Moore’s Law, proposed in 1965 by Intel’s Gordon Moore, which states that “the number of transistors doubles roughly every two years,” has accurately explained the growth of the semiconductor industry for more than 50 years. However, now that transistors are approaching the atomic level, this law is reaching its physical limits.

● Leakage current: The smaller transistors become, the thinner the gate becomes, and current leaks out unintentionally. This leads to increased power consumption and heat generation

● Soaring manufacturing costs: The construction cost of a leading-edge fab has reached the scale of 2 trillion yen, and we are no longer in an era when a single company can cover it

● The next paradigm: A shift not only toward continued scaling but toward performance improvement through “integrated technology” that includes packaging, materials, and system design

● More than Moore: A direction that raises the value of semiconductors by integrating different technologies and adding new functions, without relying on increasing the number of transistors

To go beyond the limits of silicon, research institutions and companies around the world are hurrying to develop next-generation materials and new device structures. “Post-silicon” candidates that satisfy performance, power savings, and manufacturing cost all at once have not yet been narrowed down, but several promising directions have emerged.

Graphene and 2D Materials

Ultra-thin films consisting of a single layer of carbon atoms. With extremely high electron mobility, they are candidates for ultra-high-speed, ultra-low-power devices

Ga₂O₃ (Gallium Oxide)

A breakdown electric field exceeding that of SiC. Research is accelerating on it as a next-generation power device

Carbon Nanotubes

Theoretically capable of faster, lower-power operation than silicon. Establishing mass production technology is the challenge

Self-Assembly and Atomic Layer Deposition

Manufacturing technologies that build up materials at the atomic level. Methods that go beyond the limits of conventional lithography

Research is advancing on entirely new architectures that change the very principle of conventional semiconductors, “computing with the charge of electrons.” All of them are challenges to meet the demands of the AI and data center era: “processing large amounts of data quickly and with low power.”

● Silicon photonics (optical semiconductors): Transmits information with light instead of electrical signals. Goes beyond the speed and heat limits of copper wiring. Close to practical use as co-packaged optics

● Neuromorphic semiconductors: A computing method that mimics the neural circuits of the human brain. Achieves learning and inference with ultra-low power. Intel’s Loihi and others are leading development

● Spintronics: Uses the “spin (magnetism)” of electrons as an information carrier. Being put into practical use as non-volatile, high-speed, low-power memory (such as MRAM)

● Analog in-memory computing: Eliminates the data transfer bottleneck by performing computation directly where the memory is. Has the potential to dramatically reduce the power consumption of AI chips

Today’s computers calculate with bits that are “either 0 or 1,” but quantum computers use quantum bits (qubits) in which “0 and 1 exist at the same time,” giving them computing power that vastly surpasses conventional computers for certain problems. However, highly precise semiconductor technology is indispensable for the control circuits that run quantum computers themselves, so the two are not competitors but complements.

● Quantum error correction: Qubits are extremely vulnerable to noise and prone to computational errors. Practical use requires a large number of error-correction qubits, and the integration density of control semiconductors is the key

● Cryogenic operation: Many quantum computers need to operate near absolute zero (−273°C). The design of cooling systems and control circuits is one of the biggest technical challenges

● Application areas: Surpasses conventional computers in drug discovery, materials simulation, code breaking, optimization problems, and more. Coexistence in specific fields, rather than replacing general-purpose computers, is the realistic outlook

● Quantum × AI: Quantum versions of machine learning algorithms (quantum machine learning) are being researched and are attracting attention for their potential to dramatically enhance the capabilities of AI

As a thought experiment that goes beyond technical explanation, Chapter 10 also depicts a vision of semiconductor society beyond 2050. Looking back at the history of semiconductors, things that were once “science fiction” have become reality. There is ample reason to believe that changes beyond our imagination will also occur over the next 30 years.

● AI designs semiconductors: AI is built into EDA (electronic design automation), automatically optimizing design spaces that humans cannot explore. Design periods and human costs will change dramatically

● Bio-integrated devices: An era in which semiconductors are implanted in the body and interface directly with nerves and the brain. Applications in medicine, welfare, and human augmentation are approaching reality

● “Invisible technologies” support society: Semiconductors will become smaller and more embedded, blending into every place, from infrastructure and clothing to agriculture and architecture

● The importance of ethics and sustainability: Not only technological capability but also a philosophy and sense of purpose about “for whom and for what we make them” will be indispensable for the next stage of development of the semiconductor industry

● Moore’s Law is reaching its physical limits. The paradigm is shifting from scaling to performance improvement through “integrated technology”

● Research on post-silicon materials such as graphene, gallium oxide, and carbon nanotubes is accelerating

● New architectures that change the very principles of computation, such as light, spin, and neuromorphic approaches, are emerging

● Quantum computers are not competitors but complements to semiconductors. High-precision semiconductors are indispensable for their control circuits

● Beyond 2050, AI-driven automated design, bio-integration, and blending into society will become real challenges

 

Supervised by: Koji Kakumoto (President, OTIS Co., Ltd.)

*This article integrates and summarizes the content of Chapter 10. For details, please refer to the original article for each topic.

Read Chapter 10 →