Money Changes Everything...in Fabs
A look at the ways in which economic policy can lead to leadership like TSMC has.
In which we examine the indirect subsidies that TSMC has enjoyed as the result of Taiwan’s currency, and the way in which they invested that windfall in developing their own talent pool. Also, Arm looks like it wants to put its complaint against Qualcomm to a trial.
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. 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
As much as TSMC deserves credit for its technical capabilities, we think it is important to recognize that its competitive advantage is in no small part due to Taiwan’s currency policies, which gave the company twenty years of discounted costs. TSMC management’s real super power was investing that windfall in developing its own talent, and not squandering it on poorly thought out share buybacks or far-flung acquisitions.
Almost everyone assumes that Arm and Qualcomm will settle their legal dispute on the courthouse steps. Unfortunately, Arm’s recent cancelation of Qualcomm’s license seems to indicate that Arm actually wants to take the case to trial. Because of course a jury comprised of regular citizens with no understanding of the industry will have an easy time parsing the intricacies of the case in a wholly rational manner….
RISC V continues with its steady progress, which unfortunately for us makes for some very boring headlines. We think the open framework ISA will play an important role in embedded, IoT and industrial systems, and alongside Arm in complex SoCs. Maybe someday RISC V will be ready to power standalone data center-grade processors, but our sense is that today it still lacks the compute horsepower and software ecosystem.
If you like this content, you should check out our podcast The Circuit
Semis, Hardware and Deep Tech
We think the most important topic in AI semis in 2025 will be inference economics. Everyone not named Nvidia or Broadcom, is going to make the case that CPUs (or apps processors) are the better path for inference. Unfortunately, GPUs and accelerators have such a performance advantage that CPUs will only make sense if they are very cheap, at least for now.
Electric motors have not changed in 200 years. That it is starting to change, C-Motive launched a whole new class of motors with no magnets. And there are a few more companies out there working on similarly innovative approaches.
MIT has developed a method for 3D printing electronics. This has been a long-time quest in the 3D printing world. And while this is not going to replace semiconductors any time soon, it does offer a lot of utility in many use cases.
The US government awarded a grant for establishing an EUV development center in IBM’s nano-park in Upstate New York. We like the idea of this, but as we wrote a while back, implementation details for any program like this are critical, and we think there is a better approach.
Google let researchers benchmark their new Axion CPU against Intel and Ampere instances. Performance was fairly good, if not quite exciting, but that is still a good achievement for their first attempt at a CPU.
Consumers do not know why they should care about AI.
As we mentioned above, this author says RISC V is not yet competitive with traditional CPU architectures.
Networking and Wireless
A fun graphic showing the changes in mobile phone market share over time. A trip down memory lane with some long-forgotten leaders.
Voyager 1 transmitted for the first time since 1981. Score one for hardware resiliency.
Software and the Cloud
Data from Carta confirms everyone’s intuition that it has gotten harder for Seed Stage companies to reach Series A. Looks like last year the rate of ‘graduation’ was about half of prior years, and is still in decline. Of course there are still a lot of companies that are raising multiple Seed rounds, but anecdotally unless there is ‘AI’ in the name, we know very few companies are making it out of seed stage now.
We know a lot of people who rate Gen AI “Answer Engine” Perplexity highly. And Perplexity has now launched a Finance tool, which looks very tempting and useful.
One more academic effort to move AI and LLMs away from GPU-bound compute methods. There are going to be a lot of these, so far most of them seem to sacrifice some degree of precision for their results. While that is useful, the same methods can be applied to GPU systems as well and get similar-ish results. That being said, maybe one day one of these approaches will work.
Science and Climate
A new tool for detecting intestinal cancer from a drop of blood. We have no idea how well this will work, but it did alert us to the “Lab on a Chip Journal” which looks very promising. (Or at least it will after Notebook LM explains it all to us.)
Diversions
This simulator shows how Wi-Fi signals are propagating around your home or office. It only works on laptops (not phones), but still provides hours of fun. Be sure to close the browser tab when you are done.
Image by Google Gemini.
Thank you for reading D2D Newsletter. This post is public so feel free to share it.