Bringing a new drug to market takes over a decade and costs billions of dollars. If quantum computers could accurately simulate molecular behavior, that journey would get shorter. The obstacle has been the million-qubit barrier. New work from Fujitsu and Osaka University, announced in March 2026, suggests the bar could come down to tens of thousands of qubits.

The Promise and the Problem

Quantum computers hold transformative potential for drug discovery. Unlike classical computers, even supercomputers, they can theoretically simulate the behavior of atoms and molecules directly, predicting how drug candidates interact with enzymes in the human body with unprecedented accuracy. This capability could revolutionize pharmaceutical development by slashing the time and cost of identifying effective compounds.

Quantum computers are also extraordinarily error-prone. Environmental noise constantly disrupts calculations, and correcting these "quantum errors" requires massive numbers of qubits. A fully fault-tolerant quantum computer (FTQC) capable of reliable, practical computation was generally thought to require around one million qubits, a goal that's still 10 to 20 years away.

On March 25, 2026, Fujitsu and Osaka University's Center for Quantum Information and Quantum Biology announced two new technologies that could fundamentally change this equation: "STAR Architecture ver. 3" and a molecular model optimization technique. Together, they slash the computational resources needed for complex molecular calculations by orders of magnitude.

What Is STAR Architecture?

STAR stands for "Space-Time efficient Analog Rotation quantum computing architecture." Developed jointly by Fujitsu and Osaka University starting in 2023, it takes an unconventional approach to quantum computing efficiency.

In a conventional FTQC system, calculations rely on repeatedly applying basic operations called "logical T gates" to achieve arbitrary-angle rotations, a computationally expensive process requiring enormous numbers of qubits. STAR Architecture replaces these costly T gates with a custom "phase rotation gate" that achieves the same result using far fewer quantum resources. The trade-off: phase rotation gates aren't protected by error correction, meaning they accept some degree of error in exchange for dramatically reducing qubit requirements and gate operations by more than a factor of ten.

The architecture has evolved in stages. Version 1 (2023) established the core concept. Version 2 (2024) improved phase rotation accuracy and expanded computational scale by 1,000 times, enabling simulations of solid-state materials like superconductors. But chemical materials needed for drug discovery required even more resources, until now.

Version 3: Two Innovations Break the Barrier

The latest version introduces two critical advances.

The first is an improvement to the STAR Architecture itself. Researchers discovered that when the phase rotation gate fails consecutively, it rotates in the wrong direction, severely degrading computational accuracy. Ver. 3 introduces an automatic switching mechanism: when consecutive rotation failures exceed a set threshold, the system falls back to the conventional logical T gate. This hybrid approach improves computational accuracy more than tenfold.

The second innovation is the "molecular model optimization technique." When calculating molecular energy, the molecular model is split into multiple terms, and two computational methods, "time evolution" and "random sampling", are combined based on each term's importance. The new technique reshapes the molecular model itself to redistribute these importance weights, optimizing the balance between the two methods. This minimizes the number of gates in the quantum circuit and reduces computation time by three orders of magnitude, roughly 1,000 times faster.

Real Results: Cytochrome P450

The team validated their approach on three industrially significant molecules: Cytochrome P450, an iron-sulfur cluster, and a ruthenium catalyst. The standout example is Cytochrome P450, an enzyme critical to drug metabolism in the human body. Understanding how drugs interact with P450 is essential for predicting both efficacy and side effects, making it a high-priority target for pharmaceutical research.

Under conventional FTQC architectures, simulating Cytochrome P450 at industrially useful accuracy required 740,000 qubits. With STAR Architecture ver. 3, the qubit count drops to roughly 40,000, a reduction of about 95%. Combined with the molecular model optimization, computation time shrinks to 9 days.

A ruthenium catalyst important for carbon recycling saw its requirements drop from roughly 2 million qubits to about 50,000, with computation time falling from an estimated 5,000 days to around 5 days. Across all three molecules, Fujitsu and Osaka University report qubit requirements falling to between one-fifteenth and one-eightieth of what conventional FTQC architectures would need. The physical error rate demanded of qubits was also relaxed from 0.01% to 0.1%, bringing the technology within range of quantum computers currently under development.

The Global Race: How Does This Compare?

Quantum-powered drug discovery is a global contest, with major U.S. tech companies investing heavily.

IBM has partnered with Moderna to model mRNA secondary structures using up to 80 qubits, and collaborated with Algorithmiq on photon-activated cancer drug development, a project selected as a finalist for the Wellcome Leap Quantum for Bio Challenge (a $40 million initiative). IBM's roadmap targets its "Starling" fault-tolerant quantum computer by 2029, while pursuing incremental "quantum utility" through hybrid classical-quantum workflows in the meantime.

Google partnered with German pharmaceutical giant Boehringer Ingelheim to demonstrate quantum simulation of the same Cytochrome P450 molecule that Fujitsu and Osaka University targeted. Google's "Willow" chip (late 2024) demonstrated that error rates decrease as more qubits are added, a crucial milestone. The company reported that Willow performed a computation in 5 minutes that would take classical supercomputers 10 septillion years.

IonQ announced in October 2025 that it had achieved quantum advantage in drug discovery applications, surpassing classical methods in chemistry simulations.

What sets the Fujitsu/Osaka University approach apart is its strategic angle. Rather than racing to build the largest quantum hardware, they're innovating at the software architecture level to reduce the hardware requirements themselves. While IBM plans for a million-qubit FTQC, STAR Architecture aims to make practical computation possible with just tens of thousands of qubits in the "Early-FTQC" era, machines that could arrive years sooner.

Fujitsu's Hardware: Building the Machine Too

Fujitsu isn't only working on software innovation. In partnership with RIKEN (Japan's premier research institute), the company has been developing superconducting quantum computers.

In April 2025, the RIKEN RQC-Fujitsu Collaboration Center unveiled a 256-qubit superconducting quantum computer, among the largest externally accessible gate-based machines in the world at the time. It quadrupled the 64-qubit system they released in 2023 and is now available to businesses and research institutions via their Hybrid Quantum Computing Platform.

The next milestone: a 1,000+ qubit machine planned for fiscal year 2026 (by March 2027), to be installed in a dedicated "Quantum Building" under construction at Fujitsu Technology Park in Kawasaki City. The collaboration period with RIKEN has been extended through March 2029 for long-term research and development.

However, the team acknowledges that at least 60,000 physical qubits are needed for practical Early-FTQC computation, meaning significant scaling remains ahead. Companies like Fujifilm and Tokyo Electron are already conducting joint research using the existing platform in areas including chemistry and materials science.

What This Means for Japan's Pharma Industry

The chemical materials market that STAR Architecture ver. 3 now targets is massive. While the superconductor market (the target of ver. 2) is worth approximately $9.3 billion, the combined pharmaceutical, ammonia synthesis, and carbon recycling markets are over 200 times larger.

Japan's pharmaceutical industry, home to major players like Takeda, Astellas, and Daiichi Sankyo, stands to benefit significantly from this technology as it matures. The ability to accurately simulate drug-protein interactions at quantum speed could compress drug discovery timelines from years to months.

Fujitsu's Shintaro Sato, head of the Quantum Laboratory, has outlined a staged vision: first improving calculation accuracy for relatively small molecules, then applying STAR Architecture improvements to tackle larger molecular systems. The ultimate goal extends beyond drug discovery to competitive applications across different fields.

Japan's "Quality Over Quantity" Bet

In the quantum computing race, headlines tend to follow qubit counts. Fujitsu and Osaka University are betting elsewhere: rather than waiting for the million-qubit FTQC era, they are targeting the Early-FTQC window, potentially only a few years out, to deliver industrial value with smarter software on smaller machines.

McKinsey estimates that quantum-enabled pharmaceutical R&D could create $200–$500 billion in value by 2035. The question isn't whether quantum computing will transform drug discovery, but who will get there first with a practical solution. With STAR Architecture, Japan has placed a distinctive bet.

Quantum computing for drug discovery is heating up globally. How is your country approaching this technology? Are there pharmaceutical companies or research institutions exploring quantum applications where you live? Share your thoughts in the comments, we'd love to hear your perspective!

References