We're in a golden age for chip R&D activity, and that makes it a pretty good time to be one of the companies providing the software, intellectual property and/or manufacturing plants needed by chip engineers to bring new products to market.
That thought came to mind as I read an Axios report about how Alphabet/Google (GOOGL) is working on a processor that could power its 2021 Pixel phones, along with future Chromebooks. The processor, codenamed Whitechapel, is said to feature 8 ARM CPU cores, rely on Samsung's next-gen, 5-nanometer, manufacturing process and contain dedicated circuitry for improving Google Assistant's performance.
To date, Google has relied on Qualcomm's (QCOM) processors to power its Pixel phones. And on the whole, Qualcomm has done a pretty good job of addressing the needs of the high-end and mid-range Android phone markets.
Nonetheless, Google feels that it's worth its while to develop its own smartphone processor. Just as Google has felt it's worth developing an image-processing chip for Pixel phones, and to develop proprietary silicon (its Tensor Processing Units, or TPUs) for accelerating AI/deep learning workloads within its data centers.
Apple (AAPL) , meanwhile, now relies on its own processors to power nearly its entire product line -- Macs are the exception, but perhaps not for long -- and has also developed a number of other chips for its hardware.
Amazon.com (AMZN) has designed a slew of chips that can be found inside of AWS data centers; Microsoft (MSFT) has created chip designs for both Azure data centers and its HoloLens headsets; Tesla (TSLA) is using a home-grown processor to power its third-gen Autopilot system; and Huawei's HiSilicon unit now develops chips for everything from phones to IoT devices to cloud servers.
The list goes on and on. In many cases, this surge in chip design activity among companies that aren't first and foremost chip developers simply stems from a wish to use chips that are cheaper and/or can deliver better performance/power efficiency than what third-party offerings can provide.
But other factors can also be at play. These include an explosion in the number of workloads -- in everything from smartphones and IoT devices to cars and cloud data centers -- for which no off-the-shelf silicon optimized for that particular workload happens to exist. Some of the custom ASICs developed by AWS and Azure to help underpin their data center infrastructures arguably fit this description.
In other cases, a company might be motivated by a wish for greater independence from and/or negotiating leverage with third-party suppliers. A lot of Huawei/HiSilicon's chip design work falls into this category, particularly in the wake of U.S. export restrictions, and so might some of the attempts by cloud giants to develop their own server CPUs and AI accelerators.
Regardless of a company's motivation, this uptick in chip design activity among the likes of hardware makers and cloud giants, together with attempts by independent chip developers to address a wider variety of workloads, is a positive for electronic design automation (EDA) software giants Cadence Design Systems (CDNS) and Synopsys (SNPS) . It also spells more licensing opportunities for chip IP giant ARM, which is owned by Japan's SoftBank and eying a 2023 IPO, as well as more business for independent chip developers that work with others on custom ASICs, such as Broadcom (AVGO) .And to an extent, all of this chip design work is also good news for chip contract manufacturers (foundries) such as Taiwan Semiconductor ( TSM) and Samsung. Virtually all of the hardware makers and cloud giants are relying on foundries to manufacture their chips, and even if some of these chips take share from other products being made by the foundries, having a larger/more diverse customer base is in their interests.