Moore’s law has been the golden rule of the semiconductor industry for decades. It has held true for a good four decades, but signs are emerging that Moore’s law may have run into a wall.
Moore’s law argues the number of transistors on integrated circuits will double every two years. The ability to continuously increase the number of transistors per surface area on silicon chips has driven the remarkable advances of modern information technology. It has been suggested that as much as 40% of the global productivity growth achieved over the last 20 years is due to the rapid growth in information and communication technologies enabled by continuous improvements in semiconductors.
Moore’s law has enabled semiconductor manufacturers such as Intel Corporation (NASDAQ:INTC) and Advanced Micro Devices, Inc. (NYSE:AMD) to make long-term capital growth plans secure in the knowledge that the “law” virtually guaranteed successive generations of high-performance, in-demand products.
Moore’s law bounded by physics and economics
A new report from McKinsey and Co., authored by Harald Bauer, Jan Veira, and Florian Weig, suggests there is a real possibility the the era of Moore’s law is coming to an end. They argue that the time is rapidly approaching where the capital costs required to develop the next generation of even more miniaturized chips will be prohibitive.
Bauer et al argue that that “the economics of continued miniaturization could break down as cost-per-transistor reductions flatten for nodes with feature sizes below 28nm.”
They highlight the huge costs associated with semiconductor technology development and the equipment required to create the next-generation chips. Much of these huge costs relate to the investments required for cutting-edge lithography technologies and the complexities involved in the multipatterning approaches needed for semiconductor nodes of 32nm or below.
The report cites a recent McKinsey analysis that progressing from 32nm to 22nm nodes on 300-mm wafers would make fabrication costs increase by 40%. Moving to the next generation would also increase process development expense by 45% and chip design expenses by around 50%. Their calculations suggest that process-development costs would likely exceed $1 billion for nodes less than 20nm. These figures don’t even include the new labs required to produce the next-generation devices, which McKinsey estimates will end up costing at least $10 billion. The report also points out this means there are very few companies capable of financing the development of next-generation chips.
Moore’s Law: Timeline
Bauer et al conclude by saying that nothing is written in stone, and that they anticipate that incremental process improvements and the generation of devices currently in development will keep Moore’s law valid at least until 2020. What happens after that is anybody’s guess, but the report does lay out a series of contingent future scenarios that are worth consideration.