The Lyceum: Quantum Intelligence — Jun 18, 2026

This was a week where the abstract got physical. A government lab put a commercial quantum computer through independent testing and published the result in Nature. France set a date for when legacy encryption stops getting certified. PsiQuantum poured concrete in Australia. The recurring theme: the distance between "promising lab result" and "hard regulatory deadline" is shrinking fast — and most organizations are moving slower than the calendar now demands.

• Helios (Quantinuum): a 98-qubit trapped-ion system now independently validated by Sandia in Nature, with two-qubit fidelity above 99.9% on the session.
• Gate-Model Simulator (D-Wave): the company's first error-aware programming tool, letting developers test algorithms against realistic hardware noise.
• Hybrid PQC Key Establishment (AWS): KMS, S3, and CloudFront now run classical elliptic-curve Diffie-Hellman combined with ML-KEM in production.
• Composite SSH Signatures Draft (OpenSSH): an IETF Internet-Draft published June 2 combining ML-DSA-44 with Ed25519 — still in progress, not yet ratified.

Vendor benchmark claims are easy to dismiss. An independent national laboratory publishing in Nature is harder to ignore.

Sandia National Laboratories and Quantinuum published a paper on June 17 reporting the performance of Quantinuum's 98-qubit Helios system. In one- and two-qubit operations, it demonstrated fidelity of 99.9975% and 99.921%, respectively. Fidelity, in plain terms, is how often a quantum operation produces the right answer — the closer to 100%, the fewer errors pile up as you chain operations together.

What separates this from November's commercial launch is who held the ruler. Sandia ran an independent assessment and contributed new benchmarking methodology for mid-circuit measurements — a technique essential to error correction. The headline number: Helios produced 48 fully error-corrected logical qubits at a 2:1 encoding rate. A competing demonstration would need roughly 4,800 physical qubits to make 48 logical ones. Helios does it with 98.

That 2:1 ratio is the number that will define the error-correction conversation for the rest of 2026. If it holds under outside scrutiny, the overhead assumptions baked into every fault-tolerance roadmap need revising downward. If competing trapped-ion or superconducting teams publish rebuttals, the claim was platform-specific. Either outcome is informative.

Most post-quantum cryptography news is about standards documents and migration roadmaps. France just made it about market access.

At the France Quantum conference on June 16, the national cybersecurity agency ANSSI announced that beginning in 2027, it will stop certifying security products that don't incorporate quantum-resistant encryption. ANSSI certification isn't optional — it's mandatory for French government agencies and critical-infrastructure operators across energy, telecom, finance, and transport. The policy targets RSA and elliptic-curve cryptography, the mathematical foundations of most digital security today, with a full transition to quantum-safe products targeted for 2030.

ANSSI's chief of staff Samih Souissi framed this as a matter of governance, industrial planning, and sovereignty — not merely a technical issue. France is treating cryptographic infrastructure the way it treats energy independence.

Vendors selling into European government markets now have roughly 18 months to certify or lose the contract pipeline. The observable signal: whether other regulators name dates. If Germany's BSI or the EU's NIS2 framework follow with similar hard cut-offs, the European market for non-quantum-safe products effectively closes by 2027.

D-Wave built its reputation on quantum annealing — a technique for finding low-energy solutions to optimization problems like routing trucks or scheduling factories, without the extreme precision of gate-based machines. Now it's coming for the gate-model market too.

This week D-Wave released what it calls the world's first gate-model simulator built for error-aware programming — a tool that lets developers test algorithms while accounting for the real noise hardware introduces. That matters because most simulators assume perfect hardware, so code that works in simulation often dies on a real machine.

It's part of a dual-platform strategy with a roadmap that now carries dates: a superconducting dual-rail qubit architecture designed to catch roughly 90% of qubit errors during computation, with 17 physical qubits in 2026, 49 in 2027, 181 in 2028, a 10-logical-qubit fault-tolerant system in 2030, and 100 logical qubits by 2032. The economics behind the bet: usage of D-Wave's Advantage2 annealing systems climbed 314% over the past year, per the company. D-Wave is funding its gate-model push off real annealing revenue.

The company that proved quantum optimization works in production is now betting it can build the fault-tolerant hardware everyone else is still promising. Watch for independent fidelity benchmarks on that first 17-qubit system, expected later this year — that's the test of whether the roadmap is engineering or aspiration.

Most quantum companies exist as cloud services and press releases. PsiQuantum just poured concrete.

The company began construction on its Moreton Bay, Australia site on June 17. PsiQuantum builds photonic quantum computers — systems that use particles of light as qubits instead of superconducting circuits or trapped atoms. The advantage: some photonic components run at room temperature, and the chips can be fabricated using the same processes that make conventional semiconductors. PsiQuantum's Omega silicon photonic chipset is made at GlobalFoundries in New York, meaning its supply chain scales with the broader semiconductor industry rather than bespoke quantum facilities.

The backdrop is a $1 billion Series E closed in September 2025, with a parallel Brisbane site anchored by $620 million AUD in government funding, where Linde Engineering is building the largest cryogenic plant ever assembled for a quantum computer.

The groundbreaking makes a billion-dollar bet tangible in a way no roadmap can. The open question — the thing to watch — is whether photonic error rates reach the fault-tolerance threshold before superconducting and trapped-ion rivals get there first.

The quantum hardware race isn't just about qubits — it's about the materials those qubits are made from. China just secured a critical link in the supply chain.

China has begun mass-producing ultra-pure silicon-28, an isotope essential for silicon spin qubits — a qubit type that uses the quantum spin of individual electrons in silicon, much like how transistors already work. The appeal is enormous: these qubits could eventually be built in existing chip fabs, making them cheaper and more scalable than today's approaches. The catch is purity. Natural silicon contains isotopes that create magnetic noise and disrupt coherence — how long a qubit holds its state. Silicon-28 is the "quiet" isotope, and purifying it to quantum-grade levels is brutally hard. Until now, only a handful of facilities worldwide could make it in research quantities.

Mass production reduces China's dependence on foreign suppliers and signals its hardware ambitions reach beyond superconducting and photonic systems into the modality closest to conventional manufacturing.

If silicon spin qubits become the dominant long-term platform, control of ultra-pure silicon-28 becomes a geopolitical lever. The signal to watch: whether Western governments treat this the way they've treated rare earths — as a supply-chain risk demanding domestic investment.

Government quantum programs are easy to announce and hard to grade. Korea's new Quantum Flagship comes with specific, measurable targets — which makes it interesting.

Multiple Korean outlets reported this week that the government confirmed an 8-year, 645.4 billion won (roughly $470 million) program with explicit goals: a 1,000-qubit quantum computer and a 100-kilometer quantum communication network, plus a new Quantum Strategy Committee to push from R&D into industrialization. Korea currently ranks 10th globally in quantum computing talent and 12th in communication and sensing, per a talent analysis covered by Yonhap News and Dongascience (as of 2026 survey) — respectable, but well behind the US, China, and the leading European programs.

The industrialization framing is the tell. Korea is trying to replicate its semiconductor and display playbook: use state money to build a domestic industry, not just fund papers. Korean financial analysts name Samsung Electronics, LG Electronics, and SK Telecom as potential beneficiaries.

The 1,000-qubit target gives the effort a concrete benchmark that will either validate or embarrass the investment. Watch whether the program draws international partners — Quandela, the French photonics firm, has already floated Korea collaboration.

"Fault-tolerant quantum computing" has been a slogan for years. This week, someone put a date on it.

AWS announced on June 15 that it is expanding its partnership with QuEra to bring QuEra's neutral-atom system, Libra, to Amazon Braket in 2028. Neutral-atom hardware stores qubits in individual atoms held in place by lasers — very different from the superconducting chips dominating US headlines. AWS says Libra targets "scientifically relevant applications" with hundreds of logical qubits and about a million quantum operations. That's a big claim, and for now it's a roadmap, not a shipped machine.

It matters because cloud access is how most enterprises will first touch serious quantum hardware — and because AWS is effectively saying the post-toy-demo era now has a calendar. The shift is from vague promise to scheduled infrastructure. Watch whether AWS or QuEra publish deeper technical specs before the June 30 AWS DC Summit session they teased.

This week: a national lab pointing a ruler at a quantum computer and finding it honest, France quietly informing every encryption vendor in Europe that the clock started without them, and China stockpiling the world's quietest sand. The quantum threat that keeps "breaking" RSA in the headlines hasn't broken anything yet — but the regulators acting as if it has are the ones who'll look smart in 2030, and the vendors waiting for certainty are the ones discovering certainty arrives precisely when it's too late to be useful.

Stay sharp out there.

Forward this to the colleague who still thinks they have until the 2030s to think about PQC.