Tuesday saw the event that a good chunk of the tech world anticipates every March and November when the founder and CEO of NVIDIA ((NVDA Quick QuoteNVDA - Free Report) ), Jensen Huang, takes the stage for his power-packed keynote.

The NVIDIA GPU Tech Conference (GTC) is the week-long computing/AI event where thousands of developers, innovators, and business leaders converge to experience how AI and accelerated computing are helping humanity solve the most complex challenges across industry, science, economics, and society.

Best of all, attendees and observers are never disappointed.

That's because Jensen always delivers massive delight and surprises for the enterprise, developer, and science crowds.

And Tuesday was no exception.

Here's a hit list of some of the magic that Wizard Huang released from his magic shoppe!

1) Blackwell GPUs: Full Production & Shipping

Jensen showed a chart that made it clear: 1.3 million Hopper GPUs sold in 2024 to the top 4 CSPs (cloud service providers), while this year has seen 3.6 million Blackwell GPUs already ordered so far.

He also said something he might regret. After introducing a new acceleration system called Dynamo -- that makes Blackwell 40X more powerful than Hopper -- he basically said something like "Why would you want Hopper anymore?"

Then he called himself the "Chief Revenue Destroyer."

"My sales guys are gonna kill me... but the more you buy, the more you save."

My take: But Hopper will still sell because it is still highly useful, cheaper, and immediately available. And it's still state of the art for what any enterprise needs vs what was available 2 years ago to build an LLM.

2) The Tipping Point to Exceed Apple's Sales

As I wrote about in February, the top hyperscalers have only increased their datacenter capex projections for this year.

Jensen showed a stunning chart of NVIDIA revenue scaled against past and projected datacenter (DC) capex. This is going to happen because of a new phenomenon that he calls the "AI Factory."

Every enterprise and manufacturer will need two types of factories to compete: a production factory and an "AI factory" that provides the intelligence for the former because "they manufacture intelligence at scale, transforming raw data into real-time insights."

This is why NVIDIA is bent on maintaining a "One-Year Rhythm" with new and more powerful GPUs. Jensen teased a little more about Blackwell Ultra (coming second half 2025) and the Rubin line due in 2026 because he wants his customers -- across enterprise, industry, and science -- to be able to plan ahead. This cadence will dwarf iPhone sales.

My take: If you reason that $1 trillion of old-line CPU-based DCs need to upgrade to GPU-driven acceleration, then NVIDIA sales could likely hit over $500 billion in five years -- topping Apple's estimated revenue for next year.

3) Here Come the Humanoid Robots

As described in my 5th Industrial Revolution article, Jensen has a roadmap to what he calls "Physical AI."

To make robots safe, highly efficient, and worth their cost, we need to build a complex, data-intensive training environment.

Hundreds of humanoid robotics startups are building on the NVIDIA hardware and software platforms that give them nearly unlimited tools and libraries to develop and iterate -- all open source with Isaac GR00T.

Why? Because Jensen understands that robots are inevitable in our daily lives and could become the largest industry on the planet. So he wants to make them "embodied intelligence" that knows how to work with and care for humans.

4) NVIDIA Accelerated Quantum Research Center to Bring Quantum Computing Closer

NVIDIA GPUs are not intended to replace traditional CPUs. In fact, they are designed to work together in many instances, as the Grace Blackwell "super-chips" demonstrate.

QPUs, or Quantum Processing Units, are the next frontier and Jensen will discuss those integrations on Thursday with the world's top quantum computing experts and companies.

5) NVIDIA Unveils AI-Q Blueprint to Connect AI Agents for the Future of Work

I wrote about the coming wave of Agentic AI in my 5th Industrial Revolution article where I explained that even you and I will have a dozen agents working for us in this decade.

Jensen said today that "100% of my engineers will be AI-assisted this year."

From the press release...

"AI agents are the new digital workforce, transforming business operations, automating complex tasks and unlocking new efficiencies. Now, with the ability to collaborate, these agents can work together to solve complex problems and drive even greater impact."

6) Self-Driving Cars: Omniverse & Cosmos Will Train Autonomous Vehicles Like Special Forces

What I mean is that these two NVIDIA platforms take real-world data and use it to not only train autonomous vehicle intelligence on trillions of parameters, but also to simulate those environments with synthetic data and then run billions of reinforcement training iterations of the problems they are expected to encounter and solve in the real world of pedestrians, potholes, and panic.

In other words, AI can begin to train AI relentlessly. Just like our top military operators train.

7) Exploding Frontiers of Science

This is by far my favorite application of NVIDIA accelerated computing. Because the world-changers doing research and experiments in cancer, climate, and chemistry can accelerate their work by orders of magnitude with the right data tools.

And Jensen shared a quote from one scientist that summed it all up...

"Because of your work, I can do my life's work... in my lifetime."

Vera Rubin's Name on Galactic-Class Telescopes and GPUs

Which brings me to a #8 bonus: Vera Rubin's grandchildren. If you follow me on X @KevinBCook, you may know I'm writing a book titled "13 Great Women of Science." 

Astronomer Vera Rubin (1928-2016) is #8 in the book (coincidence here, but true). She shares a story with the other twelve women of trying to do science in a man's world: their work was often ignored, sabotaged, or stolen.

And most never got the credit they deserve. What did Rubin discover? She studied galactic superclusters and rotation curves and her work on the galaxy rotation problem is cited by others as the first direct evidence for the existence of dark matter, initially proposed by Fritz Zwicky in the 1930s.

While she has received the honor of having a new powerful telescope in Chile named after her, many in the scientific community believe she deserved a Nobel prize. She had achieved most of her important work several decades before her death, and as you may know, the Nobel is not awarded posthumously.

In 1996 she received the Gold Medal of the Royal Astronomical Society, the second woman to be so honored, 168 years after Caroline Herschel received the Medal in 1828. In 2002, Discover magazine recognized Rubin as one of the 50 most important women in science. 

The irony of discovering what no one else could see: dark matter. But that's what Jensen likes to ever-so-subtly point out when he can. 

Just like when he named his best-selling Hopper GPUs after Grace Hopper, an American computer scientist, mathematician, and United States Navy rear admiral who was a pioneer of computer programming. Hopper was the first to devise the theory of machine-independent programming languages, which she used to develop COBOL.

And that's why Vera Rubin's grandchildren were in the audience on Tuesday.

May the Obi-Huang Be With You.

Cooker