Latest / Quickly Quantum / September 4, 2026 — IBM's Speed Record: Real Fix or Self-Graded Exam?
Transcript
- Can IBM make its quantum computers dramatically more useful without adding a single extra qubit? IBM says its new Nighthawk r2 chip runs over a hundred thousand circuits a second — twenty-five times faster than its current fleet — and whether that's genuine progress or a number picked to win a race IBM set for itself is exactly what we're digging into today. That's the kind of claim that sounds simple until you ask who's actually checking it. Before that, in the headlines: Quantinuum signs a new energy-sector deal with Saudi giant Aramco, Germany flips the switch on a homegrown trapped-ion machine in Jülich, UK silicon-qubit company Quantum Motion pulls in fresh cash from a growing list of investors, and the G7's top cybersecurity agencies tell industry it's out of excuses on post-quantum encryption. Welcome back to Quickly Quantum, your daily brief on the quantum frontier. It's Friday, September 4, 2026. Let's get into it. Now, first up: Quantinuum just landed one of the more geopolitically interesting deals of the year. At the LEAP tech conference in Riyadh, the trapped-ion company and Saudi energy giant Aramco signed a non-binding memorandum of understanding — an MoU, basically a formal 'let's find out together' agreement, not a contract. The plan is to benchmark Quantinuum's QCCD architecture — quantum charge-coupled device, the ion-trap design behind its Helios system — plus its Nexus software platform, against real energy-sector problems: reservoir optimization, molecular simulation, the chemistry Aramco actually needs solved. It complements what Aramco's already running on-site in Dhahran, a neutral-atom system built with Pasqal, and it adds to a string of Gulf-state quantum courtships this year. Riyadh's LEAP conference has become something of a magnet for these deals this year, and Aramco isn't shy about shopping across quantum modalities before committing serious capital to any one of them. Here's the caveat worth sitting with: this is explicitly preparatory. No committed spend, no deployment date — just what Quantinuum itself calls 'technical onboarding' and 'knowledge exchange.' How many of these MoUs actually turn into paid systems within a year? Nobody's tracking that number yet, and it's the one that'll tell you whether this Gulf quantum push is real money or just diplomacy. Germany just turned on its own ion trap. Forschungszentrum Jülich inaugurated JION — its trapped-ion quantum computer — built with Siegen-based eleQtron under a partnership called EPIQ, which pulled in about twenty-one million euros in state funding over four and a half years. JION holds its qubits in ytterbium atoms, controlled with microwaves and magnetic fields rather than the near-absolute-zero temperatures superconducting chips need, and it's being folded straight into JUNIQ, Jülich's setup that lets quantum processors and supercomputers work the same problem together. JUNIQ's whole pitch is letting researchers hand off just the pieces of a calculation that actually benefit from a quantum processor, while the supercomputer carries the rest — a hybrid approach that's becoming the default way institutions think about deploying today's noisy machines. Now, the bigger headline might be the money that came alongside it: North Rhine-Westphalia also approved roughly twenty-five million euros for eleQtron's SQALING project, to shrink that trapped-ion tech onto a chip, and another roughly twenty-five million euros for a separate effort, Q-STAR.NRW, to buy a semiconductor quantum computer with up to two hundred qubits for that same JUNIQ infrastructure. As HPCwire, the HPC trade outlet, put it on X, quote, 'Jülich and eleQtron Inaugurate JION Trapped-Ion Quantum Computer,' end quote. That's roughly fifty million more euros of regional betting stacked on top of the machine itself, with no timeline yet for when any of it proves a practical edge. Silicon-spin qubits keep pulling in money nobody expected a few years back. UK company Quantum Motion just announced the second close of its Series C funding round, adding new investors including Imec.ventures, Lansdowne Partners, Sony Innovation Fund and S3 Ventures, building on the hundred-and-sixty-million-dollar first close it announced earlier this year. The company didn't disclose a dollar figure for this second tranche. The cash goes toward custom silicon co-development, tapping into existing semiconductor manufacturing pipelines, and international expansion — including a new U.S. lab in Maryland — as Quantum Motion pushes toward fault-tolerant hardware built on standard silicon CMOS, the same manufacturing process behind your laptop's chip. The pitch that's clearly working on investors is cost: leaning on decades of existing chip-fabrication infrastructure instead of building bespoke hardware from scratch is what's letting Quantum Motion make that pitch to a widening pool of strategic and financial backers at once. The company's also part of Stage B of DARPA's Quantum Benchmarking Initiative. We flagged last month that silicon spin was suddenly attracting serious industrial money from more than one direction at once — this is that same pattern continuing, another investor list getting longer without a headline number attached to it. Now, a warning shot from the cybersecurity world. A G7 working group — signed off by agencies including America's CISA, the UK's NCSC, France's ANSSI, and Germany's BSI — says organizations, quote, 'can no longer afford to postpone' migrating critical systems to post-quantum encryption, end quote, according to CyberScoop's reporting. The concern is 'harvest now, decrypt later': attackers grabbing encrypted data today, betting a future quantum computer can crack it open later. The report also flags that some of the encryption algorithms NIST already selected as quantum-resistant have themselves been broken by classical computers or AI, which is why NIST keeps pushing what it calls crypto-agility — building systems that can swap encryption schemes without a full rebuild. The message from all four agencies together is unusually blunt for a joint report, and it leans hard on the idea that waiting for certainty about the threat timeline is itself the risk, not a reason to relax. Let's be straight about what this actually is: a coordinated warning and roadmap, not new law, and this is a single-source item from CyberScoop we haven't independently confirmed elsewhere yet. No cryptographically-relevant quantum computer exists today to prove that threat timeline right or wrong — but the agencies clearly aren't willing to wait and find out. That tension, real long-term risk with no working machine yet to size it, is exactly the gap our main story about IBM's new chip lives in too. Our main story today: is IBM's new speed record real progress toward useful quantum computing, or just a faster lap on a track IBM built for itself? IBM's newest chip is called Nighthawk r2, and the headline number is genuinely eye-catching — it runs over one hundred thousand circuits every second, twenty-five times the throughput of IBM's current Heron fleet, according to IBM's own announcement. Now, here's the part that makes this more than another spec bump: IBM didn't get there by adding qubits. Nighthawk r2 sits at the same one hundred and twenty qubits as its r1 predecessor, the same square-lattice layout too. The speed gain comes from fixing something more boring but more fundamental — how fast a qubit resets between runs. Every time you run a quantum circuit, the qubits need to be reset back to a known state before the next run starts, and that reset has historically been slow, because it relied on a conditional process — measure the qubit, check the result, then decide how to reset it, one careful step at a time. IBM says Nighthawk r2 replaces that with what it calls a dissipative reset architecture — a way of letting the qubit settle back to its ground state passively, without waiting on that measure-and-check loop. Cutting that measurement step out doesn't just save time on paper — it removes one whole point in the circuit where extra errors could sneak in, which is part of why IBM ties the throughput gain and the lower initialization error to the same underlying fix. IBM's own figures put the initialization error twenty-five times lower as a result, and less error at the start of a circuit means fewer wasted runs before you even reach the interesting part. IBM also says it's already run its 'doped Clifford sampling' experiments — a quantum-advantage-style benchmark developed with the University of Chicago — on the new hardware, hitting more than seventy-five hundred gates before errors overwhelmed the result. That gate-count figure matters because it's a target IBM set for itself on its own 2026 roadmap. Hitting your own bar is still hitting a bar. But it raises the obvious question: is this an externally agreed milestone, or an internally chosen one IBM gets to grade itself on? Here's where I want to slow down, because every number in this story — the hundred thousand circuits a second, the twenty-five-x throughput, the twenty-five-x lower initialization error, the seventy-five-hundred-gate milestone — comes from IBM's own blog and its own release. Independent third-party benchmarking of Nighthawk r2 hasn't surfaced yet. That doesn't make the numbers false. It means you're hearing IBM grade its own homework, and the company has an obvious incentive to make this year's chip look like a leap. So let's take the two questions this story actually raises. First: does raising circuit throughput without adding qubits change the timeline to fault-tolerant quantum computing — the error-corrected version of these machines everyone's actually racing toward — or does it just make today's noisy, error-prone machines cheaper to run? My honest read is the second, mostly. Reset time is real dead weight — every microsecond a qubit spends resetting is a microsecond you're not running science on it — so cutting it is a legitimate engineering win, and it'll matter to anyone paying for cloud access, since more circuits per second on the same rented hardware is a real cost story. But it doesn't touch the harder problem, which is error rates per gate, not how fast you can retry. A machine that fails faster is still failing. It's also worth remembering that circuit throughput and error-corrected performance are measuring genuinely different things, and conflating them is exactly the kind of shortcut a skeptic would flag in a roadmap built to hit self-set targets. Second: how does IBM's 2026 roadmap milestone stack up against what Google, Quantinuum, or IonQ are claiming this year? Honestly, today's reporting doesn't give me a clean comparison, and I'd rather say that plainly than force one I can't back up. What I can say is IBM chose this particular benchmark, and Clifford sampling tasks are exactly the kind of thing skeptics point to when they say a 'quantum advantage' claim really means 'we found a task classical computers are slow at simulating,' not proof the quantum computer did something fundamentally beyond reach. Time for the Hype Check. I'm putting this one at a six out of ten. The reset-architecture fix is genuine, sensible engineering — not vaporware, and if it holds up under outside testing it'll save real money for anyone running circuits on IBM's cloud. But every figure is IBM's own, the gate-count milestone is a bar IBM set for itself, and the 'quantum advantage' language is doing more work than the underlying Clifford-sampling task can support. So the next time you see a quantum company tout a big multiplier — twenty-five x this, ten x that — the question worth asking isn't whether the number's true. It's who's holding the ruler. If today's episode helped you cut through a spec sheet or two — the reset architecture, the Gulf-state MoUs, the G7's encryption warning — the easiest thing you can do is follow Quickly Quantum wherever you listen, so tomorrow's episode just shows up in your feed automatically, and if any of those specifics helped you cut through the next quantum headline you see, that's the whole point of doing this daily. 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!