D2D: CPUs for Everyone
Also a look at Qualcomm's results and the shrinking relevance of telecom equipment
Just a reminder that we operate our newsletters on a paid model. Paid subscribers will get three newsletters a month including our China Deep Tech notes and audio recordings of our posts. Paid subscribers will also get early access to the newsletter as well other benefits coming soon. Please subscribe and support our work.
Highlights from our Blog
This was a busy month for PC CPUs. Apple launched its M3 family of laptop chips, which at first blush appear to be highly performative. At the same time, the market for PC CPUs has gotten more competitive with Qualcomm launching its latest AI-heavy CPU for Windows, and the report that Nvidia and AMD will also be entering the market with Arm-based CPUs. Unless on-device AI suddenly becomes very important to consumers soon, we think Intel can probably maintain its lead in the market.
Ericsson and Nokia both reported disappointing results. In part this is a reflection of the telecom operator spending cycle. 5G is now largely deployed in most major markets, and so spending on equipment is slowing, rather quickly. However, these results also reflect the growing existential question as to where Nokia and Ericsson sit in a network that has largely moved past them. Both companies have tried a whole range of diversification strategies, none of which succeeded. So now they are essentially stuck providing a set of products that a small group of customers really need, but no one else really wants. To further drive this point home, we made an unfair comparison of the equipment vendor’s revenues and profits with those of the hyperscalers.
Qualcomm reported its first good quarterly results in a while, reflecting the end of the consumer electronics slowdown of 2023. Despite this, the thesis for the stock remains unchanged. They have a great position in the slow-growing mobile market, some real traction in the automotive market which may be material late in the decade. Beyond that there is no sign of growth.
Intel is making some changes to its internal accounting structure, separating the product side from the fabs. This may seem immaterial as investors really on care about the consolidated figures, which should not change. However, internally this will lead to some important cultural shifts. Imagine you are an Intel sales person and have to decide between ‘paying’ the fabs for a rush order and losing a big order. Small changes can have big impact.
If you like this content, you should check out our podcast The Circuit
Semis and Hardware
We wrote about SiFive a few weeks back and thought their prospects sounded good. Then they fired ~15% of their staff including several people we had just met. These cuts seem to be a well-intentioned attempt to streamline their operations and focus onto their core products, but the timing leaves a lot to be desired.
There is a new data provider for semiconductors - the anonymous Lithos Graphein. Their website has a handy semis news aggregation. They also penned a very solid, technical-but-readable, analysis of where the US EUV regulations failed on the China Talk site. No idea who this is, but they know a lot about lithography.
Microcontrollers (MCUs) were once seen as cheap chips for ‘dumb’ devices. They remain a major workhorse for most industrial and automotive applications, and they are getting smarter. For instance, Renesas just launched a line of MCUs using Arm’s latest Cortex M85, and now there is a version of the high-level Lua programming language built to run on MCUs. This will be an important battleground in automotive.
There has been a lot of bad news for electric vehicles (EV) in the press lately. Toyota cut their 2023 EV forecast by 40%. General Motors and Ford are thinking about scaling back as well. VW is retooling its software team, again. Tesla reported some messy numbers as well. But China’s EV vendors are all putting in strong numbers. So read all the bad news with a grain of salt. The US automakers are on their back foot when it comes to EVs, and will need to make some important strategic decisions soon.
A lesson in why internally-built semis are difficult. A big feature of Apple’s new M3 CPU is its “Dynamic Caching” technology. Apple filed its first patent on the idea three years ago. How does any other non-chip company compete with that kind of time horizon?
All About Circuits published a very hand overview of analog semis. One critical thing to keep in mind about analog semis is that most design is still largely manual, with a major gap between the design tools available for analog and those for digital. We have a few ideas on how to bridge this if anyone has an interest.
Networking and Wireless
Everyone finds it annoying when their Air Pods’ bluetooth connection switches arbitrarily from iPhone to Mac. Now imagine you have to build Bluetooth support for every hardware platform out there. This piece can best be described as a descent into near madness.
Privately-held Zephr came out of stealth and revealed its plan to provide cm-level positioning for cell phones - by using other cell phones. Technically, this is a neat trick. Crowd-sourcing data and compute from smartphones is a good idea, and it is encouraging to see companies bringing the idea to life with a real business model.
Starlink reached cashflow breakeven. No mention how much of that cash is coming directly from US taxpayers.
Software and the Cloud
We have long wondered how Amazon’s main e-commerce web page could be so awful and user-hostile. It turns out much of the design problems can be blamed on a deliberate strategy to ‘guide’ consumer purchase choices.
A big pain point in using the public cloud is data transfer - moving data in and out of the cloud. This is an important area of future competition as companies like Cloudflare seem to offer a much cheaper alternative.
Thank you for reading D2D Newsletter. This post is public so feel free to share it.