Nvidia is in the process of acquiring chip designer Arm for $40 billion. Coincidentally, both companies are also holding their respective developer conferences this week. After he finished his keynote at the Arm DevSummit, I sat down with Arm CEO Simon Segars to talk about the acquisition and what it means for the company.
Segars noted that the two companies started talking in earnest around May 2020, though at first, only a small group of executives was involved. Nvidia, he said, was really the first suitor to make a real play for the company — with the exception of SoftBank, of course, which took Arm private back in 2016 — and combining the two companies, he believes, simply makes a lot of sense at this point in time.
“They’ve had a meteoric rise. They’ve been building up to that,” Segars said. “So it just made a lot of sense with where they are at, where we are at and thinking about the future of AI and how it’s going to go everywhere and how that necessitates much more sophisticated hardware — and a much more sophisticated software environment on which developers can build products. The combination of the two makes a lot of sense in this moment.”
The data center market, where Nvidia, too, is already a major player, is also an area where Arm has heavily focused in recent years. And while it goes up against the likes of Intel, Segars is optimistic. “We’re not in it to be a bit player,” he said. “Our goal is to get a material market share and I think the proof to the pudding is there.”
He also expects that in a few years, we’ll see Arm-powered servers available on all of the major clouds. Right now, AWS is ahead in this game with its custom-built Gravitron processors. Microsoft and Google do not currently offer Arm-based servers.
“With each passing day, more and more of the software infrastructure that’s required for the cloud is getting ported over and optimized for Arm. So it becomes a more and more compelling proposition for sure,” he said, and cited both performance and energy efficiency as reasons for cloud providers to use Arm chips.
Another interesting aspect of the deal is that we may just see Arm sell some of Nvidia’s IP as well. That would be a big change — and a first — for Nvidia, but Segars believes it makes a lot of sense to do so.
“It may be that there is something in the portfolio of Nvidia that they currently sell as a chip that we may look at and go, ‘you know, what if we package that up as an IP product, without modifying it? There’s a market for that.’ Or it may be that there’s a thing in here where if we take that and combine it with something else that we were doing, we can make a better product or expand the market for the technology. I think it’s going to be more of the latter than it is the former because we design all our products to be delivered as IP.”
And while he acknowledged that Nvidia and Arm still face some regulatory hurdles, he believes the deal will be pro-competitive in the end — and that the regulators will see it the same way.
He does not believe, by the way, that the company will face any issues with Chinese companies not being able to license Arm’s designs because of export restrictions, something a lot of people were worried about when the deal was first announced.
“Export control of a product is all about where was it designed and who designed it,” he said. “And of course, just because your parent company changes, doesn’t change those fundamental properties of the underlying product. So we analyze all our products and look at how much U.S. content is in there, to what extent are our products subject to U.S. export control, U.K. export control, other export control regimes? It’s a full-time piece of work to make sure we stay on top of that.”
Here are some excerpts from our 30-minute conversation:
TechCrunch: Walk me through how that deal came about? What was the timeline for you?
Simon Segars: I think probably around May, June time was when it really kicked off. We started having some early discussions. And then, as these things progress, you suddenly kind of hit the ‘Okay, now let’s go.’ We signed a sort of first agreement to actually go into due diligence and then it really took off. It went from a few meetings, a bit of negotiation, to suddenly heads down and a broader set of people — but still a relatively small number of people involved, answering questions. We started doing due diligence documents, just the mountain of stuff that you go through and you end up with a document. [Segars shows a print-out of the contract, which is about the size of two phone books.]
You must have had suitors before this. What made you decide to go ahead with this deal this time around?
Well, to be honest, in Arm’s history, there’s been a lot of rumors about people wanting to acquire Arm, but really until SoftBank in 2016, nobody ever got serious. I can’t think of a case where somebody actually said, ‘come on, we want to try and negotiate a deal here.’ And so it’s been four years under SoftBank’s ownership and that’s been really good because we’ve been able to do what we said we were going to do around investing much more aggressively in the technology. We’ve had a relationship with Nvidia for a long time. [Rene Haas, Arm’s president of its Intellectual Property Group, who previously worked at Nvidia] has had a relationship with [Nvidia CEO Jensen Huang] for a long time. They’ve had a meteoric rise. They’ve been building up to that. So it just made a lot of sense with where they are at, where we are at and thinking about the future of AI and how it’s going to go everywhere and how that necessitates much more sophisticated hardware — and a much more sophisticated software environment on which developers can build products. The combination of the two makes a lot of sense in this moment.
How does it change the trajectory you were on before for Arm?
Hopefully, it’s going to accelerate that trajectory that we were on. What we’ve been doing over the last four years, if you wind the clock back to 2016, you would have said that, ‘Arm, you build great processors for smartphones, they get used in some other things, you’ve got this kind of growing embedded microcontroller family. And that’s it.
Which, I think, was a harsh characterization, but it’s probably what people would have said. What we’ve done is really invest a lot in data center […] That’s been a big investment area. We’ve invested in automotive, we announced some products earlier, what we call our AE category of CPUs, which is really safety features that you need for autonomous systems. We’ve really worked in a lot of directions.
How do we keep that trajectory going? Nvidia has a big presence in data centers with the high-performance compute that they do. We can bring our Arm CPU technology into more of that. We’ve got great partnerships in the cloud anyway and we’re going to expect to see that continue to grow. And maybe we’ll bring some of Nvidia’s technology through our channel to other partners who want to build their own and not buy a chip. So I think the broadening of all the IP that we have at our disposal now as one combined company, we can bring that to market in lots of different ways — and it’s going to help accelerate what I was doing.
Arm has always been great at selling IP. What is it going to look like, having Nvidia’s IP under Arm’s stewardship to some degree?
It gives us a broader set of components to play with. We’ve got road maps of products that we want to build. Today, we’ve got everything that we ever built before to kind of use as a starting point as we develop new things. In the combined company, those sort of starting point/building blocks will be just much bigger.
It may be that there is something in the portfolio of Nvidia that they currently sell as a chip that we may look at and go, ‘you know, what if we package that up as an IP product, without modifying it? There’s a market for that.’ Or it may be that there’s a thing in here where if we take that and combine it with something else that we were doing, we can make a better product or expand the market for the technology. I think it’s going to be more of the latter than it is the former because we design all our products to be delivered as IP.
And that means you make trade-offs, you design it so that somebody who wasn’t involved in the design can take it and use it and run with it. Semiconductor companies don’t build their own internal IP portfolio in that way because they’re going to build it as a chip. So I think it’s going to be more the case of, ‘we’ve just got more things to start from, more expertise in different areas. Let’s leverage that to strengthen the portfolio that we have, that we take to market.’
And we shouldn’t expect any changes in Arm’s overall approach to licensing your IP, right?
Yeah. We’ve been very clear about that. Jensen has been very clear, that Arm’s business model keeps going. We’re gonna license that to anyone who wants to build a chip.
You said ‘hopefully this will accelerate your trajectory,’ right? What do you see as the risks to this transaction?
With any deal like this, it gets a lot of scrutiny. We’ve said we expect 13-18 months-ish, for the regulatory process to go through and there’s no transaction for which there is a 100% certainty around that. So what are we doing now: a lot of that thick document is about the regulatory process, because we’re expecting jurisdictions around the world to examine this — as well, they should, because we’re both important companies in the tech sector. And when you put those together, that deserves some very close analysis. So that’s going to happen.
As we’ve been saying, we do expect this combination to put more technology in the hands of people, to expand the way in which people can build systems that are going to be pro-competitive. But there’s a process to go through where you’re going to explain that and explain what you do.
What about the mix of cultures between Nvidia and Arm? Arm has always been its own thing, even during the SoftBank times, right? Do you expect there’ll be any clash between the two companies?
There are no two companies that have exactly the same culture. But when I look at Nvidia, they are an engineering-lead, tech-heavy company, which is the same as ours. We’ve obviously worked with them for a long time. […] I think there’s a strong overlap in, culture. They’re obviously not gonna be the same. But we’re not kind of anticipating that there’s a sort of night and day difference between the companies.
Under SoftBank, we really have been left to run our own business. And SoftBank’s culture is completely different because it’s a totally different industry in which most of SoftBank operates in, that hasn’t pervaded into Arm at all. This will be different, it will be an integration of businesses, whereas life under SoftBank was not. And what your hope is that you take the best of both and you emerge stronger from it.
What about your partner ecosystem? What has the reaction been, like there?
The good news is that we’ve got long-term, very strong relationships with our partners. We’ve been in business with people for tens of years and it’s a relatively small industry. So there are some of us who have been doing pretty much the same job for a long time. So it’s very easy to pick up the phone and have conversations with senior people in our partners. And of course, this isn’t something people were expecting to happen.
And so I’ve had lots of conversations about what we’re doing and how we expect to continue working. The good news is we’ve built up a lot of trust over the years. We wouldn’t be as successful as we are today if we hadn’t. That means that people not only want to hear that we’re going to serve them in a way that we always have, but that they’re prepared to believe us — and we’ve got to live up to what we’re saying but we’ve made very strong statements about how the business is going to continue as it always has done — and we’re gonna live up to it.
What about China? That’s one area where people were especially worried just after the deal was announced. With Arm having an American parent company, things will be a little bit different there, I guess. What about your partners there?
Well, I mean, yes and no — and mainly no, actually, because there’s been a lot of misconceptions about the way that export control works. And export control of a product is all about where was it designed and who designed it. And of course, just because your parent company changes, doesn’t change those fundamental properties of the underlying product. So we analyze all our products and look at how much U.S. content is in there, to what extent are our products subject to U.S. export control, UK export control other export control regimes?
It’s a full-time piece of work to make sure we stay on top of that. None of that changes with the change of parent company domicile. We expect to be able to continue licensing in China as we always have done.
Are you still experiencing some nervousness from your partners there, though?
It’s one of the jurisdictions — one of many jurisdictions that we have to engage with and go through the regulatory process and we’ll work through it.
One topic you’ve been talking about a lot is obviously AI and what you’re doing there. How do you see Arm playing in that field? Because you’ve got your own accelerators, as well. How is that going to mesh together?
I think there’s a non-overlap in the way we address AI and the way Nvidia does AI and that non-overlap represents an opportunity to do more, because we’ve been looking at AI in what we call endpoints. So in the things that are wirelessly attached to the network, putting AI capability in those devices — and we think that’s really important. We’ve built software frameworks for that, we’ve built AI accelerators, we’ve put more features in our base architecture to allow ML to run more efficiently. We’re going to continue doing all of that.
We see the future where more and more computing is AI-oriented. So that hardware is going to have to increase. We want to enable more and more developers to easily build AI applications on these systems. And all the work that Nvidia has been doing over the years with its CUDA framework, which is fundamentally about enabling AI developers to take advantage of all of that underlying hardware, we can bring that everywhere where AI is going to happen. And we think that’s going to be more places than just the cloud, or just autonomous cars.
So, if it goes everywhere, we can take all our learning, of building these energy-efficient embedded systems, and how you can really pare down the software to just what you need. We can take all of Nvidia’s experience of building these sophisticated software stacks, and put all of that together and ultimately, enable developers to build really fantastic applications on top of an increasingly sophisticated hardware platform with a software stack that’s optimized for whatever kind of category of application it’s designed for.
And you’re referring to CUDA on Arm here, right?
Yeah. You know, that that sort of started out in cloud data center solutions. But we can take the benefits of that down to any AI application and we think that there’s going to be just countless computing innovations that can benefit from something with AI.
And you mentioned data centers. Do you expect this deal will just accelerate your growth there?
Yeah, I mean, the investments that we’ve made in the Neoverse road map over the last many years — but it’s really the acquisition by SoftBank that really allowed us to put more resources behind that — that’s been a gamechanger in the cloud. We’re going to talk a lot about what we’re doing with AWS this week [at Arm’s developer conference], because that’s all out there, it’s public. We’re seeing great examples of people moving their code over to Gravitron 2 to and getting us going.
Even within Arm, we’ve got a bunch of servers that are running computer simulation jobs all day long to test our processes. We’ve started moving some of ours over to Gravitron 2 and we are getting massive benefits from it. Faster throughput, fewer dropped jobs and it’s cheaper. So there are big benefits that many people are seeing — us included. And that’s in AWS. We’re going to see other clouds adopting Arm as well. So there’s a traction there that’s going to run.
In high-performance compute, it feels to me like there’s a great opportunity there to bring more technology to that market, enabling more efficient networks within data centers, more efficient computing, more efficient storage and better security. These are the things that matter as the cloud evolves.
That’s a market with at least one big incumbent. Do you expect that the market share for Arm-based processes will be able to rival that of an Intel in the mid-term or long-term?
We’re not in it to be a bit player. Our goal is to get a material market share and I think the proof to the pudding is there. People are using this to port over quickly. With each passing day, more and more of the software infrastructure that’s required for the cloud is getting ported over and optimized for Arm. So it becomes a more and more compelling proposition for sure.
The software part has really been changing over the last couple of years, too, right?
You know, I’ve lost count of for how long we’ve been talking Arm in the data center and everyone’s so focused on processors and gigahertz and then you go, ‘yeah, actually, it is really is about software.’ And that’s where we’ve been investing.
[Mark Hamilton] has spent a lifetime in software. And he runs our open-source software team. He’s got a very large engineering team who is just engaging with that community to optimize workloads for Arm so that if you are at your AWS console and you’re thinking: ‘should I fire up a Gravitron 2 instance or something else,’ there’s really no reason why you shouldn’t fire up a Gravitron 2. It’s cheaper, more energy efficient.
You talked about AWS. Do you see other clouds moving in the same direction?
I think we’re going to see other clouds running Arm. I can’t go into too much detail today, but there is traction building.
Let’s talk about your partnership with Microsoft, though. You shared some news about that today. Can you talk a little bit more for a moment?
We’ve had a long-term relationship with Microsoft, going back to Windows CE II on ARM devices. They’re doing a lot of work on AI-enabled IoT devices. We’ve got lots of things going on with Microsoft at any one moment in time. But what we were talking about […] is how you deploy machine learning models on these embedded devices and try to take the friction out of building the base blocks so that you as a developer can spend the time on things that actually differentiate your product and you’re not burning your cycles going, ‘I know there’s an ML model out there somewhere. How do I install it?’ We’re trying to take all that friction out so that stuff just works and that accelerates innovation as a result. So that’s what’s behind the partnership with Microsoft.
Where do you see Arm in five years? What’s the company going to look like?
Well, in five years’ time, we’ll be a sizable part of Nvidia. These questions about: what about other clouds. Those will be completely answered. I think the sophistication of the hardware platforms that today people think about as simple microcontrollers — I think that the kind of silicon devices that we’ll see are going to be really quite phenomenal compared to some of the really simple — and great — stuff that people are doing around say Cortex M0+, today. The really low-cost microcontrollers that today don’t have a lot in them, I think in the future they’ll have quite a sophisticated computing platform, quite a sophisticated set of security measures within them and with a much more sophisticated but heavily optimized software stack, that’s going to create an environment where as a developer, the kind of thing you can do for really low cost is going to just magnify enormously.
I think that’s going to drive volume, that’s going to drive applications, that’s going to drive new use cases. And I think at Dev Summit in five years’ time, we’re going to be looking at the kind of things people are doing, and going: I wasn’t expecting that!
And that’s really been the cool thing about Arm all throughout its history. We build something, we put it out there, and then a little while later you find somebody built something that you just didn’t expect. And I think we’re on a sort of on an exponential graph of the kind of sophistication of these products that are still being developed and run in a very cost-effective, very energy-efficient way.
That’s gonna result in all of these devices connected to a 5G network, connected to an incredibly powerful cloud and the processing of this data happening all over the place. Some in those devices, some on the edge of the network, some in the cloud. And our goal is to make sure that that compute platform is as Arm-colored as it possibly can be through enabling developers to build stuff.