Generative AI is transforming the traditional computing stack, according to Nvidia CEO Jensen Huang, and it will one day form the core in a new era of software.<\/b>
\nGenerative AI is transforming the traditional computing stack, according to Nvidia CEO Jensen Huang, who says it will eventually form the core in a new era of software development.
Speaking with BlackRock CEO Larry Fink at the World Economic Forum (WEF) in Davos, Huang compared the impact of AI on modern computing to the advent of the web, the personal computing boom, and cloud computing.
\u201cIt\u2019s helpful to reason back to the first principles of fundamentally what is happening to the computing stack. This is a platform shift,\u201d he said.
\u201cNew applications were developed to run on a new type of computer. [With the] platform shift to the internet, a new type of computing platform hosted all kinds of new applications.\u201d
With each of these transformations, Huang noted, the \u201ccomputing stack was reinvented and new applications were created\u201d.
Generative AI essentially marks the latest step on this evolutionary path, with Huang noting that ChatGPT and Anthropic\u2019s Claude chatbot will be the tip of the spear that delivers new applications and tools in the future.
\u201cNew applications will be built on top of ChatGPT,\u201d he added. \u201cNew applications will be built on top of Anthropic Claude, for example. So it\u2019s a platform shift in that way.\u201d
Later in his conversation with Fink, Huang noted that AI is \u201creally easy to understand\u201d when looking at the technology through this prism.
AI is helping redefine not only what constitutes modern software, but how it\u2019s built, as enterprises flock to the technology to transform processes and streamline operations.
Analysis from Bain & Company shows AI is already transforming areas such as software as a service, with the arrival of agentic AI further bolstering efficiency improvements and speeding up development.
The key difference in this case is that while software in the past was \u201ceffectively pre-recorded\u201d, AI tools rely on prompts, marking a step change in how users interact with them. They\u2019re more intuitive, proactive, and are capable of discerning context with minimal information.
From a user efficiency perspective, this is a marked improvement when cumbersome software often relied on specific structured data to function.
\u201cSoftware in the past was effectively pre-recorded. Humans would type and describe the algorithm or the recipe for the computer to execute,\u201d he explained. \u201cIt was able to process structured information, meaning you\u2019ve got to put the name, the address, their account number, their age, where they live.\u201d
\u201cYou create these structured tables that software would then go and retrieve information from,\u201d Huang added. \u201cWe call it SQL queries. SQL is the single most important database engine the world’s ever known. Almost everything ran on SQL before.\u201d
Because AI can understand unstructured information, allowing it to digest text, images, or sound, Huang said AI represents a far more convenient mode of human-computer interaction.
\u201cFor the first time, we now have a computer that is not pre-recorded, but it’s processed in real-time,\u201d he commented. \u201cDescribe it however you like to describe it, and to the extent that it can understand your intention, it could perform a task for you.\u201d
Huang noted that the foundations upon which AI rests are crucial. It\u2019s a delicate balancing act in which multiple \u201clayers\u201d need to function in tandem for the technology to operate efficiently.
This aspect adds a degree of complexity that previous technological shifts were typically exempt from. Simply put, it\u2019s not just a simple click-and-go technology operating independently. The infrastructure stack that underpins AI is vital.
\u201cWhen you think about AI, you think about the AI models, but it\u2019s really important to understand industrially, AI is essentially a five-layer cake – at the bottom is energy,\u201d he said.
\u201cAI, because it’s processed in real time and it generates intelligence in real time, it needs energy to do so. Energy is the first layer. The second layer is the layer that I live in. It’s chips, chips and computing infrastructure.
\u201cThe next layer above it is the cloud infrastructure, the cloud services. The layer above that is the AI models.\u201d
According to Huang, there\u2019s a common misconception that the nitty gritty of the technology happens in the model domain. However, while this is crucial, it\u2019s in the application layer above that where real value and economic growth are generated.
There are challenges here though, Huang admitted. Model performance improvements have outpaced application development, meaning organizations are struggling to unlock value from the technology.
\u201cThe reason why last year was an incredible year for AI is that the models made so much progress than the layer above it, which is ultimately the layer that we all need to succeed.\u201d
\u201cThis application layer could be in financial services,\u201d Huang added. \u201cIt could be in healthcare. It could be in manufacturing. This later on top, ultimately, is where economic benefit will happen.\u201d
Make sure to follow ITPro on Google News to keep tabs on all our latest news, analysis, and reviews.
You can also follow ITPro on LinkedIn, X, Facebook, and BlueSky.<\/p>\n