19–25 September 2026
Five takeaways from a week of deals, logical qubits and public money.
Last week’s report followed the pieces of the quantum stack starting to connect. This week was about commitments: three system deals, two with 2027 installation dates, a national competition shortlist, a new research center and half-year results. The clearest hardware result came from Infleqtion, and it arrives with an important qualifier.
The commitments are arriving faster than the machines.
Deployment
IonQ announced three deployment agreements for its Superion 256, a system of 256 physical trapped-ion qubits: a partnership with SDT, a research partnership with NVIDIA and a sale to Florida International University. The two with installation dates both point to 2027, the delivery target IonQ gave when it launched the product line. What the week showed is commitment; what it did not show is delivery.
The shaded area marks 2027. “Late 2027” is drawn as September to December. No contract values were disclosed.
The investor question: can IonQ turn these Superion agreements into installed, working systems on the dates it has given?
Logical qubits
80 atoms 10 blocks 30 logical qubits
Infleqtion entangled 30 logical qubits on its Sqale neutral-atom computer, in a circuit of about 1,000 physical operations. Outputs the ideal circuit allows appeared about 1,000 times more often than random sampling would produce them. Last week the company designed a larger code in software; this week it ran a smaller one on hardware.
Infleqtion announcement · 24 September ↗ · Technical write-up ↗
Decoding · simulated workloads
Tested on simulated workloads: up to 408 logical qubits, more than a million T gates in one circuit, and 31.5 million rounds of error checking summed across memory blocks. Decoding added under 0.3% to run time at lower error rates and up to 12% at higher ones. Not yet run on data from a real machine; the preprint dates from August.
IonQ · 22 September ↗Resource estimation · planning tools
The two are working to extend Microsoft’s resource estimator to distributed machines linked by optical interconnects, and to advanced codes such as Photonic’s SHYPS. The new capabilities have not been released yet.
Photonic · 23 September ↗Applications & benchmarks
Three results with real numbers behind them. Each holds on the case tested, and none is a claim that quantum computers now beat classical ones in general.
Q-CTRL · IBM Nighthawk R2
A 60-qubit physics simulation run through Q-CTRL’s Fire Opal software, with results within 1% of the Heron run. A hardware-generation comparison on one workload; Q-CTRL also derives a more than 400x lower bound against the classical GPU benchmark from its earlier study.
Q-CTRL · 24 September ↗IonQ · satellite radar
Spotting changes in radar images of one airfield. The simulated quantum model scored 0.41; with training and inference both on IonQ’s Forte Enterprise it scored 0.32, still above both classical baselines. On interferometric data, all methods were comparable.
IonQ · 24 September ↗ · Paper · 4 Sep ↗Classiq · INGL · IonQ
On a representative model of Israel Natural Gas Lines’ network, a simulator run recovered the classical maximum-throughput point. A reduced version run on IonQ Forte-1 produced valid candidates close to it. No advantage is claimed.
Classiq · 23 September ↗All three are company-reported. The F1 bars share a 0–0.5 scale; the Q-CTRL bars share a 0–265 second scale.
Public money
Germany shortlisted six consortia, two for each of three qubit technologies, in a competition with up to €640 million in planned funding. Its goal is at least two fault-tolerant quantum computers by 2030.
Next gate: at least one partner in each consortium must show a working system meeting the entry criteria by 31 March 2027. Being shortlisted is not the same as receiving the money; QUDORA’s €122 million is a proposed project volume.
Company results
Up about 438%, but all of it came from AME, a business acquired on 6 April, so the jump is not like-for-like growth. Gross margin was 61%, against a $6.4 million net loss. QTREX makes connectivity and electronics for quantum computers; its INSU300 material, launched 23 September, is with two government and defense organizations for validation. Results ↗
Revenue rose 14% and QPU-related services rose 34% to €3.9 million. Booked and awarded business includes grants and tax credits as well as multi-year customer contracts. The solid part of the loss bar is €37.5 million of share-based payments and one-time charges; the hatched part is the remaining €21.7 million. Pasqal results · 24 September ↗
New, resurfaced, unverified
I found no demonstration of real-time error correction on hardware and no quantum-advantage comparison against a newly run classical baseline. That is an observation about this review’s coverage, not an exhaustive count of everything published worldwide.
Claims were checked against company releases, filings and the journal or preprint of record.
The weekly view
Last week the pieces started to connect. This week brought deals, a shortlist and a new research center: commitments that, where dated, pay off in 2027. The clearest hardware result, Infleqtion’s 30 logical qubits, relied on error detection and offline loss correction rather than real-time error correction.
Now the deals have to become installed machines.
