Not That Keynote—Jensen Style
Published March 27, 2025
Global Head of Research
Key Takeaways
- At Nvidia’s GTC 2025, CEO Jensen Huang unveiled the Blackwell Ultra GPU, 40x more powerful than its predecessor, signaling a leap in AI compute capabilities and rendering last-gen hardware nearly obsolete.
- With AI data centers projected to surpass $1 trillion in buildout costs annually by 2028, Nvidia’s CUDA ecosystem is cementing its dominance by powering everything from inference computing to quantum simulation and autonomous driving.
- Nvidia isn’t just advancing artificial intelligence; it’s redefining it with “Physical AI,” a bold new frontier where humanoid robots and AI-driven systems interact with the real world, making science fiction increasingly real.
On March 18, 2025, Jensen Huang stepped onto the stage at Nvidia's GPU Technology Conference (GTC) in San Jose with the presence of a man who knows he's about to redefine the trajectory of an industry—again. His message was clear: AI has hit an inflection point, and if you're not paying attention, you're already behind.
Some keynote speeches deliver incremental updates, a fresh coat of paint on last year's ideas. This was not that kind of keynote.
This was a declaration that the future of computing—AI, data centers, robotics, autonomous systems—is accelerating at a pace that's borderline uncomfortable. And Nvidia, with its new Blackwell Ultra GPUs and Rubin chip architecture, is the engine at the heart of this transformation.
Blackwell Ultra: More Power, Less Patience
We are trained to expect gradual improvements in computing power—a bit more performance, a bit less power draw, some new buzzwords about efficiency. But every once in a while, a breakthrough happens that obliterates expectations.
Enter Blackwell Ultra.
Huang didn't just claim it was more powerful than its predecessor, Hopper. He declared that it is 40 times more powerful. Not 40%—40x.1
That's not a speed bump; that's a paradigm shift.
Huang, never one to shy away from showmanship, quipped that he is Nvidia's "Chief Revenue Destroyer," acknowledging that this leap forward makes last-gen hardware practically obsolete overnight.
For those who think this is the endgame, buckle up.
- Vera Rubin (2026): The next iteration in AI compute is already mapped out.
- Rubin Ultra (2027): A 576-GPU behemoth, ensuring Nvidia stays far ahead of the competition.
When companies make long-term AI investments, they now have to ask themselves a simple question: If we're not on Nvidia's road map, do we even have a road map?
AI Factories: The $1 Trillion Data Center Arms Race
Huang didn't just talk about chips. He talked about the world those chips are going to create.
In his words, computing has shifted from retrieving files to generating tokens—data centers are no longer just storage hubs; they are AI factories.
And these AI factories? They're going to cost a fortune.
Huang predicted that the data center buildout required for AI could exceed $1 trillion in the coming years.2 That's an eye-watering number until you consider the scale of what's happening:
- AI isn't just getting better—it's consuming more.
- Inference computing (where AI models generate outputs based on trained knowledge) is set to increase token usage by 100x.3
- And with Nvidia's proprietary CUDA ecosystem, they're making sure that every cycle of that growth runs through their hardware.
Huang made a bold claim: "We've reached the tipping point of computing—CUDA made it possible."
For those who track the industry, this isn't an exaggeration.
CUDA isn't just a technology—it's Nvidia's invisible empire.
It's the reason competitors struggle to keep up. It's why Nvidia's GPUs aren't just faster; they're smarter. The CUDA-X ecosystem, packed with domain-specific libraries, accelerates computation in everything from quantum computing to gene sequencing.
This is a classic Nvidia move—don't just sell hardware; build the software layer that locks developers into your ecosystem.
The result? If you want to build the next wave of AI models, advanced medical imaging or hyper-realistic simulations, you're almost certainly doing it on Nvidia's terms.
AI Is Driving Itself—Literally
Huang has been talking about autonomous driving for years. In 2025, it's no longer just talk.
- Nvidia announced a major partnership with General Motors to power self-driving vehicles.4
- They unveiled HALO, a comprehensive AI-driven safety system for autonomous cars.
- Huang confidently stated: "Almost every self-driving car company uses Nvidia tech." If that sounds like an exaggeration, look at the players: Tesla, Waymo, GM and more.
And here's the kicker:
"We're the first company in the world, I believe, to have every line of code safety assessed," Huang claimed.
That's a bold statement. It could possibly put Nvidia in a regulatory fast lane, letting them leap ahead of competitors who are still grappling with safety compliance.
Beyond AGI: The Birth of Physical AI
Most of the AI discussion today is about AGI (artificial general intelligence)—the idea that AI could eventually reason, learn and act like a human.
Huang is already past that conversation.
He introduced Physical AI, an emerging field where AI moves beyond digital tasks and starts interacting with the real world.
The implications are massive:
- Humanoid robots: AI that can power physical bodies, not just chatbots.
- AI NPCs: Non-playable characters in games that don't just follow scripts—they think.
- Robots training AI models: Instead of humans fine-tuning AI, Nvidia wants robots to train AI models themselves.
The world we imagined for 2050? Huang is saying it's coming a lot sooner than that.
Final Thought: Nvidia Isn't Chasing the Future—It's Dictating It
Some companies follow trends. Others create them.
Nvidia is in a category of its own—it dictates the pace of innovation.
This isn't just about faster GPUs or better AI models. It's about rewriting the rules of computing—and, by extension, every industry that relies on computing.
Huang's GTC 2025 keynote wasn't a collection of product updates. It was a vision statement, a reminder that the real AI revolution isn't coming in the future—it's happening right now.
The only question left is: Are you moving fast enough to keep up?
1 Source: 2025 GTC Keynote Address by Jensen Huang, verified here: https://www.investopedia.com/nvidia-gtc-2025-jensen-huang-keynote-live-updates-11699070
2 Source: 2025 GTC Keynote Address by Jensen Huang, verified here: https://blogs.nvidia.com/blog/nvidia-keynote-at-gtc-2025-ai-news-live-updates
3 Source: 2025 GTC Keynote Address by Jensen Huang, verified here: https://www.tomsguide.com/computing/live/nvidia-gtc-2025-live
4 Source: 2025 GTC Keynote Address by Jensen Huang, verified here: https://nvidianews.nvidia.com/news/general-motors-and-nvidia-collaborate-on-ai-for-next-generation-vehicle-experience-and-manufacturing
Categories
About the contributor
Global Head of Research
Christopher Gannatti began at WisdomTree as a Research Analyst in December 2010, working directly with Jeremy Schwartz, CFA®, Director of Research. In January of 2014, he was promoted to Associate Director of Research where he was responsible to lead different groups of analysts and strategists within the broader Research team at WisdomTree. In February of 2018, Christopher was promoted to Head of Research, Europe, where he was based out of WisdomTree’s London office and was responsible for the full WisdomTree research effort within the European market, as well as supporting the UCITs platform globally. In November 2021, Christopher was promoted to Global Head of Research, now responsible for numerous communications on investment strategy globally, particularly in the thematic equity space. Christopher came to WisdomTree from Lord Abbett, where he worked for four and a half years as a Regional Consultant. He received his MBA in Quantitative Finance, Accounting, and Economics from NYU’s Stern School of Business in 2010, and he received his bachelor’s degree from Colgate University in Economics in 2006. Christopher is a holder of the Chartered Financial Analyst Designation.