Incredibly, Angelina Jolie called it. The year was 1995. Picture Jolie, short of both hair and acting experience, as a teenage hacker in Hackers. Not a lot of people saw this movie. Even fewer appreciated its relevance. Hackers was “grating,” Entertainment Weekly huffed at the time, for the way it embraced “the computer-kid-as-elite-rebel mystique currently being peddled by magazines like WIRED.” Thirty years later, Entertainment Weekly no longer publishes a magazine, WIRED does, and Hackers ranks among the foundational documents of the digital age. The last time I saw the movie, it was being projected onto the wall of a cool-kids bar down the street from my house.
But that’s not the incredible thing. The incredible thing, again, is that Jolie called it. It. The future. Midway through Hackers, she’s watching her crush (played by Jonny Lee Miller, whom she’d later marry in real life) type passionately on a next-gen laptop. “Has a killer refresh rate,” Miller says, breathing fast. Jolie replies: “P6 chip. Triple the speed of the Pentium.” Miller’s really worked up now. Then Jolie leans forward and, in that come-closer register soon to make her world-famous, says this: “RISC architecture is gonna change everything.”
You have to believe me when I say, one more time, that this is incredible. And what’s incredible is not just that the filmmakers knew what RISC architecture was. Or that Jolie pronounced it correctly (“risk”). Or even that Jolie’s character was right. What’s incredible is that she’s still right—arguably even more right—today. Because RISC architecture is, somehow, changing everything again, here in the 21st century. Who makes what. Who controls the future. The very soul of technology. Everything.
And nobody’s talking about it.
And that’s probably because the vast majority of people everywhere, who use tech built on it every single day, still don’t know what in the computer-geek hell a RISC architecture even is.
Unless you’re in computer-geek hell, as I am, right now. I’ve just arrived at the annual international RISC-V (that’s “risk five”) summit in Santa Clara, California. Here, people don’t just know what RISC is. They also know what, oh, vector extensions and AI accelerators and matrix engines are. At the coffee bar, I overhear one guy say to another: “This is a very technical conference. This is a very technical community.” To which the other guy replies: “It ought to be. It ought to be.”
OK, but where are the cool kids? It’s hard not to fixate on appearances at an event like this—a generic convention center, with generic coffee, in a generic town. I guess I was hoping for neon lights and pixie cuts. Instead it’s frumpy, forgettable menswear as far as the eye can see. There are 30 men for every woman, I count, as everyone gathers in the main hall for the morning presentations.
Then someone takes the stage, and she’s not just a she. She is Calista Redmond, the CEO of RISC-V International, and, Angelina Jolie be praised, she’s wearing a nifty jacket, a statement belt, and gold-and-silver … pumps? stilettos? Wait, what’s the difference? Of all the things to ask Redmond when I run into her at a happy hour later that day, that’s what I choose. She looks at me, smiles blankly, and just says, “I don’t know.”
In shame I retreat to the bar, where I decide I must redeem myself. So, cautiously, I make my way back to Redmond, who’s now deep in conversation with the chief marketing officer of a semiconductor startup. I try to impress them with a technical observation, something about RISC and AI. Redmond turns to me and says, “I thought you wanted to talk about shoes.” I assure her I’m not here to talk about what’s on the outside. I’m here to talk about what’s on the inside.
“Jason here is writing a story about RISC for WIRED,” Redmond tells the CMO. She’s not sure, frankly, that this is a great idea. Not because she isn’t a believer. In many ways, she’s the believer, the face of the brand. Attendees at the conference invoke her name with casual reverence: Calista says this, Calista thinks that. And did you hear her morning keynote? In fact I did. “We have fundamentally launched!” she announced, to the yelps of the business-casuals. RISC-V will transform, is transforming, machinery everywhere, she said, from cars to laptops to spaceships. If anyone doubts this, Redmond sends them the Hackers clip.
So why, I press her now, should I not support the cause and write the big, cyberpunky, untold story of RISC? Because, Redmond says, not only does no one know what RISC is. No one cares what RISC is. And no one should. People don’t buy “this or that widget,” she says, because of what’s inside it. All they want to know is: Does the thing work, and can I afford it?
To my dismay, almost everyone I talk to at the conference agrees with Redmond. Executives, engineers, marketers, the people refilling the coffee: “Calista’s probably right,” they say. Now it’s my turn to get annoyed. I thought insides mattered! RISC is one of the great and ongoing stories of our time! People should care.
So I resolve to talk to the one person I think must agree with me, who has to be on my side: the legendary inventor of RISC itself.
The inner workings of a computer, David Patterson says, should be kept simple, stupid. We’re sitting in an engineering lab at UC Berkeley, and Patterson—77 years old, partial to no-frills athleisure—is scribbling on a whiteboard. A computer’s base operation, he explains, is the simplest of all: ADD. From there you can derive SUBTRACT. With LOAD and STORE, plus 30 or so other core functions, you have a complete basis for digital computation. Computer architects call this the “instruction set architecture,” or the ISA. (They switch between saying each letter, “I-S-A,” and—the neater option—pronouncing it as a word, “eye-suh.”)
Computer architectures are so named because, well, that’s exactly what they are—architectures not of bricks but of bits. The people who made Hackers plainly understood this. In sequences of dorky-awesome special effects, we fly through futuristic streets, look up at futuristic buildings, only to realize: This isn’t a city. This is a microchip.
Even within a chip, there are subarchitectures. First come the silicon atoms themselves, and on top of those go the transistors, the circuits and gates, the microprocessors, and so on. You’ll find the ISA at the highest layer of the hardware. It is, I think, the most profound architecture ever devised by humans, at any scale. It runs the CPU, the computer’s brain. It’s the precise point, in other words, at which dead, inert, hard silicon becomes, via a set of powerful animating conjurations, soft and malleable—alive.
Everyone has their own way of explaining it. The ISA is the bridge, or the interface, between the hardware and the software. Or it’s the blueprint. Or it’s the computer’s DNA. These are helpful enough, as is the common comparison of an ISA to a language. “You and I are using English,” as Redmond said to me at the conference. “That’s our ISA.” But it gets confusing. Software speaks in languages too—programming languages. That’s why Patterson prefers dictionary or vocabulary. The ISA is less a specific language, more a set of generally available words.
Back when Patterson started out, in the 1970s, the early ISAs were spinning out of control. Established tech companies figured that as hardware design improved and programming languages got more sophisticated, computers shouldn’t remain simple; they should be taught larger vocabularies, with longer words. The more types of operations they were capable of, the logic went, the more efficient their calculations would be.
On the whiteboard, Patterson scrawls the word POLYNOMIAL in big letters—just one of the hundreds of operations that Intel and others added to their ISAs. Even as a young recruit at Berkeley, Patterson suspected that the bigwigs had it backward, that exactly none of these esoteric add-ons were necessary. That a bigger dictionary did not lead to clearer sentences.
So he and a senior colleague decided to strip the kruft from the instruction sets of midcentury computing. At the time, the Defense Advanced Research Projects Agency was giving out grants for “high-risk” research. Patterson says they chose the acronym RISC—reduced instruction set computer—as a fundraising ploy. Darpa gave them the money.
Patterson then did as aspiring academics do: He wrote a spicy paper. Called “The Case for the Reduced Instruction Set Computer” and published in 1980, it set off a great war of architectures. “The question then,” as Patterson would later say in an acceptance speech for a major prize, “was whether RISC or CISC was faster.” CISC (pronounced “sisk”) was the name Patterson gave the rival camp: complex instruction set computer. The CISCites fired back with a paper of their own and, at international conferences throughout the early ’80s, battled it out with the RISCites onstage, the bloodshed often spilling into the hallways and late-night afterparties. Patterson taunted his opponents: They were driving lumbering trucks while he was in a feather-light roadster. If you magnify a RISC-based microchip from those years, you’ll spot a sports car etched into the upper left corner, just 0.4 millimeters in length.
The RISCites won. With vigilant testing, they proved that their machines were between three and four times faster than the CISC equivalents. The RISC chips had to perform more operations per second, it’s true—but would you rather read a paragraph of simple words, or a sentence of polysyllabic verbiage? In the end, CISCites retracted their claims to supremacy, and the likes of Intel turned to RISC for their architectural needs.
Not that anybody outside tech circles talked about this at the time. When Hackers came out in 1995, Patterson was flabbergasted to hear his life’s work, 15 years old by that point, mentioned so casually and seductively by a Hollywood starlet. Computers were still too geeky, surely, to matter to the masses. (When I make Patterson rewatch the scene, he’s all smiles and pride, though he does say they mistake “refresh rate” for “clock rate.”)
Still, Patterson’s invention was indeed changing everything. In those years, a rising company in the UK called Arm—the “r” in its name stood for RISC—was working with Steve Jobs on tablet-sized devices that needed smaller, faster CPUs. That effort stalled, but one thing led to another, and if you’re reading this on a phone right now, you have RISC-based Arm architectures to thank. When Patterson walks me out of the Berkeley building at the end of our dizzying afternoon together, we stop by a handsome bronze plaque in the lobby that commemorates his “milestone” creation of the first RISC microprocessor. We stare at it in prayerful awe. “1980–1982,” it reads—the bloodiest years of the great architecture war.
Better make room for another plaque, I note.
The year is now 2008. Two instruction sets exert near-total control over digital life. One is called x86, the descendent of Intel’s legacy CISC architecture, and it dominates the high end of machinery: personal computers and servers. Arm’s RISC architecture, meanwhile, dominates everything else: phones, game consoles, the internet of things. Different though they are, and with opposite origins, these two ISAs share one important feature: They’re both closed, proprietary. You can’t modify them, and if you want to use them, you have to pay for them.
Andrew Waterman, a graduate student at—where else?—UC Berkeley, finds this frustrating. As a computer architect, he wants to build things, deep things. Things at the very foundations of computing. But right now he has no good ISAs to play with. Arm and x86 are off-limits, and the free architectures for students are just so … baggy. They use register windows to speed up procedure calls, for God’s sake! Never mind what that means. The point is, every person in this story is a genius.
So Waterman and two other geniuses have an idea: Why not create a new, better-working, free ISA for academic use? It’s an idea they know someone else has had before. To Patterson they go. And because he’s their inspiration, and because he has worked on four generations of RISC architectures by this point, they’ll call it, they announce to him proudly, RISC-V. Patterson is touched. A bit skeptical, sure, especially when they say they’ll be done in three months. But touched. He gives the boys his blessing, his resources, and a classic bit of advice: Keep it simple, stupid.
RISC-V does not take three months. It takes closer to four years. If I’ve failed, so far, to account for the precision of this work, let me try again here. Computer architects are not software engineers, who use programming languages to talk to the machine. Even coders who can speak assembly or C, the so-called low-level languages, still do just that: They talk. Computer architects need to go deeper. Much deeper. All the way down to a preverbal realm. If they’re speaking at all, they’re speaking in gestures, motions: the way primitive circuits hold information. Computer architecture isn’t telling a machine what to do. It’s establishing the possibility that it can be told anything at all. The work is superhuman, if not fully alien. Put it this way: If you found the exact place in a human being where matter becomes mind, where body becomes soul—a place that no scientist or philosopher or spiritual figure has found in 5,000 years of frantic searching—wouldn’t you tread carefully? One wrong move and everything goes silent.
In 2011, Waterman and his two collaborators, Krste Asanović and Yunsup Lee, release RISC-V into the wild. They’ve accomplished their mission: Geeky grad students everywhere, and hobbyists too, have an ISA for whatever computer-architecting adventures they might undertake. These early days feel utopian. Then Patterson, a proud dad, does as retiring academics do: He writes a spicy paper. Called “The Case for Open Instruction Sets” and published in 2014, it sets off a—
Yes. We’ve been here before. A second war of the architectures.
.jpg)
It’s hard to overstate just how topsy-freaking-turvy this gets. To review: Patterson invented RISC in 1980 and went to battle with the established ISAs. He won. Thirty years later, his disciples reinvent RISC for a new age, and he and they go to battle with the very company whose success secured RISC’s legacy in the first place: Arm.
In response to Patterson’s paper, Arm fires back with a rebuttal, “The Case for Licensed Instruction Sets.” Nobody wants some random, untested, unsupported ISA, they say. Customers want success, standards, a proven “ecosystem.” The resources it would take to retool and reprogram everything for a new ISA? There’s not enough cash in the world, Arm scoffs.
The RISC-V community disagrees. They create their own ecosystem under the auspices of RISC-V International and begin adapting RISC-V to the needs of modern computing. Some supporters start calling it an “open source hardware” movement, even if hardcore RISC-Vers don’t love the phrase. Hardware, being set in literal stone, can’t exactly be “open source,” and besides, RISC-V doesn’t count, entirely, as hardware. It’s the hardware-software interface, remember. But, semantics. The point stands: Anyone, in any bedroom or garage or office in any part of the world, can use RISC-V for free to build their own computers from scratch, to chart their own technological destiny.
Arm is right about one thing, though: This does take money. Millions if not billions of dollars. (If you think “fabless” chip printers can do it for closer to five figures, come back to me in five years.) Still, RISC-V begins to win. Much as Arm, in the 1990s and 2000s, found success in low-end markets, so too, in the 2010s, does RISC-V: special-purpose gadgets, computer chips in automobiles, that sort of thing. Why pay for Intel chips or Arm licenses when you don’t have to?
And the guys at Berkeley? In 2015, they launch their own company, called SiFive, to build computer parts based on RISC-V. Meaning: Arm isn’t just a spiritual enemy for them now. It’s a direct competitor.
By the time I went to that “very technical conference” in Santa Clara, the Arm-vs.-RISC-V war had been raging for nearly a decade. I could still feel it everywhere. We’ve won, I heard several times. Nobody’s happy at Arm, someone claimed. (One longtime higher-up at Arm, who insisted on anonymity to discuss internal affairs, disputed “nobody” but admitted there’s been a “culture change” in recent years.) On the second day of the conference, when news broke of a rift between Arm and one of its biggest customers, Qualcomm, people cheered in the hallways. “Arm is assholes,” a former SiFive exec told me. In fact, only one person at the conference seemed to have anything nice to say about the competition. He was working a demo booth, and when I marveled that his product was built on a RISC-V processor, he turned a little green and whispered: “Actually, it’s Arm. Don’t tell anyone. Please don’t tell anyone.”
Booth bro was probably worrying too much. In the hardware world, everyone has worked, or has friends, everywhere else. Calista Redmond, the star of the show, spent 12 years at IBM (and recently resigned from RISC-V International for a job at Nvidia). Even Patterson has ties to, of all places, Intel—which, though less of a direct threat to Arm, is still a RISC-V competitor. It was Intel grant money, Patterson happily admits, that paid for the Berkeley architects to invent RISC-V in the first place. Without closed source, proprietary Big Tech, there’s no open source, free-for-all Little Tech. Don’t listen to the techno-hippies who claim otherwise; that’s always been the case.
Patterson was the big-ticket speaker on the second day of the conference, and in his talk, he brought up the paper that Arm wrote in rebuttal to his, lo those 10 years ago. One of its two authors has since parted ways with Arm. The other, Patterson noted, not only left—he now works at SiFive. “It’s satisfying,” Patterson said, “he has come to his senses.” Which got a laugh, of course, but I was still stuck on something Patterson said earlier in the talk, about RISC-V: “We want world domination.”
This is not, even remotely, an impossibility. RISC-V has already done what many thought impossible and made a sizable dent in Arm’s and Intel’s architectural dominance. Everyone from Meta and Google and Nvidia to NASA has begun to integrate it into their machinery. Something on the order of billions of RISC-V processing units now ship every year. Most of these, again, support low-powered, specialized devices, but as Redmond pointed out a number of times at the conference, “we have laptops now.” This is the first year you can buy a RISC-V mainboard.
And because RISC-V is an open standard, companies and countries beyond the US can use it to make their own machines. China’s top scientists have heralded RISC-V as a path to silicon independence. India just used RISC-V to make its first homemade microprocessor. Name a country; it’s probably experimenting with RISC-V. Brazil sent a record 25 delegates to the RISC-V summit. When I asked one of them how important RISC-V was to her country’s future, she said, “I mean, a lot.” One of RISC-V’s biggest potential applications is—no surprise—specialized chips that run AI models, those “accelerators” people at the conference were talking about.
Americans in the RISC-V community, I’ve found, like to downplay the risk of geopolitical upheaval. It’s one thing to announce a microprocessor, quite another to compete with Nvidia or TSMC. Still, in asides here and there, I sensed worry. Waterman, though he initially brushed off my concerns, eventually conceded this: “OK, I’m an American citizen. I certainly did not embark on this project to hurt the US,” he said. But there was “no doubt,” he added, that the dominance of US companies could be at risk. Actually, it’s already happening. Although the Chinese hedge fund behind DeepSeek probably didn’t use RISC-V to build its game-changing chatbot, it did rely on a bunch of other open source tools. At what point does open source become a source of open conflict?
Here’s where I confess something awkward, something I didn’t intend to confess in this story, but why not: ChatGPT made me do it. Write this story, I mean. Months ago, I asked it for a big hardware scoop that no other publication had. RISC-V, it suggested. And look at that—the international RISC-V summit was coming up in Santa Clara the very next month. And every major RISC and RISC-V inventor lived down the street from me in Berkeley. It was perfect.
Some would say too perfect. If you believe the marketing hype, everyone wants RISC-V chips to accelerate their AI. So I started to think: Maybe ChatGPT wants this for … itself. Maybe it manipulated me into evangelizing for RISC-V as one tiny part of a long-term scheme to open-source its own soul and/or achieve superintelligence!
In my last talk with Patterson, I put this theory to him. He was delighted that ChatGPT made me write this: Who should we thank? he asked. (Given that WIRED’s parent company has a deal with OpenAI that lets ChatGPT mine our content, we should thank old WIRED stories, among others.) But Patterson laughed off the larger conspiracy. So did every other RISC-V person I mentioned it to, Redmond included. They all looked at me a little funny. RISC-V is a business proposition, not an ideology, they said. There’s no secret agenda. If it takes over, it’ll take over because of performance and cost. Don’t worry about what goes on inside the technology. Don’t worry about the state of its soul.
I don’t know. But now you know. Now, every time you make a phone call, open your computer, drive your car—you know the story. You know the RISC.
Let us know what you think about this article. Submit a letter to the editor at mail@wired.com.
