The Quantum × AI Convergence Just Got Real: Japan's 1,024-Qubit Roadmap and the Post-Quantum Cliff

In December 2024 Google’s Quantum AI team showed that adding physical qubits to a logical error-correction patch lowered the logical error rate. Theorists had been waiting on that “below-threshold” result since the 1990s (Google Quantum AI, Nature, 2024). Four months later, Fujitsu and RIKEN unveiled a 256-qubit superconducting machine and committed to a 1,024-qubit successor going live in 2026 (Fujitsu, April 2025; RIKEN, April 2025). Eight months before the Google result, the U.S. National Institute of Standards and Technology had published the first three Federal Information Processing Standards for post-quantum cryptography, FIPS 203, 204, and 205.

None of these on its own makes universal fault-tolerant quantum computing imminent. Taken together they change what boards are asking. The 2026 version of the question is where quantum sits on a 2026–2030 roadmap, and why AI is landing on the same hardware.

This briefing covers three things: what quantum × AI means in practice this year, what Japan has built, and the cryptographic migration deadline most enterprises have not yet put in a plan.

What “Quantum × AI” Means in 2026

“Quantum AI” gets used a lot and defined rarely. In the NISQ era (Noisy Intermediate-Scale Quantum), nobody serious is betting on pure quantum advantage for production AI workloads. The pilots running at named institutions fall into three narrower buckets.

Quantum for AI. Hybrid algorithms such as the Variational Quantum Eigensolver and the Quantum Approximate Optimization Algorithm split a workload: a classical computer does the parameter optimization, and a short, noisy quantum subroutine handles the piece (usually combinatorial optimization or feature embedding) where quantum hardware has an asymptotic edge. Portfolio optimization, drug-discovery screening and materials simulation are where this shows up.

AI for quantum. The machine-learning techniques used to train large language models are being pointed at the engineering problems that gate quantum scaling: qubit calibration, real-time error decoding, pulse shaping, qubit material design. RIKEN and NVIDIA said in 2025 they are building hybrid AI–quantum supercomputers for exactly this dual workload (NVIDIA, 2025).

Hybrid workflow infrastructure. The 2025–2026 change that enterprises will actually feel is on the cloud side. IBM Quantum, AWS Braket, Azure Quantum and Fujitsu’s hybrid quantum platform now let a classical AI pipeline call a quantum subroutine as a module. A machine-learning team does not rebuild its stack to try quantum; it adds a layer.

Pure-play quantum advantage on enterprise AI workloads has not arrived. Hybrid workflows in narrow domains are running today, and the teams running them are accumulating the integration know-how.

Japan’s Stack

Fabrication, control electronics, error-correction software, AI compute, people. A country needs all of them to turn quantum engineering into product. Japan has assembled more of the set than most analysts credit. Four pieces.

Public–private hardware. Fujitsu and RIKEN’s 256-qubit superconducting machine, unveiled at the RIKEN RQC-FUJITSU Collaboration Center in April 2025, quadrupled the qubit count of the joint team’s previous 64-qubit machine and was opened to global users through Fujitsu’s hybrid quantum platform in Q1 fiscal 2025. The 1,024-qubit successor is scheduled for installation at Fujitsu Technology Park in 2026 (The Stack, 2025). NEC runs parallel programs in superconducting qubits and quantum annealing. NTT has funded photonic quantum computing and post-quantum cryptography research for years.

State research coordination. Q-LEAP (Quantum Leap Flagship Program) and the Cabinet Office’s Moonshot R&D Program set research priorities across universities, national labs and corporates. Moonshot Goal 6 names a fault-tolerant universal quantum computer as a long-horizon national objective. Foreign analysts tend to skip this layer, but it is why a lab result in Japan gets picked up into a national roadmap.

AI and quantum compute on one campus. The NVIDIA × RIKEN partnership announced in 2025 puts GPU-accelerated AI supercomputers next to the quantum hardware. Few countries have both at this scale in one place, so hybrid algorithms can be developed against real hardware in a single workflow.

People. Japan has produced strong cohorts of experimental quantum physicists for decades, and Fujitsu Research, NEC Central Research Laboratories and NTT Basic Research Laboratories keep senior researchers for decade-long stretches. That is how incremental hardware gains become manufacturable systems.

Japan is spreading its bet across state-coordinated research, two or three viable hardware programs, co-located AI compute and a trained workforce, much as it did in semiconductors and precision optics.

The Cryptographic Cliff: 2030–2035

Once a large enough quantum machine exists, Shor’s algorithm breaks RSA and elliptic-curve key exchange. Those underpin TLS, code signing, VPNs and most enterprise authentication.

On August 13, 2024, NIST published the first three post-quantum cryptography standards: FIPS 203 (Module-Lattice-Based Key-Encapsulation, derived from CRYSTALS-KYBER), FIPS 204 (Module-Lattice-Based Digital Signature, derived from CRYSTALS-Dilithium), and FIPS 205 (Stateless Hash-Based Digital Signature, derived from SPHINCS+).

Under NIST IR 8547, quantum-vulnerable algorithms (RSA, classical Diffie-Hellman, ECDH, ECDSA) are scheduled for deprecation by 2030 and removal from federal standards by 2035, with high-risk systems moving earlier (CyberArk, 2024). The NSA’s Commercial National Security Algorithm Suite 2.0 runs faster: National Security Systems network equipment must use CNSA 2.0 exclusively by 2030, and custom applications, legacy equipment and operating systems follow by 2033 (Morningstar, March 2026).

A vendor that sells to U.S. federal customers, sits in a defense supply chain or handles U.S. government data has to meet those dates. Certificate authorities, TLS libraries, code-signing pipelines and hardware security modules all have to be swapped or upgraded, and so does every long-lived encrypted asset.

Adversaries can intercept RSA-2048 traffic now and keep it until a quantum machine large enough to decrypt it exists. Anything that has to stay confidential for 10–20 years (legal settlements, intellectual property, health records, source code) is exposed on that basis today. Enterprise spend on the migration is now estimated in the $15 billion range.

Japan’s response, coordinated through METI, the National Center of Incident Readiness and Strategy for Cybersecurity (NISC) and CRYPTREC, aligns with the NIST standards while keeping a domestic evaluation track for cryptographic primitives. A company operating in both jurisdictions faces a single migration against the 2030 deprecation date.

What Belongs on the 2026 Agenda

The cryptographic inventory comes first. Most enterprises cannot say how many of their systems depend on RSA or ECC, and a migration cannot be scheduled until they can. The inventory has to be mapped to confidentiality lifetimes and to the jurisdictions each system falls under. Expect it to take longer than the migration planning itself.

Then a pilot. Integrating a quantum subroutine into an existing pipeline takes years of practice even though the hardware is not production-ready, so a hybrid pilot in a domain the company already knows is the cheap way to start. Portfolio optimization, materials simulation, supply-chain routing and drug-target screening have the cleanest pilot economics.

Hiring is the third item, and the one with the longest lead time. Researchers who can architect a hybrid quantum-classical system at production scale are scarce, and hiring markets lag policy and procurement deadlines by years.

The 2027 Conversation

Tech for Impact Summit 2027 returns to Tokyo in May 2027 with quantum × AI as a standing theme. The sessions are for the executives, regulators and researchers who will deploy hybrid systems and run the post-quantum migration. Fujitsu, RIKEN, NEC, NTT, Q-LEAP and the NVIDIA collaboration are all within a train ride of the venue.

T4IS2027 is an invitation-only executive summit. If your firm is building, financing, regulating, or deploying quantum-enabled or post-quantum-ready systems, we want you in the room.