IonQ Hit 1 kHz. That Is Not a Bookable QPU.
The number in IonQ’s October 9 release is a rate, not a machine you can reserve. Entanglement events above 1,000 per second, which they also write as 1 kHz, between a trapped-ion qubit and a solid-state memory, through a photonic interconnect. The company posted that from College Park. The Quantum Insider ran the same release the same day. citybiz did too. Three desks, one rate. None of the paragraphs we read included a fidelity percentage.
We wrote a different IonQ number three days ago in the 74x Clifford note. That one was a protocol speedup inside a blueprint. This one is a hardware link in a paper they say was tested on real devices. Both can be interesting. Neither is a cloud SKU with a serial number. We also wrote Duke’s 13-ion string-breaking result in that ion piece, and the factory pitch in Superion 256. The October 9 interconnect sits closer to the lab note than to the factory brochure. The CEO’s data-center analogy is the sentence that tries to drag it toward the brochure. Read the rate first.
IonQ’s 1 kHz is an entanglement rate
IonQ’s wording: entanglement rates above 1,000 per second between a trapped-ion qubit and a solid-state memory, via a photonic interconnect. The subhead names the memory more tightly. It is a silicon-vacancy quantum memory. The Quantum Insider’s reprint says the same thing in the technical-paper sentence: an end-to-end link between a trapped-ion system and a silicon-vacancy (SiV) qubit, on IonQ’s SiV quantum-memory platform.
Entanglement, in the release’s own gloss, is the quantum connection that lets separate systems work together, and the ingredient they say you need if you want to network quantum computers. A trapped ion, they write, sets the standard for coherence. A solid-state memory couples to light more efficiently. Putting them on one photonic link is the “best of both worlds” line. That is a design claim. The measured claim next to it is the rate.
1,000 events per second is 1 kHz. They use both. If a slide says “1 kHz” and a headline says “1,000 connections,” they are the same announcement, not two milestones. citybiz’s headline prefers “1,000 Quantum Connections Per Second” and then explains the kilohertz in the body. Do not count it twice in a roundup.
The release says the rate is fast enough to support distributed quantum computing and other use cases. “Fast enough” is their adjective. The paragraphs do not show the threshold they measured it against, other than the previous trapped-ion record discussed below. A rate can clear a record and still sit under the rate a specific architecture needs for a useful logical gate across a network. That second comparison is not in the text we have. I am not going to supply it.
Four times a Duke record, and the citation is 2024
IonQ says the demonstration is more than four times faster than the previous reported record for trapped-ion systems. The Quantum Insider’s copy is slightly more precise: more than four times the previous record for trapped ions, held by IonQ co-founder Chris Monroe’s research group at Duke University. That group is also a collaborator on the paper. citybiz’s version says “more than four times faster than the previous reported record for trapped-ion systems, according to IonQ,” and does not name Duke in the chunk we kept. The company release and the Quantum Insider reprint do name it. Use those, not the shorter citybiz line, if you need the holder of the old record.
The release’s reference line is O’Reilly et al., Physical Review Letters 133, 090802 (2024). That is a 2024 paper. It is the citation sitting under the previous-record claim, not an identifier for this week’s preprint. Mixing them is how a roundup ends up saying IonQ published the 1 kHz result in PRL in 2024. They did not. They announced it on October 9, 2026, and pointed at an arXiv preprint for additional technical detail.
We did not pull that preprint. The news posts do not paste an arXiv identifier in the sections we read. So this note does not invent a link, a fidelity, a distance, or a photon wavelength. If the preprint has a fidelity and you need it for a design review, go get the PDF. A press rate without a fidelity is a press rate. It can still be a real hardware result. It is not yet a number you should drop into an error budget.
Monroe’s quote, carried in the Quantum Insider reprint, puts the work in a longer company history. Interconnect efforts go back to the founding, more than a decade. Moving qubits through photons will be necessary in any large-scale quantum computer, he says, and this demonstration sets a foundation. He is quoted as Chief Scientist at IonQ and with a Gilhuly Family Presidential Distinguished Professor title. The quote is a roadmap sentence from a founder whose old group holds the record being beaten. That is allowed. It is not an independent lab’s confirmation. The collaborator relationship is disclosed in the same paragraph as the 4x claim. Keep those two facts in the same breath.
What the demo is not
It is not the 74x Clifford scheduler. That note was about operations inside an error-correcting blueprint, with Q70 and Q102 as codes rather than bookable systems. A reader who sees two IonQ headlines in one week can staple them into “IonQ is 74 times faster and also at 1 kHz.” Those units do not multiply. One is a software speedup claim on a class of gates. One is an entanglement-event rate on a photonic link between an ion and a memory. If your slide has both, it needs two footnotes.
It is not Superion. The company boilerplate on the October 9 release still calls Superion the newest generation of quantum computers. The interconnect announcement does not say the 1 kHz link is shipping inside a Superion rack. We treated Superion 256 as a factory pitch when we wrote it. This week does not convert that pitch into a delivered network. A boilerplate mention is not a product integration.
It is not the 13-ion string-breaking experiment. That was a physics result on a small ion chain. Monroe’s Duke group is the thread between that world and this press release, because the old interconnect record is attributed to his group and this paper lists the group as a collaborator. Shared people are not shared experiments. If you need the string-breaking setup, use that earlier piece. If you need a kilohertz, use this one.
Niccolo de Masi, chairman and CEO, supplies the sentence most likely to be clipped wrong. He calls 1,000 entanglement events per second a milestone for memory-enhanced quantum interconnects, then says classical data centers scaled by connecting specialized processors, memory, and networks, and quantum systems will scale “in much the same way.” The milestone language is his. The rate is the part you can check against the release’s own figure. The analogy is a picture of a data center, which this demo is not. A photonic link between one ion and one memory, at a stated event rate, is a component measurement. A networked quantum data center is the roadmap he says the measurement advances. Component, then roadmap. In that order.
SkyWater, DARPA, and the qubit-type claim
Mihir Bhaskar is quoted as IonQ senior vice president and general manager of quantum technologies at SkyWater. citybiz uses that title. The release’s version of the quote: a photonic quantum interconnect need not be a bottleneck for distributed quantum computing. He says the technology can interface with almost any qubit type, so he sees a wide range of uses, from modular computing to networked quantum sensing. citybiz’s shorter cut keeps the bottleneck line and the sensing range. The “almost any qubit type” clause is in the longer reprint. Use the longer one if you are going to repeat the compatibility claim, and keep it attributed to him.
DARPA’s HARQ program is the government line in the release. IonQ says this memory-and-interconnect approach is driving its work on HARQ, which seeks high-speed quantum interconnects compatible with multiple computing qubit types. The same paragraph says the latest demonstration used trapped-ion systems, while the underlying approach is pitched as broader. That distinction is the one to keep. A trapped-ion-to-SiV demo does not, by itself, prove the link has already been shown on every qubit type Bhaskar’s sentence gestures at. The program’s goal and the week’s hardware are related. They are not the same result.
I do not have a HARQ contract value, a period of performance, or a deliverable list in these posts. A mention of a DARPA program in a press release is a mention. It is not a funded-milestone table. If you are writing an internal brief, the honest line is: IonQ tied the 1 kHz demo to HARQ and did not, in the text we read, attach a dollar figure.
Two sales, and a forward-looking stamp
The commercial sentences are easier to date than the physics, and easier to overread.
The first commercial sale of the memory and interconnect platform went to the University of Maryland, announced in April 2026, as part of an expanded quantum networking and research collaboration. The second system sale went to SDT in South Korea, announced in September 2026. Both the company release and The Quantum Insider carry those two. citybiz carries the Maryland sale and, in the portion we extracted, does not add a third customer.
April and September are announcements of sales, not acceptance-test reports printed in this week’s posts. “Sold” and “installed and producing entanglement at 1 kHz on the customer’s floor” are different sentences. The October 9 technical claim is about a demonstration and a paper. The sales claims are about a platform the company says it has started selling. A buyer can have purchased a system whose spec sheet is not this week’s record. I cannot tell from these posts whether the Maryland or SDT systems are the apparatus that produced the 1 kHz figure. The release does not say they are. Do not add that.
The release ends with a forward-looking-statements paragraph. Research findings and future applications, including defense uses, sit under that stamp. A rate measured on a link can be a historical fact and still be wrapped in sentences about data centers that have not been built. Split them when you quote. The 1 kHz and the 4x comparison are the historical sentences, attributed to the company and, for the old record’s holder, to the collaborator relationship they disclose. The networked quantum data center is the forward sentence.
What to write down if you track this
If you keep a one-page note on vendor quantum claims, this week’s row is short.
Date: October 9, 2026. Claim: more than 1,000 entanglement events per second between a trapped ion and an SiV memory, photonic interconnect, described as more than four times the prior trapped-ion record held by Monroe’s Duke group. Paper: company says a technical paper on real hardware exists, and points to an arXiv preprint without an identifier in the news text we read. Prior-record citation printed in the release: O’Reilly et al., Phys. Rev. Lett. 133, 090802 (2024). Missing from the news text: fidelity, distance, which customer system if any was the demo apparatus. Commercial context, separate rows: Maryland sale announced April 2026, SDT sale announced September 2026. Program name: DARPA HARQ, no dollar figure in these posts.
That page is enough to stop the 1 kHz figure from being filed under “IonQ launched a networked quantum computer.” It is also enough to stop it from being filed under the 74x Clifford note. Two October claims, two units, one company that would prefer you remember the data-center sentence. Remember the rate, the missing fidelity, and the 2024 citation that is not this paper.
