Nvidia's Jensen Huang on CUDA's Future, AI Agents, and the Computing Revolution

Nvidia CEO Jensen Huang sat down with Ben Thompson for an extensive Q&A covering the company's strategic direction, competitive landscape, and the future of computing. The conversation touched on everything from Nvidia's massive sales projections to the philosophical underpinnings of the company's approach to technology development.

Nvidia's $1 Trillion AI Chip Forecast

Huang revealed that Nvidia expects to generate over $1 trillion in sales from its flagship AI chips through the end of 2027, doubling the company's previous forecast of $500 billion by 2026's end. This dramatic revision reflects the explosive growth in AI computing demand and Nvidia's dominant position in the market.

The Central Role of CUDA

At the heart of Nvidia's strategy is CUDA (Compute Unified Device Architecture), the company's parallel computing platform and programming model. Huang emphasized that CUDA has become the "core" of modern computing, enabling developers to harness the power of GPUs for general-purpose processing tasks beyond graphics.

"CUDA is not just a software layer," Huang explained. "It's become the foundation for accelerated computing. Every major cloud provider, every AI research lab, every autonomous vehicle company—they're all building on CUDA."

This ecosystem lock-in has been crucial to Nvidia's success. By making CUDA the standard for GPU programming, Nvidia has created a moat that competitors struggle to cross. Even companies like Intel and AMD, which have their own GPU offerings, find themselves playing catch-up in terms of software ecosystem and developer mindshare.

The Evolution of Accelerated Computing

Huang argued that we're witnessing a fundamental shift in how computers work. Traditional CPU-centric computing is giving way to accelerated computing, where specialized processors handle specific workloads more efficiently.

"The CPU is becoming less relevant," Huang stated bluntly. "For AI training, inference, scientific computing, even some aspects of gaming—the GPU is where the action is. The CPU's role is increasingly about orchestration rather than computation."

This perspective helps explain Nvidia's aggressive push into data center markets and its development of integrated systems like the Vera Rubin NVL72 rack. By controlling both the hardware and software stack, Nvidia can optimize performance in ways that traditional CPU vendors cannot match.

Groq and the Inference Challenge

When asked about Groq, the AI inference company that has gained attention for its fast language model processing, Huang acknowledged the competition but framed it as part of a healthy ecosystem. Nvidia recently announced the Groq 3 LPX inference server rack, featuring 256 Groq 3 LPUs (Language Processing Units) and 128GB of on-chip SRAM.

"Groq is focused on inference, which is a specific but important part of the AI pipeline," Huang said. "We see inference as complementary to our training capabilities. The market is big enough for multiple approaches."

However, Huang emphasized that Nvidia's integrated approach—combining GPUs, networking, and software—provides advantages that specialized inference chips cannot match for complex, multi-stage AI workflows.

China Market and Geopolitical Challenges

The conversation turned to one of Nvidia's most pressing challenges: navigating the complex relationship with China. With export controls limiting access to the company's most advanced chips, Huang acknowledged the difficulty but remained optimistic.

"We've developed variants of our chips that comply with export regulations," he explained. "The demand in China remains strong, and we're finding ways to serve that market while adhering to government requirements."

This balancing act has become increasingly difficult as tensions between the US and China escalate. Some analysts worry that Nvidia could become collateral damage in the broader tech cold war, but Huang's comments suggest the company is preparing for multiple scenarios.

The "Doomers" and AI Ethics

Huang addressed concerns from those who worry about AI's potential negative impacts, often called "doomers" in tech circles. He took a pragmatic stance, acknowledging legitimate concerns while emphasizing the benefits of AI advancement.

"Every transformative technology has faced skepticism and fear," Huang noted. "The key is responsible development and deployment. We're working with researchers, ethicists, and policymakers to ensure AI develops in ways that benefit society."

This response reflects Nvidia's position as a foundational technology provider rather than an AI application company. By focusing on the infrastructure layer, Nvidia can argue that it's enabling progress while leaving ethical decisions to those building applications.

Nvidia's "Nature" and Corporate Philosophy

One of the most revealing parts of the interview explored Huang's philosophy about Nvidia's corporate identity. He described the company as having a "nature" that drives its decisions and culture.

"Nvidia is fundamentally about solving hard computational problems," Huang said. "That's in our DNA. Whether it's graphics, AI, scientific computing, or autonomous vehicles—we're drawn to challenges that require massive computational power."

This philosophy explains Nvidia's willingness to enter new markets and its patience with long-term R&D investments. The company has repeatedly demonstrated that it's willing to cannibalize its own products if doing so serves the broader mission of advancing computing.

The Future of Computing

Looking ahead, Huang painted a picture of computing that's increasingly distributed, specialized, and AI-driven. He mentioned Nvidia's work on space-based data centers and orbital computing as examples of how the company is thinking beyond traditional data center models.

"The next decade will see computing move to where the data is generated," Huang predicted. "Edge devices, autonomous vehicles, even satellites—they'll all need AI processing capabilities. The centralized data center model will evolve into a distributed computing fabric."

This vision aligns with Nvidia's current product roadmap and explains its investments in areas like automotive AI, robotics, and edge computing platforms.

Competitive Landscape

While Nvidia dominates the AI chip market, Huang acknowledged the competitive pressures from Intel, AMD, and cloud providers developing their own silicon. However, he argued that Nvidia's integrated approach provides advantages that pure hardware competitors cannot match.

"The competition is healthy," Huang said. "It drives innovation. But what many competitors miss is that this isn't just about the chip anymore. It's about the entire stack—hardware, software, networking, and the ecosystem around it."

This perspective helps explain why Nvidia has invested so heavily in software development and why it continues to open new facilities and expand its workforce despite already being the dominant player in its core markets.

Conclusion

The interview with Jensen Huang reveals a company that's confident in its position but not complacent about the future. Nvidia sees itself as the steward of a computing revolution, with CUDA as the foundation and accelerated computing as the new paradigm.

As AI continues to transform industries and create new possibilities, Nvidia's role as the infrastructure provider puts it at the center of this transformation. Whether through its $1 trillion sales forecast, its approach to competition, or its vision for distributed computing, Nvidia is positioning itself not just as a chip company, but as the architect of the next era of computing.

For developers, researchers, and businesses building on AI, Huang's comments suggest that the Nvidia ecosystem will only become more central to their work in the coming years. The question isn't whether to use Nvidia technology, but how to best leverage it as AI capabilities continue to expand.

The full interview with Ben Thompson provides even more depth on these topics and offers valuable insights into how one of tech's most influential leaders thinks about the future of computing.