Update: We misread the 2016 press release announcing the Softbank acquisition. We have updated this post with the correct numbers. Turns out Softbank isn’t making such a goo return on this deal.

So the rumors turned out to be true, Nvidia is acquiring Arm. Despite this possibility being speculated about for months, we still find it a bit surprising.

Deal Overview

Nvidia is paying a total of $40 billion for Arm in a fairly complex formula – $2 billion in cash upfront, and then another $10 billion at closing, along with 44.3 million shares of Nvidia stock, currently worth $21.5 billion. Nvidia is also allocating $1.5 billion of its stock for Arm employees, and a further $5 billion earn-out for Arm owner Softbank, if Arm hits certain (unspecified) targets. Adding those employee shares to the earn-out means $6.5 billion of value is all set for some time in the future, which is pretty favorable for Nvidia and a bit unusual for deals of this size.

Nvidia says the deal will be immediately accretive, which sounds good, but of course they are paying for 2/3rd of the deal in their own stock, which is currently trading at something like 53x 2020 earnings. Some preliminary, very back of the envelope math, deal accretion implies that Arm’s earnings for the year are at most 35% higher than when Softbank acquired them in 2016.

Bottom line, this is a great deal for Softbank, who are going to almost double their investment in just under for years this is about as good a deal as Softbank is going to get .

For Nvidia, however, the deal looks much less clear.

Nvidia’s Value Proposition

We are writing this ahead of the companies’ Monday morning call, so we have yet to hear any explanation for the deal other than what is listed in the press release. And what they include there makes no sense. That is not to say the deal makes no sense, just the press release does not get us to their true rationale.

The focus of the press release is all about AI, and how the combination of Arm and Nvidia is all about AI everywhere. The flaw in this logic is that Nvidia is already the leader in semiconductors for AI, and gains almost nothing technology-wise from Arm on this front. Same holds for the data center market. And note, no mention of autonomous vehicles. There is some discussion of AI for IoT, which might make sense, except Arm is really nowhere in IoT. A point highlighted by the fact that as recently as a month ago, Arm was trying to offload its IoT business, and the fact that today’s deal does not actually include Arm’s IoT services business.

Nvidia does gain a sizable presence in mobile, a market which has eluded it for many years. However, we suspect they chose not to highlight that side of the business because that leads to some potential trouble.

Arm is the proverbial Switzerland, the neutral party that sells to all competitors in semis. An acquisition by one of those competitors puts Arm’s business with all the other competitors at risk.

Nvidia may be banking on the fact that its only direct competitors for GPUs are AMD and Intel, neither of whom do much with Arm. But we suspect that many of the other chip majors will be very uncomfortable with this deal. For example, Nvidia and Qualcomm have never been particularly close, and both have big ambitions in automotive.

In recent years, Arm has seen a big growth in business from smartphone OEMs, who are increasingly designing their own chips. These companies are probably less disconcerted by the deal. On the other hand, some of those companies already buy a lot of gear from Nvidia at high margins and may be uncomfortable to see a growing reliance on this supplier.

As we noted when we discussed the potential for this deal back in July, all the major semiconductor companies have projects working with the open source Arm competitor RISC V. We have to think that all of those projects get a major acceleration tomorrow. Arm’s technology is so central to the chip business that no customer can walk away tomorrow, it will take years to design out Arm processors, but of course that just makes the anti-trust argument all the more salient.

Nvidia touches on this briefly in the press release, pointing out the Arm will keep its brand and a separate headquarters, almost-but-not-quite-committing to keeping Arm’s licensing business independent, but they lump that comment in with their commitment to building up their presence in Arm’s original home in the United Kingdom.

All of this is a long way of saying that this deal is going to come under anti-trust and regulatory scrutiny. Anti-trust regulations are evolving in both Europe and the US, but this deal merits some attention.

A major question will be regulatory scrutiny from China. On the one hand, China has made conciliatory noises about anti-trust clearance since Phase One of the US Trade Deal was announced a few months ago. So maybe China fast tracks this. On the other hand, the term “AI” is a major trigger word in China’s tech ambitions, so painting this as an AI deal could become a problem. Arm moving to US ownership is also something that China’s technology strategists may (should) view with concern. Also, recall that Arm’s China business is operating in a state of near-anarchy right now, leaving open the question of how that situation could affect the Nvidia deal. (How much due diligence did Nvidia do on Arm China?) Put simply, China’s regulatory approval for the deal may prove challenging or come with some strings attached.

Put simply, Nvidia is buying a hugely important technology asset. And while they are paying a hefty premium for it, they have a valuable currency (i.e. their stock) with which to do a deal. Financially it looks like a great deal. That being said, our guess is that at least 30% of Arm’s revenue comes from direct or indirect Nvidia competitors, who are all now going to look for an exit. Add in another 30% of revenue from China, and there is the potential for the deal to look much more challenging over the long term.