Why AI Should Keep Getting Better

AI should keep improving because system-level scaling is replacing the slowing transistor-shrink engine of Moore’s Law.

• After nearly 39 doublings under Moore’s Law, leading-edge chips now contain more than 200 billion transistors, making the accumulated effect of compute growth large enough to enable today’s AI.
• Transistors are now only a few atoms wide and further shrinking is increasingly expensive, so compute is shifting from individual chips to systems combining processors, memory, packaging, networking, power, cooling and software.
• AI factories can coordinate thousands of processors, while larger models and datasets turn additional compute into capabilities that were previously impractical, consistent with the “Bitter Lesson.”
• The next scaling phase could draw on five simultaneous engineering frontiers—memory, packaging, networking, power and tensor compute—instead of relying on one semiconductor improvement curve.
• Franklin Templeton views AI as a potential general-purpose technology of the Fourth Industrial Revolution, with disruption extending beyond AI into robotics, genomics, clean energy and blockchain.

Takeaway: Investors should expect continued technological disruption and possible shifts in global market leadership, while recognizing that thematic technology investments carry concentrated sector and implementation risks.