Today, something a little different: a joint post in the form of a conversation between Jon Metzler and Babbage.

Jon is Continuing Professional Faculty at the Haas School of Business, UC Berkeley, where he teaches on competitive advantage in technology markets, and also writes the Connect Substack. Babbage has been writing about technology history on The Chip Letter Substack for over four years now.

We’ve been mutual subscribers for a long time so it was wonderful to be able to meet up in person, on Babbage’s home turf, at the Science and Industry Museum in Manchester in the UK. To the backdrop of a reconstruction of the Manchester Baby we discussed an amazing range of topics, including the recent history of Arm. We decided to continue that conversation online, and this post is the first part of that conversation, focusing on the Softbank takeover of Arm in 2016.

Babbage: Summer 2016 was a momentous time in British history. The UK voted to leave the EU, and control of Arm, the UK’s leading (and arguably only major) semiconductor company, started to slip into the hands of Japanese conglomerate SoftBank following a $32bn takeover offer.

The reaction to the Arm deal in the UK fell into two camps. The new government, in the wake of Brexit, portrayed the deal as a vote of confidence in the UK. Many others though saw it as an indictment of the UK’s inability to properly value and hold on to its technology champions. Hermann Hauser who founded the firm that spun out Arm, and has since become a leading technology investor, called it a ‘sad day for technology in Britain’.

How was it seen on the West Coast, Jon? A big deal, or a storm in a British tea cup?

Jon: Thanks, Babbage. I remember the day of Brexit very clearly. Well, to be exact, I remember the day the results of the referendum were announced - June 24, 2016. The news itself was obviously momentous. (Also, I have family in the UK.) But also, on June 24, I went to a meetup of the Telecom Council of Silicon Valley, and a friend who worked at BT (British Telecom) greeted me by saying “Please don’t ask me about Brexit!” He’d likely spent a sleepless night dealing with the news. (The event was on Mobile Edge Computing, which was a very road-to-5G type topic. I don’t remember who presented that day. I just remember the Brexit news.)

Then SoftBank’s offer for Arm was announced. A friend in Washington DC texted me - what’s this about? Because of Japan’s low interest rates (effectively negative at the time), he thought SoftBank might simply be looking for higher yield. While that explains some M&A of the era clearly there was more to this.

Son-san from SoftBank did have a track record, albeit mixed, of turning some acquisitions into platforms. The old Vodafone Japan became SoftBank Mobile and became a content and service distribution platform for investments SoftBank made. SoftBank had made successful investments into Yahoo and Yahoo Mobile before that. SoftBank’s investment into Alibaba was one of the most successful venture deals ever.

But the synergy that a holding company that uses a mobile telco for distribution could realize from buying ARM was not obvious to me. Son-san and SoftBank have a long-demonstrated habit of spraying more funding at investments to see if they can grow faster. This was particularly dramatic in the SoftBank Vision Fund case. In the case of startup investments, this is contrary to the tranche-based investment model that historically characterized venture capital. And, some of SoftBank’s venture investments, most famously WeWork, lost discipline once they had unfettered access to capital. Sometimes, though, pouring resources at a business can work. And here, SoftBank was promising to double headcount. (Interestingly, Gambling Man provides reporting that Arm developed plans to raise headcount by 80%. And Masa raised that to “doubling headcount”. 80% was far too modest!)

Most acquisitions involve some “rationalization” - reduction of headcount, or at least integration and streamlining of back office functions. But Son-san flew out to England and promised to double ARM’s UK head count. How did that land? Did it seem genuine? Did it seem like a bet on Brexit? Or was it totally unrelated to Brexit, and just Son-san’s animal spirits at work1?

ARM was then public. On July 18, 2016, ARM’s board recommended that they accept SoftBank’s offer, which represented a 43% premium on the July 15, 2016, closing price. And, in the words of the company, was a “Great endorsement of UK tech”.

Babbage: That was certainly the line pushed by the UK government! Thanks for mentioning Gambling Man, Lionel Barber’s biography of Masayoshi San, which has lots of terrific behind-the-scenes detail on the takeover.

I think given the size of the premium offered by Softbank, the Arm board didn’t have much choice but to accept the offer. There was certainly no hint of anyone coming in with a higher offer. An attempt to block the deal on national security grounds would likely have failed and the Conservative government never held with the idea of ‘national champions’.

But that still leaves the question of how Son could justify such a premium. I don’t think the idea of a bet on Brexit makes any sense. As you say there were no obvious synergies with Softbank’s existing business. I do recall that Son said that owning Arm gave him an information advantage, one which might help Softbank when deciding where to deploy capital in future.

The biggest element though I think was that there was value in Arm that Son - perhaps guided by those ‘animal spirits’ - believed he saw, but which Arm’s existing investors, the wider market and possibly even Arm’s management hadn’t seen or felt they were unable to access.

Son was quite explicit at the time about what he thought Arm’s big opportunity was: the ‘Internet of Things’.

I think many were suspicious of Son’s commitment to invest heavily in Arm but, as I understand it, that commitment was legally binding, and Son did follow through on his promises. That meant that Arm saw significantly reduced profitability after the takeover as a result of that investment. I’m sure that had Arm’s management - who stayed in place after the deal - tried to spend in this way as a UK quoted company then they would have struggled to justify it to their investors. Actually, I think that had their case been based on the potential of the ‘Internet of Things’ then they would certainly have failed to win London based shareholders around!

I’m sure we’ll soon discuss how things turned out in practice, but what did you, and your contacts in the US and Japan, make of Son’s initial plans for Arm?

Jon: Let me back up a level. Or rather, first address the example of an earlier SoftBank acquisition. I remember when SoftBank bought Sprint. (Like with Brexit, I remember the date clearly - I was working for a different Japanese mobile operator at the time!) It was announced in 2012, and closed in 2013. I thought they had bought the wrong operator - they should have bought T-Mobile, in part for cultural reasons (the John Legere - Masa combo would have been great - imagine the Twitter bombs! Both are/were totally fine with being perpetually online and all that entails. More seriously, both are comfortable disrupting norms, and taking share in the process), and more importantly, because of common network technology and spectrum positions. The operational synergies with T-Mobile USA were very clear, and they weren’t in the Sprint case. (Apparently Masa just wanted a bigger carrier, and Sprint was bigger in terms of subscriber base at the time. So Softbank bought Sprint.)2

I remember contacts at Sprint were both apprehensive and excited. The apprehension came from being acquired by a Japanese network operator. What would having a Japanese parent company mean? The excitement came from the prospect of finally having some budget. Sprint had been cutting costs for years. No snacks at the office. No money for anything. That kind of forced frugality wears on employees. I share this because this was a case where SoftBank’s track record of spending perhaps worked in its favor from the perspective of getting staff to accept new ownership - they knew Sprint had a lot of challenges, and had been living with the frugality that came with those challenges for years, and so an aggressive parent who would spend seemed like a welcome change.

Sprint didn’t work out well - it really was a challenged company and there were no technology or spectrum synergies with SoftBank. Sprint was a CDMA operator that had bought an iDEN operator (Nextel) and then launched WiMAX. (That sentence almost hurts to type, but it’s true.) And SoftBank was a GSM/LTE operator with AWS spectrum. Absolutely nothing matched. But SoftBank made a yeoman’s effort at addressing Sprint’s network technology challenges. In the end, they sold their shares in Sprint to T-Mobile, the company they probably should have bought in the first place. The cultural and technology and spectrum fit was perfect.

(By the way, I cover technology standards and standardization in my Strategy for the Networked Economy class at UC Berkeley. I used to use Sprint as a cautionary tale on the consequences of poor standards decisions. The capture below is from a 2018 class.)

So if there’s an insight there, it’s that as an owner, SoftBank is willing to spend. And to invest staff resources into operating its acquisitions, as it did in the Sprint case. And also, it will sell to cut losses, eventually. Or maybe to fund something else Softbank is pouring resources into, as it did in selling NVIDIA shares (early). It funded a lot of other investments with Alibaba share sales for years.

With regards to Arm being acquired by SoftBank? Well, Arm was the company you worked with if you were going to develop a battery-powered device, or an embedded device. So it was both really important but also not that big, at least in revenue terms, relative to its importance. You could argue it under-indexed relative to its importance. If SoftBank acquired it, and resourced it, what would meaningfully change, say, for US customers? Would SoftBank change Arm licensing terms? It wasn’t a vertical integration so I couldn’t really see the risk that SoftBank owning it would drive off some of Arm’s customers (whereas NVIDIA owning Arm might have been problematic). Nor, however, could I really see the synergy with SoftBank’s other properties. Loosely, maybe, there’s some intelligence that comes with knowing how and where Arm libraries are being used. But one could probably get that intelligence just by talking to Arm regularly, or taking a smaller position in the company.

Looking back at Arm’s 2016q4 investor relations, they are indeed talking about IoT, servers, and 5G infrastructure, and automotive. And VR. Qualcomm was talking about those markets at that time too. IoT has never really been a “market” - it’s historically been a collection of smaller markets, with not that much synergy between categories.

Babbage: So the common theme is that Masa sees firms with potential upside from further significant investment but where their ownership on public markets means that they don’t have access to the capital required to make that investment.

Your reminder that Arm pre Softbank in 2015 was talking about IoT - and servers, automotive, 5G and VR - is fascinating. Looking at the comments in their last annual report as a public company, I sense a management that is enthusiastic about the opportunities and wants to invest but is also somewhat cautious, or at least wants to demonstrate a degree of discipline, whilst making the case for that investment. For example, here is CEO Simon Segars on IoT:

“Although we are confident that the Internet of Things is a significant opportunity for ARM, there are uncertainties too. We do not know how fast the Internet of Things will grow, how big it will eventually become, or who will be the market leaders. However, we would rather be investing in it now; creating technology that we believe will form a key component for many connected devices, and creating the new ecosystem of companies that will provide the products and systems from which these networks will be formed. We want ARM to be a market leader in the Internet of Things, and the only way to achieve that is to be fully involved.”

In other words we want to invest, but also recognise the uncertainty.

So Masa didn’t really bring a change of focus, but instead provided the capital to support management’s ambitions and encouraged them to be less cautious.

As an aside, I’m always interested in the way in which a company’s origins and history influence its culture. Arm started as a firm that had to exist on tiny royalties and I suspect that cost control and being frugal were part of the reason it survived and prospered. Switching to a more free-spending mode would have, perhaps, been somewhat counter-cultural.

If we widen the discussion beyond IoT then I think there is a clear case that Arm had potential to grow in markets like servers, automotive and 5G. The first Arm design used in a server was as far back as 2011 but it took almost a decade for the firm to build any sort of meaningful presence. I wrote about some of Arm’s early server efforts in Arm at Amazon:

AMD : Launched its Skybridge Opteron Arm server in 2014 only to drop the initiative a year later.

Cavium : Launched ThunderX server CPU in 2014. Followed with ThunderX2 in 2018.

Marvell : Acquired Cavium in 2018. Launched ThunderX3 Arm server design in 2020, only to exit the business later that year.

Qualcomm : Centriq designed launched in 2017 only to be shut down as Qualcomm faced a hostile, and ultimately unsuccessful, bid from Broadcom.

The common feature in all of these cases was that Arm was dependent on third parties who quickly got cold feet and closed the project down.

Even when Arm did start to get traction at AWS the design still used repurposed mobile cores in the form of smartphone-origin Cortex A-72 designs:

In November 2018, AWS announced its first Arm-based CPU cores, branded as ‘Graviton’ at its re:Invent conference. The first version of Graviton featured 16 Cortex A-72 based cores running the 64-bit ArmV8-A instruction set and running at 2.3 GHz. AWS offered a single Graviton instance type, the A1.

I think it’s hard to deny that there was latent opportunity - for example with more powerful cores that were specifically designed for servers - which a bolder strategy might have taken advantage of sooner. You also mentioned other revenue enhancing opportunities too that a less cautious management might have grabbed, for example increasing licensing fees.

Let’s pause here, and we can explore what happened next to Arm under Softbank’s ownership, next time.