NVIDIA's CEO, Jensen Huang, has posted a rather interesting blog post about the state of the AI industry from a much broader perspective, and he has summarized it in a "five-layer" cake.
Jensen Has Basically Summed Up AI Into a Five-Layer Cake, Claiming that the Opportunity Still Hasn't Been Realized
There is no doubt that Team Green is one of the biggest catalysts of the ongoing AI infrastructure buildout, given that the company not only provides essential compute resources but also several other utilities to the world that we'll discuss ahead. As we move towards GTC 2026, NVIDIA's CEO recently posted a blog titled "AI is a 5-Layer Cake", and the main idea being conveyed in a lengthy post is that AI as a technology is now moving towards a front that will be influential and massive. According to Jensen, neither the workforce nor businesses are factoring in how large AI will actually become.
We are still early. Much of the infrastructure does not yet exist. Much of the workforce has not yet been trained. Much of the opportunity has not yet been realized.
- NVIDIA's CEO
The distribution of individual layers focuses on one key aspect: each is fundamentally evolving and dependent on the others. Jensen categorizes "energy" as the bottom layer of AI, calling it a "binding constraint" for intelligence/$ figures. Of course, without electricity, you cannot power the infrastructure, which is why there is no "AI cake" without energy. The next layers of this cake are chips and infrastructure, and while both of these overlap to some extent, Jensen says the latter is more about power delivery, cooling, construction, and networking.
NVIDIA is known for providing the hardware to fuel specific AI stages, meaning that from training to inference, the company has designated portfolios to address market demand and ensure hyperscalers get the very best. A prime example of this is how, with Blackwell Ultra, Team Green specifically focused on reasoning workloads for high-density compute. Similarly, with Vera Rubin, we'll see an emphasis on long-context applications, which are yet again targeted at agentic AI. So, when you look at NVIDIA's AI portfolio and how technical specifications have evolved, you'll realize that the firm has kept up.
Models and Applications are the final layers of the AI cake, and this is where not just frontier labels, but companies like NVIDIA earn the most. The primary 'bear' case with AI as a technology has been its diffusion towards the masses, mainly because, with early versions of "AI as a service", many saw the technology primarily limited to chatbots. However, in recent times, new use cases have emerged that have influenced the lives of several consumers. OpenClaw, Manus AI, Lovable, Replit, and many other services demonstrate how beneficial AI can be when used correctly.
However, according to Jensen, AI is still very early, and the technology is still early when it comes to mass adoption. NVIDIA believes that, for AI to be as massive as the internet, the world needs infrastructure worth "trillions of dollars," which is why it says the opportunity hasn't yet been realized.
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
