Rene Haas on Arm’s Big AGI Chip Bet: ‘There’s Never a Perfect Time to Do Anything. You Just Have to Do It’
Rene Haas is half-propped on a couch in his San Jose, California office, a basketball cupped in his hand that partially obscures his face. When WIRED’s photographer first suggested this pose, Haas pulled an immediate grimace. The snarky headline wrote itself in his head, he tells me: “People are going to say Arm’s CEO sleeps on the job.”
Even so, he obliges. After carving out 46 minutes for our interview, he hurries us out to make a scheduled call with Masayoshi Son, SoftBank’s CEO and chairman of Arm’s board.
Our sit-down comes just days before the chip giant’s momentous announcement: Arm is building its own silicon. For a company that built its entire fortune licensing architectures to other chipmakers, never dipping into fabrication itself, this move is one of the biggest gambles in the firm’s history. Apple, Tesla, Nvidia, Microsoft, Amazon, Samsung, and Qualcomm all build or sell chips based on Arm’s IP, whether licensing full designs or paying royalties to the firm. By some estimates, there are three Arm-based chips for every person on Earth.
Looked at another way, though, launching an in-house chip marks a full-circle return to Arm’s roots. The company traces its origins to the late 1970s, when two computer architects founded Acorn Computers, which built a microprocessor based on the RISC architecture we know today. By the early 1990s, Acorn was flailing, and its then-CEO pivoted the business to licensing designs to outside firms. Jump to the mid-2010s, and Arm’s power-efficient mobile chip designs catapulted it to become the most valuable chip IP company on the planet.
Arm’s trajectory hasn’t always been smooth. After SoftBank acquired Arm and took the then-public company private in 2016, smartphone market growth cooled, forcing Arm to push aggressively into new lines of business. In 2020, Nvidia attempted to buy Arm, only for the deal to be blocked by global regulators. When that merger collapsed in 2022, Haas stepped into the CEO role. He took Arm public again soon after, though SoftBank still retains a 90 percent stake in the firm.
Haas joined Arm from Nvidia in 2013, where he led the computing product business unit, before eventually taking over Arm’s cash cow: the IP products group. Much like Nvidia CEO Jensen Huang, who often leans on his decades of industry perspective to frame current challenges, Haas brushes off concerns about today’s geopolitical chaos by referencing the upheaval of the 1980s—he’s not worried. He tells me he’s met former President Donald Trump half a dozen times, and doesn’t lose sleep over the U.S. government interfering in the UK-founded firm’s operations. Tall but not intimidating, he often wears heeled Saint Laurent boots, a blazer, and a Panerai watch for work.
Now 63, chip industry insiders describe Haas as a master networker who counts many of tech’s biggest names as close friends. The Wall Street Journal once labeled him a “natural-born diplomat.” But with this new chip project—one of Silicon Valley’s worst-kept open secrets—Arm and Haas risk alienating some of the company’s most loyal long-term partners. Can you stay close with customers after years of collaboration, only to announce you’re entering their market? Haas is convinced he can.
This interview has been condensed and lightly edited for clarity.
Lauren Goode: Since you became CEO, many observers say there’s been a major culture shift at Arm. Do you agree with that assessment?
Rene Haas: One thing I’ve learned—I understood this intrinsically when I worked for Jensen, but it really sunk in when I took the top job here—is that the CEO sets the entire tone for the company.
My leadership identity was shaped 30 years ago, when I moved to Silicon Valley, worked at a handful of startups, then joined Nvidia. The throughline across all those early roles was that I worked for founder-led companies. At the time, I couldn’t articulate that founder-led environments were what I resonated with most, but looking back, that’s where my professional DNA was forged, and where I thrive.
What does that environment look like exactly? It embraces risk, chases bold growth, moves fast, accepts mistakes, and is willing to place big bets. That’s exactly the culture I wanted to build when I took over here.
LG: Was this Arm silicon project already in the works when you became CEO?
RH: No.
LG: How much of this was your idea, versus Masa Son’s?
RH: Was it his decision? No. I’m the CEO. Was he fully aware and involved in discussing tradeoffs and all the different options we weighed? 100 percent. But this was more collaborative brainstorming—he acted as a thought partner, not a boss dictating “yes this, no that.” He doesn’t get into that level of detail.
One benefit of SoftBank holding 90 percent of Arm is that he’s our chairman and our largest shareholder, so I talk to him all the time. We’re pretty close.
LG: One analyst told me you talk to Masa 10 to 12 times per day.
RH: Some days. He does like to talk on the phone. I keep odd hours anyway, and he knows my personal schedule really well—when I wake up, when I work out, when I go to sleep. He’s very respectful of that.
LG: What made you decide this is the right time for Arm to build its own chip?
RH: Over time, we’ve evolved from just an IP company to a full compute platform company. What I mean by that is: when you look at the role a CPU plays in any ecosystem, there’s such deep interdependence between CPU hardware and the software ecosystem, whether that’s Windows, macOS, iOS, Android, or Linux. We didn’t always fully acknowledge or understand how important that interdependency was, but when I took over, I knew this was core to who we are, and it was something we had to advance.
So why build our own chip? When you’re a compute platform company, there are times when the entire ecosystem benefits from you building physical hardware yourself. We’ve seen this before: Microsoft builds Surface laptops to move the entire Windows ecosystem forward, even as HP, Dell, and Lenovo still build their own Windows devices. Google builds Pixel phones, while Samsung still builds top-tier Android devices.
LG: Those products are still a small sliver of their overall businesses.
RH: But they benefit the whole ecosystem. Think about Gemini optimized for Pixel and Android— that makes every Android device better, including Samsung’s. When Microsoft builds new tools or features for Surface that improve Windows, that benefits HP, Lenovo, and Dell too. That’s the part people don’t always get about us. They say “Oh, Arm’s just an IP company.” But the IP we deliver is a full compute platform, and platform companies have to do things to move their ecosystem forward.
LG: What’s the new chip called?
RH: It’s the Arm AGI CPU.
LG: AGI as in artificial general intelligence?
RH: That’s right. You like it?
LG: I don’t dislike it. I wonder if AGI is a term that will go out of fashion eventually, as the industry settles on new language for advanced superintelligence.
RH: I’m betting it won’t.
LG: Who is your target customer for this chip?
RH: The first customer is going to be Meta. But we also have SK Hynix, Cisco, SAP, Cloudflare. We have multiple customers lined up.
LG: And this is designed specifically for data centers.
RH: Data centers.
LG: What makes this chip special?
RH: A number of things, but first off, it’s unbelievably power-efficient. This company was born building chips to run off batteries in mobile phones, so efficiency is baked into our mindset. And everything we hear about AI growth centers on how much energy it requires. So delivering the world’s most power-efficient server CPU is all upside. This will be the most power-efficient data center CPU on the market.
The other thing it’s incredibly good at is running agentic AI. There’s a myth out there that with the rise of AI, the GPU or accelerator is everything. But when you look at how agents run inside data centers, that’s work only a CPU can do. GPUs don’t go away, but they need far more CPUs to run all those agents.
LG: Meta just signed two huge deals with Nvidia and AMD for both CPUs and GPUs, including Nvidia’s new Vera CPUs, part of the bigger Vera Rubin superchip.
RH: That's right. Which is Arm-based.
LG: Pardon my French, but does launching your own CPU piss off a customer like Nvidia?
RH: Yeah, so, back to the platform discussion … a rising tide raises all boats.
LG: That is a very convenient answer.
RH: But it’s not inaccurate. The more software that’s written for Arm, the more optimizations that are done for Arm—all of that helps anybody that builds on Arm. The opportunity is massive. When you look at how much money is tied up in data center and AI growth, it’s just enormous. To your question, does it piss off Nvidia? I’d imagine it will piss off Intel and AMD more than it pisses off Nvidia.
LG: Why is that?
RH: Because it takes market share from them.
LG: You’re referring to Arm architecture competing against Intel’s legacy x86. So you don't think you’ll piss off your pal Jensen, but AMD and Intel may have an issue with this.
RH: I use “piss off” as tongue-in-cheek. It's beneficial to the Arm ecosystem and it's beneficial to Jensen that we build a chip. If you've got [Nvidia’s] Vera chip, which is a great product, and you've got Arm AGI CPU, which is a great product, it's not great for Intel and AMD, that's all I know.
LG: Will Nvidia actually buy your CPU? Amazon execs appeared in a video at your event, vouching for your chip—will Amazon, which already has its own chips, buy yours?
RH: I’d love it if Jensen bought some, but the Vera CPU is already very closely coupled to his own full stack, and they have their own networking technology to connect all their chips. And Amazon will probably continue to use their own Graviton chip.
But we didn’t build this for Amazon, we built it because we think the market for Arm CPUs is really huge. It’s basically an underserved market. If you look at some of our early customers, like Meta, which is going to be using our chip in an air-cooled rack—they don’t need Nvidia’s NV Link. Or Cloudflare—they can’t buy Amazon’s Graviton chip, and they don’t have the resources to build their own.
LG: Who is fabricating your chip? Who are you working with for co-packaging and networking?
RH: TSMC. All TSMC. We’re also building a reference design of a server—we’re working with partners like Super Micro and Foxconn on that. In this day and age, doing a chip by itself isn’t enough. So we worked with ecosystem partners to deliver a full solution for a complete server rack.
LG: You are now going to have to answer questions about volume and yield and gross margins.
RH: Yes. When I joined Arm from Nvidia in 2013, I remember coming here and thinking, "Wow, there's no scrap, there’s no yield, there’s no returns, there’s no constant forecasting. You just hit Send and the royalty report comes in. What a wonderful business.”
So there’s a part of me that says, “Am I crazy for getting back into this?” But yes, all of that is stuff we’ll now have to manage, but I know how to do that. I know what’s required, and we’ve been building a team inside the company from an operations and go-to-market standpoint, everything that’s necessary to support that work.
LG: How many employees have you dedicated to this?
RH: We've added about 2,000 engineers into a group that does backend design, implementation, and subsystem work. They're not all exclusively on this chip, but we leverage a lot of the work that we do with our compute subsystems into this project. I don't know if I want to pin it down and say it's a 500-person or a 700-person project. It leverages a lot of work that we've already done.
LG: A lot of folks in the industry say that when you first launch a chip, it can take two to three generations before it has the impact that you're hoping for. Do you agree with that?
RH: Yeah. So, since I've become the CEO, we’ve been building what we call these compute subsystems. They essentially build all the building blocks of the compute engine that we can hand to customers for them to build their own systems-on-chips. In some cases, it's 85 percent of the work already done for them. It's a huge amount of pre-existing development.
The compute subsystem that the Arm AGI CPU is built on has already been shipped in silicon by other partners. So the confidence level that we have that this chip is going to work is extremely high. To your point, yes, it can take a certain number of iterations. That’s more true if you're getting into a brand-new market and you're trying to build something that's never been built before, never been tested, you don't really know how the market's going to accept it.
This one, our confidence level that the first one out of the chute is going to be very, very good and be a volume candidate is very, very high.
LG: Lord, give me the confidence of a chip CEO. OK. I'm going to do a little rapid-fire with you. I'll say a term, you give me a quick response. Intel.
RH: Historic.
LG: RISC-V.
RH: Nascent.
LG: Still nascent? It started in 2014.
RH: I said I’d give you one word. Nascent.
LG: I don't know if you happened to read the WIRED article that my colleague wrote last year about the history of RISC and the founding of RISC-V?
RH: Yes.
LG: And how some people in that community weren't very nice about Arm. They said you guys were assholes.
RH: Him or me? [he gestures at his chief of staff, Saumil Shah.] We could say such things about them.
LG: Back to rapid-fire: Sam Altman.
RH: Brilliant.
LG: What makes you think he's brilliant?
RH: I know him. He is a long, long-game type of guy. He really thinks quite big. He's quite visionary in terms of how he thinks about problems, how he thinks about technology. Yeah, I think he's brilliant.
LG: When you look at politics, geopolitics, the macroeconomic climate, and also the memory shortage—it's a crazy time to be launching a chip. How much has that changed your plans as you get ready to put this chip out into the world?
RH: None. Because this is a long journey for us. The Arm AGI CPU is the first product. This is just the beginning.
My first job out of college was with Texas Instruments in 1984. It was a boom year for the semiconductor industry. But the early 1980s were one of the worst recessions ever. Then in 1985 Reagan took office for the second term, and he cut defense spending, which was a big part of TI's business.
What was the lesson I learned about that? It's never a good time to do anything. Seriously. There's never a good time to do anything. There's never a bad time to do anything. You’ve just got to do it.
LG: I read that you’re a big fan of the famed basketball coach Phil Jackson. Why?
RH: What I love about Phil Jackson is that he took two great teams that had a lot of talent and turned them into champions. Michael Jordan was not a champion before—
LG: Michael Jordan won a title at North Carolina.
RH: He won the title when he was at North Carolina, but he wasn't the best player on the team. And Kobe Bryant was not a champion before Phil Jackson. He turned them into champions.
LG: Have you met Phil Jackson?
RH: No, I have not.
LG: How has this not happened?
RH: [nodding toward his chief of staff] Someone doesn't know how to do his job.
LG: What irritates you as a leader? Please don’t say “stupid questions.”
RH: People happy with the status quo.
LG: What's an example of that?
RH: “We've always done it this way. Not sure we could do any better. Can't see any other way to improve this.”
LG: The antithesis of risk. What inspires you as a leader?
RH: Making a big difference. Doing things that change the world.
LG: Everyone says that in Silicon Valley. What does that actually mean?
RH: Yeah, what does it actually mean? Well, if I distill it down to Arm, what inspires me is what we're doing with this chip. It's a big, big change for the company.
LG: Is this a bet-the-farm moment?
RH: No.
LG: So even if this chip doesn't work, you're still confident in your IP business?
RH: Absolutely. And the chip's going to work. It works.
Let us know what you think about this article. Submit a letter to the editor at [email protected].
