← All episodes

September 22, 2026 · 10 min

IonQ's Korea Stack Bet Plus RSA-260 at $400k

About this episode

IonQ bundles a Superion 256 and SiV memory with SDT and a planned Gumi facility, plus a classical RSA-260 claim, and Phasecraft's Virginia move.

Source links

Quickly Quantum is an AI-voiced podcast, built and run by a real person. Nothing in this episode is financial advice.

More from Brian Lampert: Concrete Compute, the daily AI infrastructure briefing, and Space Stakes, the business of the new space race. Transcripts and every episode: quickly-quantum.kngoworld.chatgpt.site.

In full

Episode transcript

Plain-text version ↗

Is the near-term quantum business selling computers, or selling whole in-country stacks with local factories attached? Today on Quickly Quantum: IonQ supplying a Superion 256 plus quantum memory to SDT in Korea, with assembly planned in Gumi, and what that bundle says about who wins the data-center buildout. Before that, in the headlines: a classical RSA-260 factoring claim and a Virginia headquarters move. Welcome back to Quickly Quantum, your daily brief on the quantum frontier. It's Tuesday, September 22, 2026. Let's get into it.

Starting with that factoring record, according to Quantum Zeitgeist, and I want to flag this one is single-source, our show has not independently confirmed the company's cost math, so take the business framing as their claim. According to Quantum Zeitgeist, Cognition says its research team factored RSA-260, a 260-digit number, topping the previous RSA-250 solved in February 2020. According to Quantum Zeitgeist, Cognition says it used a new GPU lattice siever built with its Devin AI system. Now, for you if you're new here, lattice sieving is the grinding step inside the general number field sieve, or GNFS, the best-known classical factoring recipe, and according to Quantum Zeitgeist, Cognition says the trick was not new math but making that step run on graphics processing units instead of central processing units only. So how does it actually compare? According to Quantum Zeitgeist, the company puts RSA-260 at approximately $400,000 and 13.5 GPU-years of compute, at ten times lower cost than previously public methods by the company's account, and according to Quantum Zeitgeist the team estimates hyperscalers could now factor RSA-1024 numbers for around $30 million. That 30 million dollar estimate stopped me cold. And the implication? Here's my read, this is classical acceleration, not a quantum break, but if AI-built sievers keep making old attacks cheaper, how should you set your quantum-safe migration clock for data that has to stay secret for years?

Now, one planned item on today's slate did not carry reportable specifics we could air, so we're keeping it to that line and moving on to what we can source.

Crossing to Virginia, according to The Quantum Insider, with the same caveat, this is single-source and we have not independently confirmed it beyond that outlet. Phasecraft plans to establish its U.S. headquarters in Arlington County, Virginia, with a $113,000 investment expected to create 19 new jobs. Phasecraft, founded by researchers from the United Kingdom's quantum science community, develops algorithms for quantum computers, meaning the software recipes for materials simulation, chemical reaction modeling, and optimization, and the company says it picked Virginia for skilled workforce, research universities, and proximity to industry and government stakeholders. Why does a 19-job move matter to you? Because location is strategy here, the release ties the expansion to the Defense Advanced Research Projects Agency's Quantum Benchmarking Initiative program and the Advanced Research Projects Agency-Energy's Quantum Computing for Computational Chemistry program, so this reads, and here's my read, as planting a flag next to the federal buyers who decide what utility-scale actually means.

@Qtumanalysis posted on X: "John Roese, Dell's global CTO and chief AI officer, on the company's new quantum program, 'Generative AI was a big deal. Quantum is potentially a bigger deal because it will build on the shoulders of giants. Once we have viable, cost-effective, at-scale quantum systems, it changes everything.'" Roese attaches a condition, though: it changes everything once those systems are viable, cost-effective and at scale. That makes his bigger-deal line a forecast that depends on getting there. The same post says Dell has deployed an HPC cluster inside Diraq's Sydney laboratory, co-located with Diraq's silicon spin qubit processor, to test whether classical servers and a quantum chip can run as one machine. And that stack framing tees up our main story, are vendors now selling a computer, or a whole stack that lives in your data center?

Our main story today I am calling Korea's Bundle Play, compute plus networking plus local factory. According to The Quantum Insider, IonQ announced a multi-year partnership with Korea-based SDT, Inc. under which IonQ will supply SDT with a Superion 256 quantum computer as well as IonQ's silicon-vacancy, or SiV, quantum memory module. If you're new, Superion 256 is IonQ's trapped-ion system, meaning charged atoms held by fields and steered with lasers, while SiV memory is diamond-based storage aimed at quantum networking, holding quantum states so future quantum data centers can link machines together. Now, the structure matters more than the names. The Superion 256 system is planned for deployment with an SDT customer in South Korea, where the companies intend to develop a hybrid quantum-classical data center, meaning quantum hardware working side-by-side with your familiar servers, and SDT plans to establish a quantum manufacturing and system-integration facility in Gumi for quantum memory packaging, system assembly, integration and commissioning. IonQ Chairman and CEO Niccolo de Masi said in the release, "Demand for our new Superion 256 system is growing globally, and this agreement in Asia Pacific is yet another example of our clients combining IonQ's leading quantum computers with our quantum networking and quantum memory solutions." SDT CEO Jiwon Yune said, "Our objective is to make our new Gumi facility a key operational hub for IonQ's global supply chain and regional deployment by hosting manufacturing operations and assembling world-class quantum systems." Yune also said SDT will participate in assembly and operation of IonQ systems and pointed to a separate in-country cancer center project where they intend to develop hybrid quantum-classical infrastructure for medical and biomedical research. The release also lists IonQ's other Korea ties including KISTI, SK Telecom, Hyundai Motor Company, Seoul National University and Sungkyunkwan University.

So what changed today versus what regulars already understand? Vendors stopped competing only on qubit counts a while ago and started competing on who can actually land a machine, plug it into a data center, and keep the supply chain moving, and this deal pushes that logic one step further by bundling the computer with the networking piece and the local assembly line. My read, clearly my opinion, is IonQ is using SDT to buy footprint in Asia-Pacific, you get an undisclosed first customer deployment to learn on, a Gumi line for SiV packaging and system integration that could lower future deployment friction, and a biomedical hook that plays well with governments that fund infrastructure. But here's the counterweight the release itself hands you, the customer is unnamed, no delivery dates, no prices, no measured performance specs, and both the hybrid data center and the manufacturing facility are described as planned and intended, not completed installations. Placement announcements do not equal installed, benchmarked systems, and until SDT names who is taking the Superion 256, when acceptance happens, and what it actually runs in Korea, this is commercial positioning with hardware attached. So what would change my mind? A named customer, a delivery and acceptance date in Korea, or system specs measured on that installed machine, any one of those turns a memorandum into momentum. Time for the Hype Check. I give this a 6 out of 10 on substance, real hardware supply plus a concrete Gumi scope and a plausible data-center thesis, docked hard for anonymity on the buyer and for plans presented as future intentions rather than steel on the floor.

What this means for the next headline you see is simple, ask whether they sold you a computer or sold you a foothold, with a customer name and a date. If today's episode helped you sort signal from bundle-speak, follow Quickly Quantum wherever you listen. 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!

I also host Space Stakes: the business of the new space race, every day. What actually flew, what the contract is really worth, and who has customers. Find it wherever you get your podcasts.