A New Normal for Semiconductor Growth?

我们的分析表明,半导体行业revenue growth inflected from a low-to-mid-single-digit CAGR in the 2000-2015 period to a mid-to-high-single-digit CAGR in the 2015-2021 period (see Chart 1). Based on our战斗啊f Exponentials frameworkand proprietary semiconductor industry model, we wouldn’t be surprised if semiconductor revenues continue to grow at this higher rate for the foreseeable future.

Chart 1: Inflection in Semiconductor Trend Line Revenue Growth Rate. Source: SIA, SEMI, TechInsights (VLSIresearch), Applied Materials - Strategy & Market Intelligence.
Chart 1: Inflection in Semiconductor Trend Line Revenue Growth Rate.
Source: SIA, SEMI, TechInsights (VLSIresearch), Applied Materials - Strategy & Market Intelligence.

Memory Growth Accelerated by 900 bps, Foundry/Logic Growth Accelerated by 400 bps

We estimate memory (DRAM + NAND) revenues represented ~30% of total semiconductor revenues in CY21. Our analysis suggests the memory revenue CAGR stepped up to 13% in the 2015-2021 period from 4% in the 2000-2015 period. We estimate the Foundry/Logic (~70% of semiconductor revenue in CY21) revenue CAGR stepped up to 7% from 3%.

2015 Was an Inflection Year

Our Battle of Exponentials framework assumes 2015 was an inflection year when traditional Moore’s Law (a.k.a. classic 2D scaling) stopped working well. In around 2012, memory companies started describing a deceleration of unit cost reductions in DRAM. The NAND industry started its transition from planar to 3D NAND in 2013. In around 2014, the铸造行业转型toward “inter nodes” and “plus nodes” started in principle with competitive debate about 16nm vs. 14nm offerings.

Note on Our Semiconductor Industry Model

Our proprietary model is based on aggregating semiconductor-related revenues for 80+ individual companies to approximate the Semiconductor Industry Association’s aggregate revenue estimates. This approach allows us to segment revenues by device type (i.e. DRAM, NAND and Foundry/Logic), and further segment Foundry/Logic revenues by leading-edge and non-leading-edge “ICAPS” nodes that principally serve the IoT, Communications, Automotive, Power and Sensor markets. Given the cyclicality of revenues, especially in memory, we decided that using CAGRs provides the most meaningful comparisons across the periods of interest on a through-cycle basis.

In the next blog post I’ll share some data to help translate semiconductor revenue growth into wafer fab equipment (WFE) spending growth.

Receive updates in your inbox!

Subscribe Now

Comments

Hi Sundeep, Thanks for your informative posts on the subject. I'd be curious to know how the confluence of the 3 vectors that you refer too in the "exponentials framework", translate to the absolute level of wafer demand required to accomodate these vectors - any rough math or thoughts you can share? Rgds,

Want to join the discussion?

Add new comment:*

You can also fill out thisformto contact us directly and we will get back to you.