AI demand is highly elastic: How cost-effective AI inference hardware will open massive markets

Description

As prices drop, new markets open up. This has been the defining characteristic of the semiconductor industry over the last 40 years for core hardware components like compute, transmit, and storage. AI compute is no exception. AI inference will likely follow the same trends as general purpose compute, WiFi, or storage, and the market for AI hardware and software stacks could approach $100B in the next 10 years.

The key drivers of price elasticity include the explosion in capabilities of the algorithms, the continuous drop in the cost of AI hardware, and easier design and deployment through intuitive software frameworks and abundant compute resources. From a software perspective, two trends in AI—abstraction and increased usability—mirror other disruptive multipliers in computing (GUIs, databases, networking, etc.). From a hardware perspective, the end of Moore’s law means that new computing paradigms (e.g., analog computation and nonvolatile information storage) are emerging as viable paths to drive down costs. Together, the hardware and software ecosystems have all of the ingredients of previous explosions in market value driven by disruptive technology.

Michael Henry explores AI’s $100B potential, diving into AI innovation at a hardware and software level and examining how the emerging trends in AI affect the value propositions of a wide range of products and markets.