Following up our post on Nvidia buying Arm, we had the opportunity to speak with a few friends based in China recently. Admittedly, this is anecdotal, but their take is uniformly skeptical that the deal can pass regulatory approval in China.
Their concerns break down roughly along two lines – regulatory and straight-out political. Given the state of the world today, these two trends are deeply entangled.
If we step into China’s viewpoint, the concerns are fairly obvious. Nvidia touted the deal as strengthening the company’s position in AI. The word appears a half dozen times in the initial press release and another dozen times in their investor presentation. AI is clearly presented as a key rationale for the deal. China has made it very clear that it views “AI” as a strategic domain, not just for their technology industry but for their national security. Set aside the truth of that concern (and the US is just as prone to some ‘expansive’ thinking on the attributes of AI as well), this is a field they care a lot about. It is reasonable to think that AI is one area of semiconductors where China can become a global player, leapfrogging its lack of domestic CPUs and GPUs. Even absent a Trade War, this deal would merit intense scrutiny from Beijing.
However, it is not entirely clear how important China believes Arm is to AI, or to semiconductors more broadly. The semis industry views Arm as very important, but some of this rests on path dependency. The US has built its semis industry on the back of Arm, but if you are starting from scratch maybe Arm is not quite as important. Or maybe it is not perceived as important.
We have a few reasons to think this. First, most people we speak with believe that the biggest users of open source Arm alternative RISC V are Chinese companies. We have written in the past of how expensive it is for a new company to obtain an Arm license, and so it makes sense that many of China’s 1,300 and growing semis companies are starting up with something more affordable. Certainly, the big Chinese electronics companies like Huawei, BBK (Oppo, Vivo, et al) and Xiaomi are already deep into the Arm ecosystem, but many, newer companies are likely taking a different path.
Second, we have to wonder how well Arm China is actually doing. Arm says that China contributes 30% of its revenue, but we now believe that much of this goes to Arm directly, bypassing the local JV. This means, if push comes to shove, Arm could divest its stake in the JV, without incurring a 30% revenue hit. Given the convoluted situation around Arm China, it is possible that the authorities in Beijing do not see Arm as quite as strategic as we assume they do.
That being said, the current political environment all but assures that any semiconductor transaction with a US buyer will be heavily scrutinized in Beijing. We have no insight into how decisions are made at the highest levels of China’s government, very few people do. But this case seems easy to guess.
Nvidia has says it is planning on closing the acquisition in 18 months which is long by most standards, but short given that Qualcomm waited two and a half years before throwing in the towel on NXP. Since then, obviously a lot of has happened in US-China relations. As part of Phase One of the US-China “Trade Deal”, China signaled that it would not hold back regulatory approvals of deals. Unfortunately, that means the Nvidia/Arm deal is likely to be decided by the political winds. If China’s government wants to signal that it wants to improve its relationship with the US, either permanently or for some immediate negotiation tactical reason, the deal may get approved. By contrast, if they want to signal a harder line, approvals for Nvidia could drag on indefinitely. At the very least, nothing is going to happen ahead of the US election and 2021 inauguration.
Now, we have to assume that Nvidia, Arm, their bankers and lawyers are all smart people. And that some of them are very up to date on what is happening in China. So they may have some reason to believe that they can close this deal. For what it’s worth, Nvidia has never attracted the ire of China’s government the way that others, like Qualcomm, have. From what we can tell, they have a fairly good business in China and a good relationship with China’s government. And let’s not forget Softbank has a role to play here, and they definitely seem to have some insight into working with China. So maybe there is an angle that is not visible to outsiders. Alternatively, we have seen plenty of M&A deals where problems like this get willfully ignored, with no one wanting to spike the deal with a last-minute objection.
Put simply, we think it is going to be very hard for Nvidia to get deal approval in China. Not impossible, but considerably long odds.
Pingback: China has won the AI Accelerator Race | DIGITS to DOLLARS·