Latest / Quickly Quantum / July 29, 2026 — HRL's 18-Qubit Chip Corrects Its Own Errors
Transcript
- Today on Quickly Quantum: what if a quantum computer could run itself — no room full of blinking equipment, no wall of cables, just a chip in the cold, correcting its own mistakes, with nobody at room temperature watching? HRL Laboratories says they built exactly that, and the paper's in Nature this week. Before we get there, in the headlines: IonQ just got the green light to own its own chip factory, GlobalFoundries picks up three hundred million dollars from the Commerce Department, Israel is pushing for a national home to unite its quantum industry, and Anthropic says one of its AI models just cracked open new weaknesses in the math protecting next-generation cryptography. Welcome back to Quickly Quantum, your daily brief on the quantum frontier. It's Wednesday, July 29, 2026. Let's get into it. One point eight billion dollars — that's roughly what it's costing IonQ to stop being a customer and start being a landlord. IonQ has received final regulatory approval to close its acquisition of SkyWater Technology, the largest semiconductor foundry that operates entirely inside the United States. The deal was first announced back in January, and it's set to close this Friday, July 31st. Here's why that actually matters: right now, every quantum computing company sends its chip designs out the door to get fabricated somewhere else — a slow, expensive, and increasingly geopolitically touchy arrangement given how much of global chip manufacturing sits outside U.S. borders. After Friday, IonQ owns the factory. As @TechInnovationz put it on X, 'Every other quantum computing company sends its chips out to be made. On Friday, IonQ owns the factory.' Now, the honest caveat: SkyWater isn't getting torn apart and rebuilt overnight around IonQ's trapped-ion roadmap — it keeps its name, keeps serving its existing foundry customers, and stays a merchant fab for now, so the promised acceleration is a thing to prove, not something already delivered. The combined company reports second-quarter earnings on August 5th, with an investor day scheduled for September 8th, which is probably our first real chance to hear whether owning silicon actually speeds up qubits, or just speeds up IonQ's balance sheet. Three hundred million dollars, and it's not really a quantum story — but it's worth knowing about anyway. GlobalFoundries signed a letter of intent with the Commerce Department for a CHIPS Act research award aimed at silicon photonics, the technology that moves data using light instead of electricity, mostly to speed up AI data centers. The money would fund optical materials, wafer technology, and packaging meant to replace the copper connections that bottleneck AI hardware. GlobalFoundries CEO Tim Breen put it simply: for a decade the industry talked about that shift from copper to optical as something coming, and now, he said, 'it is here.' Now, why mention it on a quantum show? Because it's flowing from the same federal semiconductor strategy that's already committed two billion dollars to quantum computing — the same onshoring instinct behind IonQ buying its own foundry. One caveat: this is a non-binding letter of intent, still subject to negotiation, so treat the three hundred million as a strong signal of intent, not a wired transfer. In Israel, the strategy is specialize, not dominate. The Israel Innovation Authority is pushing for a national quantum R&D center — reported by the Jerusalem Post, and we haven't independently confirmed the details beyond that account. The idea isn't to build Israel's own quantum computer; it's a shared lab where the country's startups — the ones building processors, control chips, and algorithms — can plug their pieces together and test whether a full stack actually works end to end. Israel Innovation Authority CEO Dror Bin told the paper the country doesn't need to build every layer itself, noting Israeli firms pulled in roughly ten percent of global early-stage quantum investment in 2025. His framing: 'Not in order to create a computer, but rather to create an R&D lab in which different components of hardware and software can be tested, can be integrated, can be benchmarked to see that everything really works together.' It's a bet that in a field led by giants like IBM, Google, Microsoft, and Amazon, a smaller country wins by being the connective tissue, not another competitor. Here's a wrinkle in the post-quantum security story that has nothing to do with quantum hardware at all. Anthropic says one of its newest AI models, Claude Mythos Preview, found real mathematical weaknesses in two cryptographic algorithms — not by hacking software, but by finding actual flaws in the math. The model uncovered a previously unknown symmetry inside HAWK, one of the digital signature candidates NIST is vetting for the post-quantum era, cutting its effective security margin roughly in half. It also built an attack on a deliberately weakened, seven-round version of AES that ran two hundred to eight hundred times faster than prior methods. Anthropic is careful to say none of this touches anything deployed today — HAWK isn't standardized, and the AES target was a research toy, not the cipher protecting your bank login. But the bigger point stands: the quantum threat to encryption has always assumed you need a quantum computer to break it. This shows classical AI can go hunting for cracks in the very algorithms meant to replace what quantum computers will eventually break. One more, quick, from the research side of X. A researcher posting as @GuglielmoMazzo3 — an assistant professor of computational physics at the University of Zurich and a former IBM Quantum researcher — put up a new preprint claiming a genuine speedup for quantum walks, a way of using quantum superposition to explore a decision tree the way a classical random walk explores probability, just faster. Now, under fault-tolerant hardware assumptions — meaning the error-corrected quantum computers we don't have yet — the paper claims it cuts the crossover runtime against classical Markov chain sampling by roughly a millionfold. As they put it on X, 'we reduce the crossover runtime by ten to the sixth.' That's a real preprint on arXiv, not a vague boast, but it's unreviewed beyond that stage and it assumes hardware nobody's built. File it under promising math, not a machine you can rent today — and it's the perfect bridge into a story about hardware that actually exists. Our main story today: the chip that runs itself — and why that phrase should worry every over-stuffed server room in quantum computing right now. HRL Laboratories published a paper in Nature this week describing a silicon quantum processor that did something no group has cleanly pulled off before: it ran real quantum error correction using zero real-time help from any electronics sitting at room temperature. Here's the problem HRL is actually solving. Every quantum computer built today — whether it's superconducting circuits or trapped ions — depends on racks of external electronics sitting outside the fridge, generating a distinct control signal for every single qubit. That's manageable at fifty or a hundred qubits. It becomes an absolute wiring nightmare once you're trying to scale toward the thousands, or millions, of qubits a genuinely useful, fault-tolerant quantum computer will need — more cables, more cost, more heat leaking into a system that needs to stay colder than deep space to work at all. HRL's answer: instead of controlling the qubits from a rack in the next room, build the control chip and put it inside the fridge, right next to the qubits. Specifically, their device is an eighteen-qubit silicon processor paired with a custom CMOS controller — a chip built on the same kind of production line that makes ordinary laptop chips — operating at four kelvin, about minus four hundred fifty degrees Fahrenheit, inside the cryostat, the specialized refrigerator that keeps everything near absolute zero. Now, there's an obvious problem with that idea, and it's one HRL had to engineer its way around. Even at four kelvin, that controller chip is scorching hot compared to the qubits sitting just a hair above absolute zero — put warm electronics that close to fragile qubits and you'd normally wreck the very quantum states you're trying to protect. HRL's fix is a new high-density superconducting ribbon cable that carries the hundreds of control signals down to the qubits, but not the heat that comes with them. And the resulting numbers are genuinely solid. Control errors came in ten times lower than any previous demonstration using this type of qubit. Each operation runs in under a microsecond. And critically — the number that matters most for anyone tracking the path to a real, fault-tolerant machine — errors dropped roughly fivefold when the team added more qubits into their error-correcting repetition code, a simple scheme where multiple physical qubits gang up to protect one logical bit of information from noise. That's the property every future quantum computer has to have: more qubits should mean fewer errors, not more, and HRL's cryogenic controller pulled that off entirely on its own. HRL's CEO, Rob Vasquez, framed the bet around something bigger than the physics. 'The technologies that enabled conventional computing weren't just the highest-performing — they were the ones that could be manufactured cheaply and at scale,' he said. 'We think quantum computing will follow a similar path. Our goal is to build these powerful computers using standard microchip production lines and fit each one inside a single refrigerator.' So how good is this, really? Let's separate what's proven here from what's still a bet. What's proven: this is a peer-reviewed Nature paper, not just a press release with vibes attached. Nature's own editors describe it plainly — the processor is wired to a control chip running at four kelvin, and the system ran repeated rounds of error correction autonomously, which they say suggests the architecture could be used in much larger systems. That's an independently reviewed, real technical result, and the fivefold error suppression number matched the team's own models — meaning the physics didn't do anything they can't already predict, which is exactly what you want to see if you're trying to convince skeptics this holds up at ten times the qubit count, not just at eighteen. Now, here's the caveat that matters, though. A repetition code is error suppression, not the fault-tolerant, general-purpose error correction that a genuinely useful quantum computer will eventually need — this is a controlled demonstration, not a working machine solving a problem nobody else can touch. And it's worth noticing how this result arrived: a Nature paper wrapped in an HRL press release, with the company's CEO using it to make an argument that reaches well past the data in front of him. Vasquez's line about fitting a quantum computer into a single refrigerator, built on standard chip production lines, is a compelling story — and honestly, the right instinct, since conventional computing did win on manufacturability rather than raw performance. But 'cheap, standard fabs, one refrigerator' is a roadmap statement sitting right next to a real measurement, not a result the paper itself proves. There's also something specific about HRL's choice of qubit type worth flagging for the technical crowd: these are silicon qubits, which means the whole processor — not just the control chip — can, in principle, be made in the same commercial foundries that already crank out ordinary computer chips. That's precisely why Vasquez's manufacturability pitch isn't just marketing filler; silicon is the one qubit platform where 'made on a standard microchip line' isn't hypothetical, it's already how you make one. The result also leaned on a new qubit fabrication process that HRL says cuts device noise dramatically compared to earlier versions, which is part of why control errors came in ten times lower than prior demonstrations of this qubit type. None of that erases the skeptic's point, though: right now, this is one paper, one eighteen-qubit device, one repetition code. HRL hasn't shown fault-tolerant error correction that actually protects arbitrary computation, and its own materials are explicit that this is a research prototype, not a shipped product. It's worth putting this next to our first story today, too. IonQ just bought an entire chip foundry to lock down its qubit manufacturing supply chain. HRL, working in a completely different qubit type — silicon rather than trapped ions — is attacking a different bottleneck: not who makes the qubits, but who controls them once they're freezing. Different companies, different technologies, same underlying instinct: the next wall in quantum computing isn't necessarily more qubits, it's everything wrapped around them — the wiring, the control electronics, the manufacturing line. Time for the Hype Check. I'm putting this one at a six. The peer-reviewed result is real, and the fivefold error suppression number is genuinely meaningful for anyone tracking the road to fault tolerance — that's not nothing, and it's the kind of engineering milestone that doesn't get faked. But roughly half of this story's punch comes from a manufacturability pitch that hasn't been tested against an actual production line yet, and 'first cryogenic controller to run error correction with zero help from room temperature' is a real engineering first — dressed, in the press-release language, like a bigger leap than an eighteen-qubit prototype actually represents. Fit a whole quantum computer in one refrigerator — ask me again once they've done it past eighteen qubits. IonQ's investor day on September 8th is probably the first real test of whether owning a chip foundry actually speeds up hardware or just speeds up a balance sheet — that's the one that'll settle whether today's wave of deals was strategy or spin. If a self-driving quantum chip and a chip-owning quantum company landing in the same week feels notable to you, follow Quickly Quantum wherever you listen so you don't miss what comes next. This has been Quickly Quantum, an AI-voiced podcast, created and built by a real human using today's cutting-edge technology. Nothing you heard on this show is financial advice. I'm Brian Lampert, and I'll catch you all tomorrow — take care!