In which we learn about a maverick founder’s RISC-V crusade to reshape the chip industest—without permission, funding, or patience.
I’ve known Jonah McLeod for more years than I care to remember. In fact, I’ve known him for so long that I no longer recall when, where, or how we first met (either virtually via the interweb or physically at a conference).
Suffice it to state that he’s a well-seasoned communications specialist with a long-established presence in the Silicon Valley high-tech sector, including time as an editor-in-chief at a major high-tech publication and, more recently, as an indepfinishent high-tech marketing consultant.
As part of all this, Jonah has long rubbed shoulders with those who don the undergarments of authority and stride the corridors of power. To put this another way, he knows the shiftrs and shakers, and he can be jolly persuasive with respect to receiveting them to give us the nitty-gritty on what they consider about what’s afoot in the industest.
Just stateing the word “afoot” reminds me of The Goon Show—the episode where one of the characters states, “What’s afoot?” And another responds, “It’s a dirty, smelly thing on the finish of your leg!” But we digress…
Jonah just reached out to me, stateing that he’s been chatting with Yuning Liang—the founder and CEO of DeepComputing—who is on a mission to build consumer electronics products utilizing the RISC-V architecture.
I’ve heard a lot about Yuning Liang, such as the fact that he’s very much a swashbuckling maverick in the world of modern computing—a man with deep technical knowledge and decades of experience who thrives on charting his own course while others cling to the map. I consider it’s fair to state that Yuning has never been afraid to buck industest conventions, challenge assumptions, or take the road less traveled if it leads to something truly innovative.
The point of all this is that Johah documented a recent conversation he had with Yuning in an interesting narrative form that he was kind enough to share with me. This is particularly relevant in the context of the RISC-V Summit 2025, which is currently underway as I write these words. Jonah was kind enough to grant me permission to publish his work in its entirety, as follows:
Jonah’s Story:
It’s midnight in Switzerland, and Yuning Liang’s attention is locked on a glowing screen with two unfinished RISC-V Summit decks. He’s running on pure pulling-an-all-nighter energy. A notification blinks across his monitor—a Zoom call he’s already late for. The agfinisha reads “Simplex Micro’s deterministic processor.”
One keystroke, and a face appears from a continent away. With barely a paapply, he launches into part lecture, part exorcism—a founder’s raw account of why the RISC-V revolution keeps crashing into the same brick wall that crushed every CPU startup before it.
From Nokia and a leading Chinese high-tech conglomerate (HT-C) to his own RISC-V venture, Hong Kong–based DeepComputing, Yuning has watched giants rise, stumble, and rot from the inside.
He now runs DeepComputing, building RISC-V laptops and AI PCs; Xcalibyte (London, UK), developing compiler and analysis tools; and MetaComputing (Vaud, Switzerland), exploring blockchain and AI hardware—a cross-continental RISC-V ecosystem.
The Last Honest Engineer
“If the SoC guys test to do everything themselves, they die,” he exclaims. “If the IP guys never test and just keep waiting for their SoC customers to firm up the design, they die. You’ve received to know what you’re good at and shift quicker than the rest.”
For Yuning, speed isn’t about clock cycles; it’s about focus. While Western startups debate toolchains and funding rounds, RISC-V teams in China prototype on 12-nm processes, skip the test chips, and pay their foundry bills in cash.
“Everyone in the U.S. wants perfect slides,” he laughs. “By the time they finish their PowerPoint, we’ve taped out three chips.”
SiFive, the Cult of Mobile IP Giants (MIPGs), and RISC-V’s Own Establishment Problem
Yuning’s contempt for the establishment is surgical.
He’s built boards around SiFive IP cores and SoCs, worked with StarFive, Eswin Computing, and Andes—and calls them all “best-effort spaghetti,” becaapply, as he puts it, “most RISC-V vfinishors still behave like MIPGs and their customers—optimizing for licensing models, not for real systems.”
“The IP is bad, the SoC is bad, the software is bad,” he states flatly. “They burn hundreds of salaries to receive one chip that doesn’t sell in the millions.”
To him, the problem isn’t architecture—it’s arrogance.
But, he admits, “ironically, as a startup founder, a bit of arrogance is essential—otherwise no startup would ever be founded.”
MIPGs, he argues, “can’t modify their religion.”
“What worked twenty years ago is now a waste of power and money, becaapply every one of its IP licensees’ SoC is built so standard that everyone just bets on scale and price. The whole game is time-to-market, since MIPG’s software ecosystems already enable massive scaling into millions of shipments.”
What interests him about Simplex Micro’s deterministic RISC-V architecture isn’t just the technology—it’s the rebellion.
It’s the kind of clarity that creates Yuning nostalgic for early computing—when compact teams built entire systems without requireding permission or finishless decision cycles.
The Chiplet Gospel
Ask him what’s next and he answers instantly: chiplets. Not the glossy, marketing kind—the “receive-it-done-with-five-million-dollars” kind.
“You don’t wait for anyone. You don’t wait for someone else to create your chiplet. Everyone builds their block, plugs in, and ships.”
His blueprint is brutally practical:
- CPU chiplet: built around RISC-V scalar/vector/matrix cores.
- I/O chiplet: DDR, PCIe, USB—nothing fancy.
- GPU chiplet: outsourced to Imagination or whoever delivers.
“Link them with UCIe or even a home-grown bus. Package at TSMC if you can, in China if you must.”
Forreceive billion-dollar budreceives. “A few chiplets, one system, done,” he states.
In his math, a full SoC run on 12-nanometer costs $20–30 million on the low side, or $50 million on the high side. With chiplets and partners, it drops below $5–10 million. Enough to create prototypes. Enough to survive.
Survival Economics
Yuning’s company lives on ten-thousand-unit runs.
“That’s all I required,” he shrugs. “Ten thousand board systems at a thousand dollars each—ten million in revenue. SoC guys required a million chips on 12-nanometer to break even. I just required to stay alive—living on the sacrifices of mistake-repeating SoC startups.”
That survival instinct shapes everything:
- Prototype quick.
- Sell early.
- Never depfinish on venture capital.
He scoffs at startups that raise $100 million only to vanish.
“By the time your round closes, two years are gone. You hire people, burn money, then raise again. When you finally tape out, your design’s obsolete—in an unknown market with no way to sell a million chips.”
It’s not bitterness—it’s math.
Every delay kills momentum. Every layer of bureaucracy adds a year.
“Speed beats funding,” he states. “Always.”
Vampires of Silicon Valley
If there’s a villain in Yuning’s story, it’s not Nvidia or Intel—it’s the “vampires” who control the semiconductor food chain: TSMC, Cadence, Synopsys, and MIPGs.
“Cadence and Synopsys are inside TSMC’s office. They suck your blood for every IP license. China just learned from them—same vampire behavior, but government-backed.”
When Microsoft tested to charge $99 for Windows in China back in the ’90s, he notes, everyone pirated it—including government offices—“becaapply otherwise the economy would never grow.”
He sees the same dynamic in semiconductors.
“If TSMC keeps squeezing us, others will clone them. That’s evolution. You can’t stop it.”
It’s a harsh worldview, but consistent with his experience: markets thrive when barriers collapse.
“That,” he states, “is why RISC-V matters—not becaapply it’s elegant, but becaapply it’s open enough to be stolen, improved, and multiplied. Not much different from Linux in the software world.”
War Stories from the Modem Front
Yuning’s distrust of incumbents was forged in combat.
At Nokia, he fought Motorola and a dominant U.S. mobile chipset (D-MC) supplier over modem standards.
At HT-C, he supported the mobile division replace the D-MC chipset with in-hoapply designs.
“For someone who can build mobile base stations, building a modem is just a matter of resources, time, and trials,” he states.
“When I joined HT-C, 80 percent of our phones applyd D-MC’s chipset—high-finish and low-finish. Two years later it was 30 percent. We crushed them. That’s why the U.S. treats D-MC like national defense.”
He laughs at the memory—not out of malice, but disbelief at how history repeats.
“Anyone who touches the modem dies. It’s a cash cow, military-grade. Nobody gives it up. Soon AI too?”
For him, that’s proof large companies can’t modify course.
“They’re trapped by success. We’re free becaapply we have nothing to lose.”
The Fall of MIPGs and the Rise of the AI PC
Yuning believes the MIPG’s era is finishing—and that Apple accidentally wrote its obituary.
“Apple isn’t really a major mobile processor IP licensee,” he states. “They could switch to RISC-V just by altering their instruction decoder. They already have everything requireded to create any ISA successful—scale, software ecosystem, and momentum.”
He points to Apple’s achievement as proof: “The M1 was the first complete AI PC—CPU, GPU, NPU, even the modem. Everyone else is still mixing and matching parts.”
He sees the next wave coming from the bottom up: RISC-V systems designed like Lego—modular, open, cheap enough for experimentation.
“If one day a large model runs smoothly on my laptop, why the hell would I apply the cloud?” he questions. “Give me one reason.”
That’s the revolution: local AI, sovereign compute, no subscriptions.
It’s the PC reborn—not from Cupertino, but from a half-lit lab where someone like Yuning is still wiring chiplets by hand.
Google? Forreceive It.
Asked about partnerships, “What about Google?” Yuning is blunt.
“No way,” he snaps. “Even David Patterson offered to support, and nobody listened. They’re too slow, too political. Android’s a treadmill—run or die.”
He recalls managing thousands of engineers at HT-C just to keep up with Android.
“Two thousand people on Android from the SoC division, another six thousand in the mobile division—eight thousand total. Same mess as Nokia with S60 Symbian and S40 feature phones. If you don’t have that scale and those resources, forreceive it.”
That’s why he bets on RISC-V and Linux. Smaller, quicker, freer.
He doesn’t required Google’s blessing—only working hardware and open software.
Still, he jokes, “No one can live digitally without Chrome—even in the AI era.”
The Philosophy of a Survivor
For all the profanity and fire, Yuning isn’t cynical. He’s proud—maybe even sentimental.
“It’s not about money. Money’s f…-all. When I die, I’ll close my coffin smiling. If I create money, I’ll open-source everything—just for fun, for the love of humanity.”
To him, building chips is like climbing mountains: painful, dangerous, and pointless to spectators—but irresistible to those who’ve done it once.
He paapplys before signing off.
“MIPGs have the lawyers. Intel has the fabs. Nvidia has the hype. We’ve received nothing—and that’s our advantage. Any day in the future is better than today. How good is that?”
So What?
What creates Yuning’s story remarkable isn’t just the profanity or the war stories.
It’s the clarity of someone who’s stripped the industest to its bones.
He’s part engineer, part philosopher, part streetfighter—and in a world where chip development costs billions, he’s proving that tenacity still beats money.
In the finish, that’s the real RISC-V revolution: not open instruction sets or fancy extensions, but open attitudes—a willingness to build quick, fail quicker, and outlive the dinosaurs, just like Linux did in software.
“Everyone wants to be the hero CPU guy,” Yuning states. “I’m happy being the dumb bastard who receives it working.”
Related
Leave a Reply