The prelude to the Superintelligence Revolution is opening.
For more than half a century, "Moore's Law" — an established symbolic milestone of technological advancement — has played a core role in explaining the pace of development in the semiconductor industry. Moore's Law originated from the empirical observation that the number of transistors in semiconductor integrated circuits doubles approximately every 18 to 24 months, enabling explosive performance improvements and cost reductions throughout the IT industry. Indeed, Moore's Law has been accepted not merely as a simple technology prediction, but as a kind of "technology myth" that has driven the accelerating progress of digital civilization.
However, today we are witnessing this predictive model facing its limits. Semiconductor microfabrication technology is becoming increasingly complex, and physical barriers at the atomic level are gradually slowing the increase in transistor density. In other words, Moore's Law has reached a point where it can no longer be considered a viable formula for technological evolution. Against this background, a new paradigm is emerging that presents a new direction and pace for technological evolution.
At its center lies a law of an entirely different character: "the evolution of intelligence in the AI era." In particular, NVIDIA CEO Jensen Huang defines AI's pattern of development, which surpasses the pace of Moore's Law, as the evolution of "hyper-accelerated" intelligence. Here, "hyper-acceleration of speed" does not simply mean improvement in computational performance. It is a concept that explains the phenomenon of non-linear amplification of capabilities across intelligence as a whole — from data processing, model training, and algorithm optimization to human-level reasoning ability.
Thus, we are now entering a new era in which the old formula of linear development no longer applies — an era in which intelligence itself becomes the law. This article examines the concept of the "Law of Intelligence," which is replacing the existing Moore's Law centered on this phenomenon of intelligence hyper-acceleration, and analyzes in multiple dimensions the meaning that this change has for technological progress, industrial structure, and social systems.
The Economics of the Internet Revolution
The important theoretical foundations that define the turning point of the Information Revolution and the essence of the digital era originate from various technology laws. These laws play a core role not only in observing technological trends, but in explaining the structural evolution of the internet and the digital economy as a whole. Among these, the oldest and most widely known concept is Moore's Law itself. Proposed in 1965 by Intel co-founder Gordon Moore, this law originated from the prediction that the number of transistors on semiconductor chips would double approximately every two years, and that computing performance would accordingly improve exponentially. This prediction was realized for over 30 years, laying the foundation for modern digital society including personal computers, smartphones, and cloud computing.
If Moore's Law drove the development of computing, in terms of networks there is Gilder's Law, proposed by American economist and futurist George Gilder. According to this law, bandwidth increases at least three times faster than computer performance. That is, if computer performance doubles every 18 months according to Moore's Law, communications performance doubles every 6 months. Gilder's Law has been used to predict the successful convergence of digital entertainment and broadband, and furthermore explains the core driving force of the internet economy, where "the flow of information itself" beyond simple communication connectivity becomes an economic asset.
In contrast, Michael Ruettgers, who was CEO of global storage specialist EMC, paid attention to data storage capacity. According to Ruettgers' Law, the data storage capacity demanded by companies and governments doubles every year. This can be said to be an insight that foresaw the arrival of a data-centered society where data is not merely generated, but stored, analyzed, and converted into economic value. Today we are generating vast amounts of data amid the universalization of AI, IoT, and smart devices, and the development of cloud storage, distributed storage technology, and high-speed SSDs for storing and efficiently managing this data is supported by Ruettgers' Law. In particular, blockchain-based distributed storage technology ensures the security and integrity of data, making new forms of data economy possible.
While the laws mentioned above explain the technical foundations of physical infrastructure, storage capacity, and network transmission speed, Metcalfe's Law, proposed by Robert Metcalfe, co-inventor of Ethernet and co-founder of 3Com, centers on explaining the social and economic value created by those technologies. Metcalfe argued that the value of a network is proportional to the square of the number of connected nodes. That is, as the number of users participating in the network increases, the utility and economic value of the entire network increases exponentially. This law not only explains how the value of social media platforms such as Facebook, Instagram, and KakaoTalk rises with increases in the number of users, but also applies to the scalability and economic structure of blockchain ecosystems like Ethereum. Recently, this network effect, combined with AI and the data economy, is functioning as a core principle explaining the virtuous cycle structure in which the more data is accumulated and connected, the more sophisticated intelligence is formed, which in turn creates new value and services.
Ultimately, Moore's Law, Gilder's Law, Ruettgers' Law, and Metcalfe's Law are essential concepts for understanding the structure and evolutionary aspects of the internet economy, going beyond simple technological trends. These laws form the foundation of the Information Revolution in terms of computational capability, transmission speed, storage capacity, and connection value respectively, constituting the foundation of the intelligent information society in which we live today. These four laws are not separate but complementary to each other, and by working together they form the foundation of the complex mechanism that enables the exponential expansion of the digital economy and the hyper-acceleration of intelligence.
The Era of Hyper-Acceleration
Today, however, we are entering an entirely new phase of evolution that can no longer be explained solely by the laws of the digital economy that drive the Internet Revolution discussed above. It is a transition to an era in which "intelligence" itself — which learns, judges, and evolves autonomously, beyond the simple processing and transmission of information — leads, an era governed by the "Law of Intelligence." And at the center of this flow is the evolution of GPU architecture led by NVIDIA. The technological lineage from Hopper to Blackwell, and then to Rubin, heralds not merely improved chip performance, but a shift in the computing paradigm.
The Hopper architecture (H100 GPU), which announced the first departure, was announced in 2022, featuring over 80 billion transistors based on a 4nm process, up to 80GB of HBM3 memory, and performance of up to 4 petaflops (PFLOPS) in FP8 computation. Subsequently, the Blackwell architecture (B100 GPU), announced in 2024, adopted an MCM (Multi-Chip Module) structure integrating two GPU dies, and dramatically improved performance and scalability based on next-generation HBM3e memory and 5th-generation NVLink technology. CEO Jensen Huang mentioned in his GTC 2025 keynote address that Blackwell achieved up to 68 times performance improvement compared to Hopper. This signifies not only an increase in petaflops (Blackwell's maximum computational performance of 20 petaflops is roughly five times that of Hopper in simple comparison), but a comprehensive advance in data center-level efficiency, computational parallelism, and performance per watt. In particular, FP8 computation was further optimized in Blackwell, designed to dramatically shorten training times for GPT-4 and GPT-5-class models.
The next generation following Blackwell's technological innovation is the Rubin architecture. This next-generation architecture, named after Vera Rubin, is expected to be announced around 2025-2026 and is designed as infrastructure for the superintelligence era leading to AI computation and AGI (Artificial General Intelligence). Rubin is expected to offer up to 900 times performance improvement compared to Hopper — a hint that not just speed but the philosophy and structure of AI computing itself may change. The emergence of Rubin, which includes a smaller process (below 2nm), more integrated 3D packaging technology, more sophisticated mixed-precision computation systems, and AI-friendly algorithm optimization, will be the occasion to redefine existing GPUs from simple accelerators to the engines of AI.
The key is that all this evolution is completely surpassing the linear prediction models represented by Moore's Law. AI's development is now determined by the triple driving force of "computational capability × data learning × model optimization." And the self-reinforcing innovation loop is working: algorithms create better models, those models understand more data, and that data in turn improves algorithms. These changes mean that the Law of Intelligence is becoming the new standard for technological advancement.
The evolutionary history of AI GPU architectures from Hopper to Blackwell to Rubin is not limited to merely the history of the semiconductor industry. It is a symbolic flow showing how intelligence is implemented on a mechanical foundation and how it amplifies its own pace as it evolves. All these technological innovations are ultimately directed toward one goal: building the foundation for AI that thinks autonomously and learns creatively — "Artificial Superintelligence."
The Prelude to the Superintelligence Revolution
We are now standing at a turning point where the paradigm of technological advancement is decisively shifting. At its center is the "hyper-acceleration of speed," which signifies the advent of an entirely new era that can no longer be contained by the laws that explained the past Information Revolution. Most importantly, the emergence of AI that surpasses human intelligence is becoming a reality. AI is evolving beyond simple computational assistance tools into entities that think autonomously, create new scientific theories, and solve problems in ways humans cannot understand.
Also, this flow of hyper-accelerated AI development is gradually inducing structural changes throughout the economic system. Today, AI no longer remains a simple passive technological resource, and in some areas it is functioning as a quasi-autonomous economic actor. In particular, as AI directly intervenes in the operation of digital assets, automation of token-based transactions, execution of smart contracts, and real-time analysis of market data through computational capability and algorithmic judgment, economic process automation that minimizes human intervention is becoming possible.
And another change that must be noted is the "Co-Evolution of humans and AI." AI is no longer a simple external assistant, but is establishing itself as a member of a knowledge community that creates new intelligence together while interacting with humans. In the past, humans designed and trained AI; now, AI educates humans, compensates for human limitations, and is opening new horizons of knowledge and creativity. This co-evolution is acting as a force that restructures even human ways of thinking, problem-solving approaches, and even social structures, and future society will enter a complex intelligence system based on human-AI cooperation. The prelude to the "Superintelligence Revolution" is beginning.
