What Is a Quantum Computer? Fault-Tolerant Quantum Computation and Where Each Company Stands
Looking only at "how many qubits" a machine has doesn't tell you when a practical quantum computer will actually arrive. Here's a rundown of the difference between physical and logical qubits, the characteristics of each implementation approach, and where each company currently stands.
Physical Qubits and Logical Qubits
The qubits that exist on today's quantum computer hardware are called physical qubits. Physical qubits constantly accumulate errors from noise, interaction with the environment (decoherence), and imperfect gate operations, so they cannot correctly carry out a long computation as-is.
To address this, a redundant encoding of a single "ideal" qubit built from multiple physical qubits is called a logical qubit. In a representative encoding scheme called the surface code, a technique called syndrome measurement detects and corrects the location of errors "without directly measuring — and thereby destroying — the quantum information itself." The number of physical qubits required varies widely, from several hundred to over 1,000, depending on the target error rate and code distance.
Implementation Approaches for Quantum Computers
| Superconducting | Uses Josephson junctions, with gate operations performed via microwave pulses. Gate speed is fast and the ecosystem is mature, but cooling to near absolute zero (around 10 millikelvin) is required |
|---|---|
| Trapped-ion | Uses the internal states of ions trapped by an electromagnetic field as qubits. Excellent gate fidelity and coherence time, but gate speed is relatively slow, and scaling up requires ion shuttling or photonic interconnects |
| Neutral atom | Uses optical tweezers to arrange neutral atoms, with gate operations performed via Rydberg states. The reconfigurable array makes it comparatively easy to scale to hundreds or thousands of atoms |
| Photonic | Uses photons themselves as qubits. Capable of room-temperature operation and well suited to communication, but realizing deterministic multi-qubit gates is technically difficult |
| Silicon | Uses electron spins confined in semiconductor quantum dots as qubits. Can repurpose existing semiconductor manufacturing infrastructure (CMOS processes) and has an advantage in integration density, but has historically lagged other approaches in gate fidelity — though this has been improving rapidly in recent years |
NEC's History of Foundational Research, and Its 2026 Strategic Shift
The superconducting approach to qubits has an important history of foundational research that originated in Japan. In 1999, a group led by Yasunobu Nakamura at NEC's Tsukuba Laboratories, using a superconducting Cooper-pair box, demonstrated the world's first coherent electrical control of a qubit in a solid-state device (the paper, "Coherent control of macroscopic quantum states in a single-Cooper-pair box," graced the cover of Nature in April 1999). This achievement is widely regarded as one of the technical origins of today's superconducting quantum computers.
NEC continued foundational quantum computing research for more than a quarter century afterward, and in 2023 unveiled an 8-qubit annealing-type prototype. However, on September 5, 2026, NEC announced that it would withdraw from hardware development of quantum computers (QPUs) that use qubits. The decision was a management judgment that recouping the investment would take too long; NEC will end direct QPU hardware development as of the fiscal year ending March 2026, and will instead focus on quantum annealing and quantum-inspired computation and related services running on classical computers. According to reports, many of NEC's quantum researchers have moved to Fujitsu, which is actively pushing forward with hardware development.
Where IBM, Intel, imec, Hitachi, and Fujitsu Stand
| IBM | Aims to demonstrate "quantum advantage" by the end of 2026. In 2026, it plans to introduce a 1,386-qubit multi-chip processor called "Kookaburra," linking 3 chips via chip-to-chip couplers to build a system of 4,158 qubits in total. It has laid out a roadmap that improves coherence, gate fidelity, error correction codes, control systems, and decoders together as a single effort |
|---|---|
| Intel | Advancing the silicon spin qubit approach. In January 2026, it supplied a 12-qubit chip called "Tunnel Falls," fabricated on 300mm wafers, to Argonne National Laboratory in the US (the results were published in Nature Communications). As of 2024, it reported achieving 99.9% gate fidelity and roughly 95% yield across the wafer. Going forward, it aims to achieve high-fidelity two-qubit gates and larger-scale two-dimensional arrays |
| imec | In May 2026, unveiled the world's first quantum dot qubit device fabricated using High-NA EUV lithography. It realized silicon quantum dot spin qubits with a gate-to-gate gap of about 6nm, and is working to shift its focus from lab-scale demonstration devices toward reproducible qubit manufacturing compatible with 300mm fabs |
| Hitachi | In July 2026, began a silicon quantum computer R&D project together with Intel K.K. and the National Institute of Advanced Industrial Science and Technology (AIST), supported by Japan's NEDO. The project targets a 100-qubit-scale silicon chip design, manufacturing platform, and 3D integration technology, with a 100-qubit error-correction prototype targeted for fiscal 2028 and a 1,000-qubit-scale prototype targeted for fiscal 2030. It builds on pioneering research into quantum dot charge qubits from Hitachi's Cambridge Laboratory in the UK |
| Fujitsu | Partnered with RIKEN in 2025 and is actively pushing forward with hardware development. With NEC's withdrawal from hardware, Fujitsu is also said to be serving as a destination for NEC's quantum researchers |
Looking across these developments, even within Japan alone it's clear that a major reorganization is underway — both in terms of technical approach (superconducting vs. silicon) and in the division of roles between companies — as NEC, despite being a pioneer of foundational research, withdraws from hardware, while Fujitsu and Hitachi (in partnership with Intel and AIST) each carry the work forward using different approaches.
Summary
- Physical qubits accumulate errors, so logical qubits — made redundant across multiple physical qubits — are necessary
- By the threshold theorem, if the physical error rate stays below a certain level, increasing the code distance can suppress the logical error rate exponentially
- Multiple implementation approaches — superconducting, trapped-ion, neutral atom, photonic, and silicon — are being developed in parallel
- NEC has a history that includes the world's first solid-state qubit demonstration in 1999, but withdrew from hardware development in September 2026
- IBM, Intel, imec, Hitachi, and Fujitsu each continue development using different approaches and organizational structures
Learn the Fundamentals of Quantum Algorithms
Grover's algorithm is explained from a concrete 2-qubit example through to the general case.
Read the Next Article