Yesterday we wrote about the ongoing tension with Nvidia’s acquisition of Arm. That post was already sprawling, but as we wrote it news emerged about Arm’s troubled joint venture (JV) in China, the company apparently now known as Anmou. Arm China is, to put it mildly, a troubled JV, and the latest developments put a whole new wrinkle in that saga, and could potentially play an important part in deciding what happens with the Nvidia acquisition.
Arm established its JV in China in 2018. Motivation for this deal are an all-too-familiar story for semis companies in China. It is unclear if Arm was pressured into the deal or just read the writing on the wall. Either way, Arm corporate wanted to position itself as supporting China’s semis and intellectual property (IP) position. They sold 53% of the JV to a syndicate of private equity funds. The syndicate was led by Hopu, a well-established PE fund with a team of seasoned semis professionals, but much of the syndicate was a grab bag of smaller, lesser-known funds.
Pro tip for anyone out there who still “wants” to start a China JV. If regulations require, or “encourage”, foreigners to own less than 50% of the JV, a good strategy might be to pack the controlling 51% stake with multiple entities, so as to not have one dominant partner. In this setting, the foreign company only has to corral a few votes from that syndicate to win any board votes.
Arm appears to have done exactly that leaving it with a fractured board whose members had highly divergent interests. They appointed a long-time Arm executive, Allen Wu, to be CEO of the JV. He had worked for Arm for a over a decade and done a solid job of building up Arm’s sales team in China. He also played a strong role in rounding up that investment syndicate. Cue ominous music.
Wu then staged a coup. The Financial Times does a great job of covering this, and we discuss the process in the post linked above. Arm China’s board voted to fire Wu, but he hacked a quirk of China’s corporate laws and held onto power. Now he surrounds himself with bodyguards (because kidnapping is not an unheard of business practice in China), and apparently has locked corporate Arm out of Arm China’s offices. Over the weekend he hosted an event in which he essentially established Arm China, now christened Anmou, as a leading independent semis leader in China.
Here is a link to Anmou’s statement (in Chinese, but Google translate does a credible job). And here is an analysis of the saga from a good analyst (although we do not agree with the “greatest heist” in the title and a couple other points).
What does all this mean?
Can Anmou succeed? Almost certainly not. First, the Arm China team has long been largely a sales organization. They have some talented engineers, but neither the depth nor the long-term experience in actually building the processor cores which are the heart of Arm’s business. Anmou is not getting updates to the Arm portfolio and it is highly unlikely they can create their own solutions to compete. The statement we linked to above is full of acronyms of future products, all the things they claim they are going to build. File most of that list under…aspirational. Where Arm China has done best is in small, embedded devices for IoT and industrial uses. Not a bad market, but much smaller than core Arm’s business.
And of course, none of it can be exported. Anything produced with Anmou-based chips will be faced with a lawsuit in any jurisdiction with a functioning legal system. How likely is it that some industrial company selling million dollar equipment will risk shipping something that contains a $1 Anmou chip?
There is also the big question of what IP actually resides within Arm China. Arm derives about 30% of its business from China, but most of that goes to the large chip makers, especially around mobile phones. However, many people we have spoken to about this say that the JV does not touch the bulk of that revenue, and only focuses on those smaller IoT and embedded, emerging chip companies. Arm China’s accounting is too opaque at this point, and so we can not say with certainty how this very important question gets answered.
Finally, there is the question how is this imbroglio tied to China’s broader semis ambitions? There seems to be an assumption in coverage of Arm China/Anmou that China’s government is backing this somehow, and this somehow plays in to China’s semis ambitions. We do not think this is the case. True, China wants more semis and more semis IP, but as far as we can tell the central government has stayed removed from the dispute. Earlier this year, Wu published a letter framing his dispute as one between China and foreigners, but we read this as a plea for help, and nothing seems to have come of it. China’s government is not a monolith, and while Wu seems to have backers from the local government in Shenzhen, his efforts do not appear to have landed in Beijing. This is likely due to the fact that the dispute is not one of Foreigners vs. China. Arm China’s board is 53% controlled by PRC entities, and they voted against him too. There are quite possibly people in Beijing who are as unhappy with the situation as Arm UK.
Almost everyone we speaks with believes this is just a fight over money, and the expectation is that Wu is just trying to use whatever negotiating leverage he can muster, and he has a lot of leverage.
Photo credit: Ridley Scott and Jordan Cronenweth
Yes Another great post. I know a lot of the focus on posts has been about the core processors in devices, but wondering about your thoughts on processors used in different parts of the network? I recently came across this company called Kneron and thought it was interesting but unclear if they are just taking their own book (currently raising a big series C round funded by Alibaba, Qualcomm, delta electronics, Foxconn and others)…
I only know a bit about Kneron, but there is a lot of activity around this part of the network, albeit for a market that is still not well defined. Everything is AI now and it is hard to say what is going to require a discreet chip and what will just end up as AI accelerator blocks in someone else’s SoC. It is clear that everything electrical is going to get more electronics (i.e. chips) in it, but we have been doing pretty well with MCUs for a long time. I think it makes sense that edge and industrial and embedded equipment makers would like to add more compute to their devices (plus AI for anyone raising money) and a CPU is overkill, but this is such a fragmented space that the economics don’t work out for some sort of general purpose ASIC to cover all of it. This is where a lot of the RISC V activity has been, for good reason.
I should probably write a piece about this, but where to start…