In this exclusive interview by Cleo Abram a youtuber, we sit down with Jensen Huang, CEO of NVIDIA, a company that has redefined computing and is now at the forefront of AI, robotics, and beyond. From the invention of the GPU to the promise of a robotic future, Huang shares his vision for how technology will shape our lives in the coming decades. Conducted by Cleo, this conversation dives into the past, present, and future of NVIDIA’s groundbreaking work, offering insights for optimists and innovators eager to build a better tomorrow.
Introduction: A Visionary Shaping Our Future
Jensen Huang, the CEO of NVIDIA, is no stranger to making bold bets that redefine technology. Under his leadership, NVIDIA has transformed from a gaming graphics company into a global powerhouse driving advancements in AI, robotics, self-driving cars, and medical research. In this interview, we explore the key insights that led to NVIDIA’s dominance, the current state of computing, and Huang’s ambitious vision for a future where “everything that moves will be robotic.” With a focus on optimism and accessibility, this conversation aims to inspire our audience—whether they’re experts, students, or curious 12-year-olds dreaming of becoming the next Huang.
Part 1: How Did We Get Here? The Birth of the GPU
Cleo: Let’s start in the 1990s with video games. Developers wanted more realistic graphics, but the hardware couldn’t keep up. NVIDIA’s solution changed not just gaming but computing itself. Can you take us back and explain the insights that led to the first modern GPU?
Jensen Huang: In the early ’90s, we noticed that in any software program, about 10% of the code does 99% of the processing, and that 99% could be done in parallel. The other 90% needed sequential processing. We realized the perfect computer would handle both. That was the big observation. We set out to solve problems that normal computers couldn’t, and that’s how NVIDIA began.
Cleo: I love the Mythbusters video on NVIDIA’s YouTube channel showing a robot shooting paintballs one by one for sequential processing (like a CPU) versus all at once for parallel processing (like a GPU). Why focus on gaming first?
Huang: Video games required parallel processing for 3D graphics, and we loved the idea of simulating virtual worlds. We also saw gaming as potentially the largest entertainment market ever, which it became. A large market meant we could invest in R&D, creating a flywheel of technology and market growth that made NVIDIA a leading tech company.
Cleo: You’ve called GPUs a “time machine.” What do you mean by that?
Huang: A GPU lets you see the future sooner. A quantum chemistry scientist once told me, “Because of NVIDIA, I can do my life’s work in my lifetime.” That’s time travel. GPUs make applications run so much faster—whether it’s weather prediction or simulating self-driving cars—that you’re essentially seeing the future.
Part 2: What’s Happening Now? From CUDA to AI
Cleo: By the early 2000s, GPUs were revolutionizing industries beyond gaming, but researchers had to trick GPUs into thinking their problems were graphics-related. You created CUDA to make parallel processing more accessible. What was the vision behind CUDA?
Huang: CUDA came from a mix of inspiration and desperation. Researchers, like those at Mass General using GPUs for CT reconstruction, showed us the potential. Internally, we wanted dynamic virtual worlds—water flowing, explosions acting realistically—which required more than graphics. We knew our GPUs, driven by the gaming market, would be the highest-volume parallel processors. CUDA let programmers use languages like C to tap into that power, making it accessible to everyone.
Cleo: In 2012, AlexNet, a neural network trained on NVIDIA GPUs, crushed an image recognition competition, kicking off the AI revolution. What was that moment like for you?
Huang: AlexNet was a turning point. We’d been struggling with computer vision internally, and when we saw AlexNet’s leap, we asked, “How far can this go?” We reasoned that if deep learning could scale, it could solve vast problems, potentially reshaping the computer industry. That led us to reinvent the computing stack with systems like DGX. After 65 years of general-purpose computing, we’re redefining how computers work.
Cleo: It took a decade from AlexNet to AI’s mainstream moment. Why so long, and what was that decade like for you?
Huang: It felt like today—always some problem, always a reason to be impatient, but also reasons to keep going. We reasoned from first principles: parallel computing and deep learning’s scalability were empirically true. We believed in it, even when others didn’t. We invested tens of billions, and though it was tough—especially when investors wanted profits over R&D—it was fun. Our core beliefs never wavered, and evidence of success kept us committed.
Part 3: What’s Next? A Robotic Future
Cleo: You’ve said, “Everything that moves will be robotic someday, and it will be soon.” NVIDIA’s Omniverse and Cosmos are enabling this by training robots in digital worlds. Can you explain how these tools work and what they mean for robotics?
Huang: Think of ChatGPT: it generates text but can hallucinate without grounding. Ground it with a PDF or search, and it’s truthful. For robotics, we need a “world model” like Cosmosastanza: Cosmos is that model, encoding physical common sense—gravity, friction, object permanence. Omniverse, a physics-based simulator, grounds Cosmos with Newtonian physics. Together, they generate physically plausible futures, letting robots learn in digital worlds—faster, safer, and under varied conditions like lighting or obstacles.
Cleo: What will this mean for our daily lives in 10 years?
Huang: Everything that moves will be robotic—cars, lawnmowers, humanoid robots. They’ll learn in Omniverse-Cosmos and act in the physical world. Imagine your own R2-D2, with you in your glasses, phone, car, or home, growing up with you. It’s a certainty, and it’s exciting.
Cleo: What challenges must we overcome to ensure this future is safe?
Huang: We need to address bias, toxicity, hallucination, and impersonation. For physical AI, like self-driving cars, it’s about engineering robust systems—redundancy, like triple flight computers in planes, and community-wide safety systems. It’s a big engineering challenge, but we’re working on it.
Part 4: The Future of Computing and Humanity
Cleo: What technological limits are you thinking about now?
Huang: It’s all about energy efficiency. Flipping and transporting bits takes energy, and that’s the physical limit. We’ve improved AI computing efficiency 10,000 times since 2016—imagine a light bulb using 1/10,000th the power. Energy efficiency is our priority to build smarter systems.
Cleo: As AI models like transformers become specialized, how do you balance designing specific versus general-purpose chips?
Huang: We believe transformers are a stepping stone, not the final architecture. History shows no single algorithm lasts forever. We design flexible architectures to let researchers innovate, like the attention mechanism evolving into flash or hierarchical attention. Flexibility is key.
Cleo: How do you push the limits of what’s physically possible in chip design?
Huang: We build deep expertise in-house—semiconductor physics, aerodynamics for cooling, plumbing for liquid cooling. We work closely with partners like TSMC to push limits, assuming we need their level of knowledge to innovate.
Cleo: What new bets is NVIDIA making now?
Huang: We’re fusing Omniverse and Cosmos for generative world systems, critical for robotics. We’re advancing humanoid robotics, digital biology to understand molecules and cells, and climate science for high-resolution weather prediction. These time machines let us predict and optimize the future.
Advice for the Future
Cleo: For our audience—students, optimists, future innovators—how should they prepare for this AI-driven world?
Huang: Learn to interact with AI like ChatGPT, Gemini Pro, or Grok. Prompting AI is an art, like asking great questions. Ask, “How can AI make me a better [lawyer, doctor, chemist]?” AI lowers barriers to knowledge and empowers you. Get an AI tutor—it’ll make you superhuman.
Cleo: Finally, what legacy do you hope NVIDIA leaves?
Huang: I hope people say we made an extraordinary impact. We’re at the epicenter of transforming digital biology, material sciences, robotics, and autonomous vehicles. We want our technology to empower everyone—big companies, small researchers, profitable or not. The next generation will know us for gaming and for revolutionizing their world.
Conclusion: A Call to Build the Future
Jensen Huang’s vision is clear: a world where computing empowers us to become superhuman, where robots assist us, and where AI unlocks new possibilities across industries. For our audience, the message is equally clear—embrace AI, ask how it can enhance your work, and join the optimistic journey to build a better future. As Huang says, “If you don’t build it, they can’t come.” Let’s build it together.