Definition
Moore’s Law Redefinition refers to the paradigm shift in computing where the original definition—the doubling of transistors on a chip every eighteen months—is superseded by the “Software-Hardware Symbiosis” of parallel processing. As physics-based limits (see Dennard Scaling Collapse) killed the speed gains for individual processors, the “law” survived only by moving to massively parallel architectures like the GPU, where intelligence is scaled by multiplying processing cores rather than shrinking them further.
Why It Matters
If we assume computational power will continue to grow exponentially for free, we stop innovating in software efficiency. As physical limits are reached, the redefinition of Moore’s Law toward architectural and algorithmic efficiency becomes the only path forward for tech progress.
Core Concepts
- The End of “Free” Speed: For forty years, Moore’s Law meant that existing software simply ran faster on new chips. This “free lunch” ended in 2005.
- Transistor Scaling vs. Intelligence Scaling: While transistors continue to get smaller (at a slower rate), the effective intelligence of systems now scales with the amount of compute and data (The Bitter Lesson).
- The Intelligence Premium: In the The Software Century, the “law” is no longer about transistor count, but about the “OODA loop” speed of AI models trained on parallel hardware.
- Intel’s Denial: The historical case study of Intel’s failure to recognize the Dennard collapse, continuing to predict linear gains while Nvidia pivoted to parallelism.
- Infinite Demand: Unlike audio or networking, 3D graphics and AI have “infinite demand” for processing power—users aren’t finished until they are “living inside the Matrix.”