Introduction
Moore’s Law states that the number of transistors on integrated circuits doubles every two years. This law, predicted by Intel co-founder Gordon Moore in 1965, has largely held true for over 50 years. However, some experts believe that Moore’s Law is reaching its physical and economic limits. In this article, I will examine the evidence and arguments on both sides of this debate.
What is Moore’s Law?
Moore’s Law originated with a 1965 paper by Gordon Moore, in which he observed that the number of components per integrated circuit was doubling every year. He later revised it to every two years. This exponential growth has allowed computers to become cheaper, faster, and more energy efficient over time.
Some key points about Moore’s Law:
- Predicts that transistor density on chips will double every 2 years
- Has held true since the 1960s, allowing computing power to increase exponentially
- Driving force behind digital revolution and technological advances
- Not a physical law, but an observation and projection
Moore’s Law has been a self-fulfilling prophecy in the semiconductor industry. Companies set R&D targets to keep up with the pace of Moore’s Law. Innovation at the materials and manufacturing level has sustained the exponential curve thus far.
Signs That Moore’s Law is Slowing Down
Though Moore’s Law has largely held true for over 50 years, some signs indicate that its exponential growth may be reaching an end in the 2020s decade.
Physical Limits
One factor is physical limits on transistor size reduction. Transistors have become so small that quantum effects and leakage current disrupt function below 5 nanometers. EUV lithography allows further miniaturization, but is extremely complex and costly.
As Gordon Moore himself noted, Moore’s Law will eventually reach limits:
“I am confident that Moore’s Law will hold solid for another two decades and beyond.”
Economic Factors
The costs of each generation of semiconductors also continue rising. Building new fabs for smaller transistors now costs over $10 billion. The industry may lack financial incentives to press forward at the pace of Moore’s Law.
Slowing of Historical Pace
Reviewing the historical data shows that the doubling period has slowed slightly over time.
- In the 1970s, doubling occurred approximately every 2 years
- In the 2000s, doubling occurred approximately every 2.5 years
- From 2012-2018, some say doubling occurred every 3.5 years
This suggests exponential growth may be gradually decelerating.
Countering Arguments – How Moore’s Law Continues
However, experts on the other side argue that innovations will allow Moore’s Law to continue in some form for years or decades more.
Paradigm Shifts
- Major paradigm shifts, like 3D transistors or quantum computing, could restart the exponential curve even if 2D scaling slows.
- “There’s plenty of room at the bottom” – entire computing paradigms have changed before.
As Gordon Moore himself wrote in 2015:
“So while in the decades ahead Moore’s Law will be challenged, prospects are still very bright for continuing doubling every couple of years well into the future.”
Expanding Scope
- Moore’s Law may apply to more than just transistor density – things like data density, communication bandwidth, even genome sequencing show similar exponential trends.
- As long as exponential growth continues for some metric, Moore’s Law remains valid in an expanded form.
The Path Forward – Innovations That Could Prolong Moore’s Law
If physical and economic constraints slow the doubling of transistor density, what innovations could allow a continued exponential increase in computing power?
3D stacking
- Moving into the third dimension with 3D stacked chips provides more space for growth.
- This involves stacking layers of processors vertically to increase density.
- Companies like Intel are developing new 3D transistor designs.
New materials
- Moving from silicon to new materials like graphene or carbon nanotubes could enhance properties and allow smaller transistors.
- IBM already created a graphene chip with transistors below 5 nanometers.
Quantum computing
- Quantum computing offers an entirely new computing paradigm using quantum bits or qubits.
- Adds exponential growth in computing power by harnessing quantum phenomena.
- D-Wave, IBM, and others are advancing quantum computing technology.
Specialized AI chips
- Chips specifically optimized for AI/machine learning, like Google’s TPU, offer major gains.
- Expert systems and neural networks have their own version of Moore’s Law advancing at a rapid pace.
Conclusion
In summary, while traditional Moore’s Law scaling faces challenges, a range of innovations in materials, 3D integration, and new computing paradigms offer paths to prolong exponential growth in computing power. The spirit of Moore’s Law – fast paced innovation driving tech progress – is alive and well even if specific projections require updating. While nothing can increase exponentially forever, Moore’s Law still has years or even decades of life left in it.
