on top of this from the start with Omniverse training so it can, like autonomous cars, be safely trained for even extreme fringe events, making them both safe and performant with more and more capabilities, greater safety, and massively improved performance.
This next generation of robots will be able to work more as a system and learn real-time as they operate. They announced Mega, which tests robots at scale, and Isaac GROOT N1, which is inspired by human cognitive processing, making these robots vastly easier to spin up and apply to new environments.
And they closed with an update on a Disney Star Wars robot that pretty much works like the simulations in the movies. They announced Newton, a collaboration between Deep Mind, Disney Research and NVIDIA. If they can get the cost of this thing down to something affordable, I can see variants in most every home, particularly those that don’t allow pets.
Wrapping Up:
NVIDIA GTC is where you go to see the future of computing, but increasingly, it is where you go to see the future of the developed world. From pervasive AIs to pervasive robotics to simulation and everything in between, if you want to see the future, check out the GTC keynote. It doesn’t cost you anything but your perception of what is possible and how amazing the future will become. Particularly during times like this, a bit of “amazing” can help restore hope for a better tomorrow.
