Welcome to issue #24 of This Week in Quantum — a roundup of the most important developments in quantum computing from August 17-23, 2026.
This week’s stories cluster around a theme: the unglamorous engineering and manufacturing work that has to happen before any of the field’s bigger promises can ship. A materials fix that makes a good qubit material foundry-compatible, an actual foundry opening with a stated qubit roadmap, a certification that gives post-quantum cryptography a real audit trail, a decoder that makes existing error-correction hardware perform better without new chips, and a physics result that shows what these platforms are good for beyond computing. Below is what happened, with the specific numbers behind each claim.
Research Highlights
Cornell makes tantalum qubits compatible with standard chip fabrication
Published in Nature Materials on August 18 (“Krypton-sputtered tantalum films for scalable high-performance quantum devices”), a Cornell team led by Assistant Professor Valla Fatemi solved a specific manufacturing problem: tantalum is prized for superconducting qubits because of its strong resistance to decoherence, but forming its useful body-centered-cubic (α-phase) crystal structure on silicon normally requires heating the substrate above 400°C during sputtering — hot enough that it exceeds the thermal budget of standard semiconductor foundry back-end-of-line (BEOL) tooling, and hot enough to cause tantalum-silicon intermixing that degrades qubit coherence. The Cornell team switched the sputtering gas from argon to krypton. Because krypton atoms are heavier, they transfer more kinetic momentum to the ejected tantalum atoms, which stabilizes the α-phase crystal lattice at substrate temperatures as low as 200°C — half the previous requirement. Transmon qubits fabricated from the resulting films, with compact 20-micron capacitor gaps, achieved internal quality factors up to 16.9 million, and the films also showed higher electronic conductivity than argon-sputtered tantalum. The practical upshot: a material foundries can now plausibly integrate into existing automated tooling without a dedicated high-temperature process line.
Caltech directly measures a 40-year-old physics prediction for the first time
Published in Nature on August 19, a Caltech-led collaboration (Manuel Endres’ experimental group and Jason Alicea’s theory group, with Université Paris-Saclay and the Technical University of Munich) used a neutral-atom quantum simulator to directly measure the finite-size energy excitation spectra predicted by conformal field theory (CFT) — specifically the Ising and tricritical Ising models — at quantum phase transitions. The team trapped chains of up to 35 strontium atoms in optical tweezers and developed a “many-body modulation spectroscopy” technique to read out the energy level ratios these CFTs have predicted since John Cardy’s framework in the 1980s, but which had never been directly observed experimentally. The measured ratios and their scaling with system size matched the universal predictions. This isn’t a computing benchmark — it’s a demonstration that neutral-atom platforms originally built for quantum computing are now precise enough to test open questions in fundamental physics, including, the team says, two-dimensional CFT regimes where even classical computers may not be able to calculate the answer.
An AI decoder beats Google’s benchmark on real hardware data
Quantum X Labs reported that its AI-driven quantum error correction decoder outperformed Google’s PyMatching benchmark on Google’s own surface-code hardware dataset, using NVIDIA’s CUDA-Q platform for low-latency, real-time decoding. The model was trained exclusively on synthetic data and still generalized to real hardware error patterns — relevant because decoding speed and accuracy are as much a bottleneck to practical fault tolerance as physical qubit count, and a software-only improvement doesn’t require new chips to deploy.
Industry News
QpiAI opens what it calls Asia’s largest quantum chip foundry
On August 17, Bengaluru-based QpiAI inaugurated the second phase of its 8-inch quantum processing unit foundry in Jakkur, part of a 70,000-square-foot R&D center with Class 100 and Class 1,000 cleanrooms. The facility currently fabricates flip-chip superconducting processors with up to 128 physical qubits and handles the full device manufacturing chain — lithography, etching, patterning, assembly, and packaging — in-house. QpiAI has already produced four processors at the site: the 8-qubit QVidya, the 25-qubit Indus (integrated into hybrid classical HPC data centers), the 64-qubit Kaveri (using proprietary low-loss flip-chip interconnects), and the 9-qubit Yukti, a fluxonium-based chip built specifically to evaluate fault-tolerant surface codes and logical qubit encodings. The company has invested roughly $20-25 million in the facility so far, with another $10-15 million planned; a third phase, targeted for 2027, aims to scale the foundry to fabricate single QPUs with up to 10,000 physical qubits. About 20% of required inputs are currently imported, according to founder and CEO Nagendra Nagaraja.
Security Watch
Crypto4A validates the first FIPS 140-3 Level 3 quantum-safe HSM
On August 20, Canadian cybersecurity company Crypto4A announced that QASM, the cryptographic module at the core of its QxHSM hardware security module, received NIST FIPS 140-3 Level 3 validation — the first time a hardware security module supporting the complete suite of NIST-standardized post-quantum cryptography algorithms (FIPS 203, 204, 205, and LMS) has reached this validation level. The distinction matters: FIPS 140-3 Level 3 requires demonstrated physical tamper resistance and strong identity-based authentication, not just correct implementation of the underlying algorithms, which is what separates a certified product from a PQC software announcement. DigiCert has already integrated the validated module into its DigiCert ONE platform to secure certificate issuance, digital signing, and automated PKI infrastructure.
The Bigger Picture
Every story in this issue is a supply-chain or verification problem, not a qubit-count record, and that’s a fair description of where the field’s bottleneck currently sits. A foundry-compatible tantalum process, an actual 10,000-qubit-capacity foundry with a stated timeline, a certified HSM an auditor can sign off on, and a decoder that improves an existing error-correction stack in software are the categories of work that determine whether headline milestones from other weeks ever become deployable products — none of them will generate the kind of press a new logical-qubit record gets, but collectively they’re a better read on how close the industry actually is to shipping at scale.
Upcoming Events
- September 1-4 — QUANT 2026 (QUANCOM), Trento, Italy
- September 9 — Quantum Roundabout, Nottingham, UK
- September 9-11 — Quantum Expo Korea 2026, Seoul
- September 13-18 — IEEE Quantum Week, Toronto
- September 22-24 — Quantum World Congress 2026, College Park, Maryland
- September 24-25 — Q2B Paris 2026, Paris
- September 28-30 — Quantum Innovation Summit, Dubai
- October 1 — Quantum in Business (QiB), Chattanooga, Tennessee
- October 12 — Quantum Computing & Quantum Technology (QUANTUM), Berlin
- October 25-27 — IQT Quantum+AI 3.0, New York City
- November 2-4 — International Conference on Quantum Annealing (INQA) 2026, Bled, Slovenia