Eleven years ago, I began my PhD in theoretical physics and honestly had never considered or written about quantum computers. Meanwhile, New Scientist was busy crafting the first “Quantum Computer Buyer’s Guide,” always ahead of its time. A glance through reveals how things have changed—John Martinis from UC Santa Barbara was recognized for developing an array of merely nine qubits and earned a Nobel Prize in Physics just last week. Curiously, there was no mention of quantum computers built using neutral atoms, which have rapidly transformed the field in recent years. This sparked my curiosity: how would a quantum computer buyer’s guide look today?

At present, around 80 companies globally are producing quantum computing hardware. My reporting on quantum computing has allowed me to witness firsthand how the industry evolves, complete with numerous sales pitches. If choosing between an iPhone and an Android is challenging, consider navigating the press lists of various quantum computing startups.

While there’s significant marketing hype, the challenge in comparing these devices stems from the lack of a clear standard for building quantum computers. For instance, potential qubit options include superconducting circuits, cryogenic ions, and light. With such diverse components, how does one assess their differences? This aspect will hone in on each quantum computer’s performance.

This marks a shift from the early days, where success was measured by the number of qubits—the foundational elements of quantum information processing. Many research teams have surpassed the 1000-qubit threshold, and the trajectory for achieving even more qubits appears to be becoming clearer. Researchers are exploring standard manufacturing methods, such as creating silicon-based qubits, and leveraging AI to enhance the size and capabilities of quantum computers.

Ideally, more qubits should always translate to greater computational power, enabling quantum computers to tackle increasingly complex challenges. However, in reality, ensuring each additional qubit doesn’t impede the performance of existing ones presents significant technical hurdles. Thus, it’s not just the number of qubits that counts, but how much information they can retain and how effectively they can communicate without losing data accuracy. A quantum computer could boast millions of qubits, but if they’re susceptible to errors that disrupt computations, they become virtually ineffective.

The extent of this “glitch” or noise can be measured by metrics like “gate fidelity,” which reflects how accurately a qubit or pair can perform operations, and “coherence time,” which gauges how long a qubit can maintain a viable quantum state. However, we must also consider the intricacies of inputting data into a quantum computer and retrieving outcomes, despite some favorable metrics. The growth of the quantum computing industry is partly attributed to the emergence of companies focused on qubit control and interfacing quantum internals with non-quantum users. A thorough buyer’s guide for quantum computers in 2025 should encompass these essential add-ons. Choosing a qubit means also selecting a qubit control system and an error correction mechanism. I recently spoke with a researcher developing an operating system for quantum computers, suggesting that such systems may become a necessity in the near future.

If I were to create a wish list for the short term, I would favor a machine capable of executing at least a million operations: a million-step quantum computing program with minimal error rates and robust error correction. John Preskill from the California Institute of Technology refers to this as the “Mega-Quop” machine. Last year, he expressed confidence that such machines would be fault-tolerant and powerful enough to yield scientifically significant discoveries. Yet, we aren’t there yet. The quantum computers at our disposal currently manage tens of thousands of operations, but error correction has only been effectively demonstrated for smaller tasks.

Quantum computers today are akin to adolescents—growing toward utility but still faced with developmental challenges. As a result, the question I frequently pose to quantum computer vendors is, “What can this machine actually accomplish?”

In this regard, it’s vital to compare not only various types of quantum computers but also contrast them with classical counterparts. Quantum hardware is costly and complex to manufacture, so when is it genuinely the sole viable solution for a given issue?

One method to tackle this inquiry is to pinpoint calculations traditional computers cannot resolve without unlimited time. This concept is termed “quantum supremacy,” and it keeps quantum engineers and mathematicians consistently preoccupied. Instances of quantum supremacy do exist, but they raise concerns. To be meaningful, such cases must be applicable, facilitating the construction of capable machines that can execute them, while also being demonstrable enough for mathematicians to assure that no conventional computer could compete.

In 1994, physicist Peter Shor devised a quantum computing algorithm for factoring large numbers, a technique that could potentially compromise the prevalent encryption methods utilized by banks worldwide. A sufficiently large quantum computer that could manage its own errors might execute this algorithm, yet mathematicians have yet to convincingly demonstrate that classical computers can’t efficiently factor large numbers. The most prominent claims of quantum supremacy often fall into this gray area, with some eventually being outperformed by classical machines. Ongoing demonstrations of quantum supremacy appear currently to serve primarily as confirmations of the quantum characteristics of the computers accomplishing them.

Conversely, in the mathematical discipline of “query complexity,” the superiority of quantum solutions is rigorously demonstrable, but practical algorithms remain elusive. Recent experiments have also introduced the notion of “quantum information superiority,” wherein quantum computers solve tasks using fewer qubits than traditional computers would require, focusing on the physical components instead of time. Though this sounds promising—indicating that quantum computers may solve problems without extensive scaling—they are not recommended for purchase simply because the tasks in question often lack pivotal real-world applications.

It’s undeniable that several real-world challenges are well-suited for quantum algorithms, like understanding molecular properties relevant to agriculture or medicine, or solving logistic issues like flight scheduling. Yet, researchers lack full clarity on these applications, often opting to state, “it seems.”

For instance, recent research on the prospective applications of quantum computing in genomics by Aurora Maurizio from the San Raffaele Scientific Institute in Italy and Guglielmo Mazzola at the University of Zurich suggests that traditional computing methods excel so significantly that “quantum computing may, in the near future, only yield speedups for a specific subset of sufficiently complex tasks.” Their findings indicate that while quantum computers could potentially enhance research in combinatorial problems within genomics, their application needs to be very precise and calculated.

In reality, for numerous issues not specifically designed to demonstrate quantum supremacy, there exists a spectrum in what constitutes “fast,” particularly when one considers that quantum computers might ultimately run algorithms quicker than classical computers, despite overcoming noise and technical challenges. However, this speed may not always offset the hardware’s significant costs. For example, the second-best-known quantum algorithm, Shor’s search algorithm, offers a non-exponential speedup, reducing computation time at a square root level instead. Ultimately, the question of how fast is “fast enough” to justify the transition to quantum computing may depend on individual buyers.

While it’s frustrating to include this in a purported buyer’s guide, my discussions with experts indicate that there remains far more uncertainty about what quantum computers can achieve than established knowledge. Quantum computing is an intricate, costly future technology; however, its genuine added value to our lives remains vague beyond serving the financial interests of a select few companies. This might not be satisfying, but it reflects the unique, uncharted territory of quantum computing.

For those of you reading this out of the desire to invest in a powerful, reliable quantum computer, I encourage you to proceed and let your local quantum algorithm enthusiast experiment with it. They may offer better insights in the years to come.