D2D: Small Progress for Intel and Qualcomm
Lots of progress for Nvidia
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
We attended Intel’s analyst event. It was largely lacking in news, other than the fact that their manufacturing process is on time.Good but not news. We did find it useful as a preview for how Intel is going to position its next generation of products. Lots of talk about “freedom” and “flexibility” which is code for “Intel does not lock customers in with Cuda” aimed at Nvidia. And a recognition that they are now fighting on multiple fronts, going so far to even include Qualcomm in their discussion of PC competition.
A few weeks ago we participated in a live stream discussing Nvidia. It was filled with a lot of baseless rumor and showed the ugly side of the Internet rumor mill. So we laid out our take on Nvidia. The company is enjoying significant real demand. It will not last forever, semis are cyclical, but Nvidia is incredibly well positioned for the major transition in compute workloads which everyone is calling AI.
The smartphone software landscape needs a major shake-up. We are stuck in a slow-growth market, and even Apple is struggling to differentiate each generation of the iPhone. Someone on the hardware side (looking at you Qualcomm) needs to make a major effort to shake things up.
If you like this content, you should check out our podcast The Circuit
Semis and Hardware
Digitimes say what we all suspected. Qualcomm is picking up a lot of design win share in automotive.
There is a lot of speculation as to what else Arm has planned for growing revenue. They put out some fairly big projections ahead of their IPO. They have to raise prices, but they are also exploring new business models. One rumor we have heard repeatedly is that they will start building their own chips. Many take this to mean they are going into competition with their customers. Much more likely, is that they are simply starting an ASIC services business to help their customers take chips to production.
Qualcomm published a blog post extolling the virtues of RISC V. They highlight the Flexibility, Control and Visibility the ISA offers chip designers. This includes an ‘interview’ with Ziad Asghar, who seems to be the designated RISC V point person for the company. It does not include a single mention of Arm. It does add a tiny bit of detail around the JV Qualcomm, Bosch, NXP, Nordic and others put together last month - admitting that it will target automotive and provide reference designs and certify compatible hardware.
The Register went to Intel’s Innovation Day and came away with a similar impression to ours. They just came up with a better headline. “Intel slaps forehead, says I got it: AI PCs. Sell them AI PCs.”
The US Department of Defense has allocated funds to eight regional semis R&D hubs. This is a good start, but the government can and should go much further.
US policy makers are (maybe) working hard to craft a coherent industrial policy for semis. Government procurement contracts will probably carry more impact. So Global Foundries signing a $3 billion, 10-year contract with the US government is worth noting.
Networking and Wireless
Cisco is acquiring Splunk for $28 billion. Cisco once built the hardware that powered the entire Internet. They still build a lot of that gear, but for a decade now they have been slowly transitioning to a software-based model. Or more specifically, a licensing model. It reminds us a lot of Oracle. There is probably a lesson here that technology companies transition from being run by engineers, to sales people and then to lawyers. The joke we heard was that many Splunk customers are excited for the deal because they think they will be able to apply their Cisco status to reduce their Splunk bill. While the deal is accretive to Cisco EPS by Year 2, worth noting that Splunk has never made a GAAP profit.
Over the past 6 years, the number of phone brands has fallen by 65%. The analysts at Counterpoint Research attribute this to pandemic and ensuing components supply crunch. Good analysis, but we think they are undercounting the number of Chinese brands and the raw economic impact of China’s industrial policies. This trend began even earlier and is very much the result of the biggest brands getting bigger and exerting their market power.
Tarana just raised $50 million in a ‘growth’ round, to fund its expansion and next generation product. Tarana makes wireless networking gear for WISPs and other fixed wireless providers. They have been slogging a way at this market for a long time and seem to have finally reached traction, or more precisely, the market has reached the point where it can generate substantive demand for Tarana.
Not so long ago, Internet traffic in Latin America all had to go through an Internet exchange in Miami, even if both ends of that traffic were in the same country. The market has evolved massively since then, and we can probably thank smartphones.
OpenRAN pioneer Rakuten is undergoing a significant wave of executive departures. Coupled with the struggles at Mavenir we linked to last issue, it is clear that OpenRAN is facing some precarious times.
Software and the Cloud
We write about the realities of porting software from one platform (x86) to another (Arm and RISC V). Our point is that the last 20% of the optimization work, which makes the whole effort worthwhile, is also the hardest to do. Game developers face the same sort of challenges when porting from one platform (PC) to another (Nintendo Switch). This blog post gives a good sense of the fine grained work involved. Different hardware, same problem.
The big software and Internet companies are optimized for growth. This developer argues that comes at the expense of good software, as applications are dumbed down to attract the marginal user. We do not agree with all of this, but it is a useful framework for thinking about all the apps we use everyday.
We can all agree that wireless standards are useful and important, but can they really compare in significance to the standardization of electrical sockets?