MicroCloud Hologram announces a quantum multiplier, but testing is lacking

MicroCloud Hologram claims to have developed an approximate quantum multiplier designed for NISQ computers—the current machines that still operate with noisy qubits and lack full error correction. The proposal reduces circuit depth and the number of gates needed in exchange for accepting small deviations in the results.

From a technical standpoint, the idea makes sense. In an imperfect quantum processor, an exact but very long operation may yield worse results than a shorter, approximate one. Each additional gate increases the time qubits are exposed to noise, decoherence, and control errors. However, MicroCloud Hologram’s announcement still lacks sufficient data to consider this progress as demonstrated.

HOLO’s Quantum Multiplier: Key Points in 20 Seconds

• MicroCloud Hologram claims to have created an approximate quantum multiplier for NISQ devices.
• The design uses multiple adders that simplify or interrupt carry propagation.
• The company states it has achieved circuits with constant depth in some modules.
• Users could choose different levels of precision and resource consumption.
• Circuit reduction could improve fidelity when noise dominates execution.
• The announcement mentions testing on simulators and real quantum hardware.
• It does not specify which computers, how many qubits, or the input sizes used.
• It also does not provide tables with depth, gate counts, fidelity, or error distribution.
• No scientific paper, source code, DOI, or reproducible documentation is published.
• For now, it should be treated as a corporate announcement rather than an independently validated result.

The company is listed on Nasdaq under the ticker HOLO and comes from the holographic technology, LiDAR, and digital twin industries. In recent years, it has expanded from its core businesses into quantum computing, artificial intelligence, and blockchain. The new announcement presents the multiplier as a step toward practical applications but does not identify the researchers involved or provide an academic publication for review.

Why Multiplication Is Difficult on a Quantum Computer

Quantum processors do not multiply numbers the same way as classical CPUs. Operations are built from reversible circuits composed of gates acting on qubits.

A multiplier typically generates partial products and then sums them. This structure mirrors, with important differences, the process used in classical binary arithmetic. The more bits the numbers have, the more sums, carry operations, auxiliary qubits, and gates are required.

Circuit depth measures how many consecutive layers of operations it contains. Gates that can run simultaneously count within the same layer, while those dependent on previous results must wait.

This detail is especially critical in the NISQ stage. John Preskill coined this term to describe computers with an intermediate number of qubits that are too noisy for reliable long circuits. In these machines, noise limits how many operations can be completed before results degrade.

A traditional ripple-carry adder processes bits in a chain. Determining the output of a given bit may require waiting for a carry from the previous bit, causing circuit depth to grow with number size.

MicroCloud Hologram’s approach, according to their statement, involves weakening the accuracy of less significant bits and simplifying parts of that chain. Errors in less relevant positions have limited impact on the overall value, enabling the circuit to be completed with far fewer layers.

This idea is not entirely new in quantum research. A preprint published in 2024 proposed various approximate adders with reduced depth. These designs computed some outputs with a single CNOT gate or even no gates for certain approximations. Noise simulations with Qiskit showed shorter circuits could better preserve fidelity than exact, deeper adders.

The contribution now attributed to HOLO appears to involve using these kinds of approximated adders to construct a full, configurable multiplier. The announcement mentions four levels of precision, allowing each application to choose between more accurate results or smaller circuits.

This principle resembles techniques long used in classical image processing, audio, neural networks, and other systems. Many tasks tolerate small numeric errors if it means saving energy or speeding up computation.

An interesting paradox in quantum computing is that relaxing mathematical precision can improve the real quality of the result. An exact circuit accumulating too many physical errors might end up further from the correct answer than an approximate, shorter circuit that preserves qubit coherence.

The T Gate Challenge and a Nuanced Claim

MicroCloud Hologram emphasizes reducing the number of T gates as a primary advantage of their multiplier. While this is relevant, the announcement conflates two different stages in quantum computation.

In fault-tolerant designs, circuits often use the Clifford+T gate set. Clifford operations are relatively easy to protect with error correction codes, but T gates require additional resources like magic state preparation and distillation.

This process can consume many physical qubits and operations, making T-count and T-depth key metrics when designing future fault-tolerant quantum computers. Literature has long sought to reduce these costs in arithmetic modules like multipliers and dividers.

However, current NISQ devices do not typically implement T gates using large fault-tolerant magic state factories. Compilers translate high-level circuits into native hardware gates—rotations and two-qubit gates—depending on the architecture, whether superconducting, trapped ions, or neutral atoms.

Reducing T-count may still be useful for comparing designs and planning future fault-tolerant implementations, but it doesn’t directly translate into actual resource savings on NISQ hardware. To accurately assess the multiplier, it would need data after compilation for a specific machine: physical gate counts, connectivity, final depth, error rates per operation, and execution time.

Such detailed information is absent from the announcement. HOLO claims testing on various simulators and real devices, but does not specify processors, number of runs, or the fidelity measurement method.

Nor does it specify what reference designs it compares against when claiming superiority. This is complex, as different proposals optimize for different metrics—auxiliary qubits, total depth, T-depth, gate count, or state preparation costs.

For example, a 2026 publication proposed an exact multiplier with polylogarithmic depth and T-depth, at the cost of quadratic gate and qubit resources. This indicates that the research on exact and approximate methods continues, and no single metric can determine superiority conclusively.

Essential Data for Validating Progress

Without these elements, improvements might only be apparent on small inputs, under favorable noise conditions, or relative to a weak baseline implementation.

Hypotheses on Practical Applications Remain

HOLO mentions quantum machine learning and optimization problems as potential applications. Both can tolerate some numerical imprecision, but this doesn’t guarantee that an approximate multiplier will improve performance or results.

In classical neural networks, reducing precision from 32 to 16 or 8 bits can speed up training because processors include specialized units for these formats. In quantum computing, it must first be shown that the algorithm relies on reversible integer multiplications and that this component significantly impacts total cost.

Many NISQ algorithms use parametrized rotations, measurements, and classical optimization. They do not always incorporate multipliers like those described here. In some cases, data preparation costs may outweigh the arithmetic operations themselves.

Optimization problems also behave differently. Small deviations may be acceptable in smooth functions but can completely alter the solution landscape under strict constraints or with very close parameter values.

Benefits should be evaluated end-to-end. A multiplier’s fidelity alone isn’t enough; it must be integrated into a full algorithm, run on actual hardware, and shown to improve overall solution quality, execution time, or the number of samples needed.

Approximate computing remains an active area of research to mitigate the limitations of current quantum hardware. Existing academic work supports the idea that shorter, less precise circuits can perform better under certain noise levels.

However, the gap between this possibility and a truly “practical” multiplier remains large. MicroCloud Hologram has proposed an interesting architecture but has yet to present the results needed for validation.

Publishing detailed circuit designs, tested input sizes, hardware used, and reproducible comparisons would help determine whether this is an actual breakthrough or just an application of known approximate adder techniques.

Until then, the most accurate statement is that HOLO claims to have a solution and outlined its approach. The experimental validation is still pending.

Frequently Asked Questions

What is an approximate quantum multiplier?
It’s a circuit that multiplies numbers using qubits but simplifies some operations and tolerates small errors to reduce circuit depth and gate count.

Why might an approximate result be better on a quantum computer?
Because a shorter circuit spends less time exposed to noise. On some machines, the controlled, managed errors can be smaller than the errors accumulated in an overly long exact implementation.

Has MicroCloud Hologram publicly demonstrated its technology?
The announcement states tests on simulators and real hardware, but details necessary for reproduction and verification are not provided.

Can it be used in commercial applications now?
There is not enough public evidence to support that. It should first be evaluated within full algorithms, on specific hardware, with appropriate metrics.

via: prnewswire