Will Quantum Software Eat the World?

The emergence of quantum computing hardware in recent years has produced an explosion of quantum startup businesses hoping to cash in on the burgeoning industry. At present, there are well over 100 private companies focused on quantum computing, not counting the large public companies, government entities, and academic groups who are also pouring resources into the technology. While this avalanche of funding, combined with the lack of any practical applications at the moment, leads many to worry about a coming “quantum winter”, I want to explore a different aspect here: what does the interplay between quantum hardware and quantum software look like, and how can we expect this to evolve over time? Or, to borrow a phrase from Marc Andreessen, will quantum software eat the world?

Quantum circuit

Technology strategy: software and hardware

A few years ago, I took a detailed look at the scenarios in which technology companies decide to build complementary hardware and software products, and when these efforts are most likely to succeed. In general, the strategy pays off when the software product is:

(a) essential to the use of the hardware product, that is, either it is the only software in existence, or it is fundamentally better than other existing software options (think of early mainframe computers),
(b) tightly integrated with a piece of specialized hardware to perform a function more efficiently than more general-purpose devices can accomplish (think of iPod and iTunes in the early days), and/or
(c) core to the business, with the hardware product serving as a funnel to anchor users in the company’s ecosystem (think of Surface, Windows, and Office).

So how can we apply these insights to the world of quantum computing? Specifically, should we expect the companies building quantum hardware to have an inherent advantage in building the software for those devices, or is it possible that companies focused solely on software and applications will be able to achieve dominance?

Quantum computers are specialized devices

Here, I think it’s worth pointing out a key difference between traditional (classical) computers and quantum computers: Quantum computers are unlikely to ever be used as general-purpose computing devices — at least for the next several decades. Yes, in theory, a quantum processor can do anything a classical processor can do, but our classical computing technology is so fast, so small, so cheap, and so satisfactory for most purposes, that it will require many orders of magnitude of advances in quantum technology to begin to even consider replacing classical computers in the broad sense. So, for now, we can think of quantum computers as special-purpose devices that are optimized for a few particular tasks — essentially, as “co-processors” whose operation will be controlled by a classical computing framework. This certainly describes the current area of research into near-term “hybrid quantum-classical” systems, but it will continue to be true even after quantum computers reach the stage of fault-tolerance and universality.

With this in mind, it’s relatively easy to see that the companies building quantum hardware should also be building the core software for these devices, since the scenario falls neatly into cases (a) and (b) in the framework laid out above. The two are necessarily intertwined, and it’s impossible for someone else to effectively build the core software — especially in a world where companies don’t even sell the quantum computing hardware itself, but only sell access to it over the cloud.

The future rise of quantum software and applications

So is that the end of the story? Should all of the quantum software companies cease to exist? Of course not! The quantum computing hardware companies must build the core software, but this software is more analogous to a device driver — it provides access to the underlying hardware, but it doesn’t necessarily provide value on its own. It takes applications to do this. In fact, I would argue that in the most likely scenario, quantum software and applications companies will extract most of the value from the quantum computing market in the long-term — that is, I would argue that quantum software will indeed “eat” the quantum world. Reasons for this include:

Quantum computers will eventually become a commodity. Once quantum hardware has reached the stage of fault-tolerance, device-specific attributes become much less important. More and more companies will begin building universal quantum computers whose capabilities are all roughly equivalent. At this point, it helps to recall the history of the personal computer: IBM enjoyed enormous success initially, but as more and more companies began producing systems with equivalent capabilities, the unique value of the IBM PC began to disappear. This dynamic allowed a software company, Microsoft, to rise and become one of the dominant beneficiaries of the explosion of the PC industry.

Most applications for quantum computers have not been discovered yet. If we already knew everything a quantum computer could do, then the companies building the quantum computers could also build the applications, and everything would be done. But this certainly is not the case. In fact, most quantum computing companies today are focused on near-term algorithms and applications, and it’s not even clear yet whether there is anything useful that these near-term quantum computers can do. It’s far more likely that the most relevant applications will emerge later, once the hardware has matured and we have fault-tolerant devices. At that point, since the hardware will have already begun to commoditize, the companies discovering and developing these applications will likely be able to provide the most value to the marketplace.

Please comment with your thoughts and perspectives!

Why Quantum Winter Is Not Coming

Amid the recent explosion of startups and venture capital investment into quantum computing, there has been much talk of an inevitable “quantum winter”.

Map of Winter Storm Quantum

No, this is not some bleak doomsday scenario where our enemies win the race to develop a quantum computer and thrust our society into a winter of defeat and despair.

Instead, it’s the fear that the hype around quantum computing will far outpace the realities, investors will get frustrated by the failure to meet the inflated expectations, and funding for the industry and associated research will collapse. Let’s take a look at the basis for this fear.

The famous AI winter

The most famous “winters” through the history of technology were the AI winters of the late 20th century. From the invention of the first learning algorithm — the perceptron in the late 1950s — the popular press was enamored with the potential of this new breed of technology. Famously, the New York Times reported in 1958 of the potential for a computer that “will be able to walk, talk, see, write, reproduce itself and be conscious of its existence.”

Government funding for AI research exploded in the 1950s and 1960s, but people got frustrated (no robots yet? where are my robots?!), funding was cut, and the 1970s became sort of the first “winter” for AI. By the 1980s, funding had returned, and the field was again on an upslope. But experts worried that hype was again outpacing reality, and indeed, the late 1980s and 1990s brought another collapse in funding and the failure of many AI-focused companies.

Why we fear a quantum winter

Levels of government funding and industry investment in quantum computing are unprecedented, and it seems nearly weekly that some announcement of new funding is made. But those of us in the field know all too well that, for practical purposes, quantum computers are still completely useless. Sure, there’s a ton of great work being done which will pay dividends in the future, but most realists don’t expect widespread quantum adoption for practical problems for many years or even decades. Will investors be patient? We hope so. For every hypefilled article, there are plenty of experts trying to manage expectations and avoid inevitable disappointment.

This is by no means a unique situation. Gartner publishes a “Hype Cycle” every year for emerging technologies, with a prominent “Peak of Inflated Expectations” — a typically crowded list of over-hyped technologies just waiting for their proverbial bubble to burst. In their most recent analysis, quantum computing is still on the rising edge of this curve, almost unnoticeable among a tidal wave of AI-related technologies (seems like the AI spring has sprung).

So at some point, we should certainly expect the hype around quantum computing to subside. This is the typical trend for new technologies, and building a quantum computer is a long slog that has a much more extended time frame than, say, the next blockchain. The current hype is unsustainable. But does that also mean that, like with AI in the past, funding will dry up? Is a quantum winter is around the corner? I would argue that this is extremely unlikely.

Why quantum computing will not suffer the fate of AI

1. Technological maturity. Quantum computing today is a far more mature field than AI was in the 1950s. Modern AI research didn’t really begin until around the 1940s, and so the field was only about a decade old when the first massive wave of investment came in the 1950s. People had grand dreams, but no one knew what AI would truly be capable of.

By contrast, quantum computing research began in earnest in the 1980s (spurred in part by Richard Feynman), and so at this point the field has nearly four full decades of research behind it. And the technological feasibility of quantum computing is not just wishful thinking (some people would beg to differ with this statement, but they are a small minority). The principles of quantum physics underlying quantum computing have been around since the 1920s and have been experimentally tested many times over the last century, with astonishing success. And quantum error correction — the key to making quantum computers fault-tolerant and scalable — has been on a firm mathematical foundation since Shor and Steane developed their codes in the 1990s.

2. Frankenstein’s monster. It’s easy for people’s imaginations to run wild when thinking about robots. When the perceptron was invented in the 1950s, no one had any realistic plan for developing a conscious machine. But people bought into this idea, in part because it had been the stuff of science fiction for so long. People had an intuitive idea of what this technology could look like, and what impact it could have. If you were expecting a walking, talking, reproducing robot, and all you got are a few algorithms that can classify images, you’d lose faith, too.

Most technologies are unable to capture the imagination like AI. This includes quantum computing. Sure, there is an international spy novel written about quantum computing (currently ranked #1492 on the Espionage Thrillers bestsellers list at Amazon!), and there are plenty of misunderstandings of what a quantum computer will be able to do, but we don’t run a serious risk of investors being influenced by their knowledge of science fiction.

3. Factoring, factoring, factoring. For those of us in the field, it has become cliché to mention Shor’s algorithm, by which quantum computers will be able to quickly factor extremely large integers, and thereby break the RSA encryption scheme that is used to secure basically everything on the Internet. And while most algorithms research today is focused on near-term applications of smaller (“NISQ-era“) quantum computers, it’s impossible to overstate the importance of factoring to the field as a whole. Essentially anything that’s transmitted over the Internet today (or for the foreseeable future, until a quantum-safe encryption standard broadly replaces RSA) — if an attacker wants to decrypt it, all they have to do is store the encrypted data and wait for a fault-tolerant quantum computer to exist. Sure, this may be decades away — but maybe not. The potential value to industry investors is enormous, and will no doubt be worth the risk and the wait. And no government can afford to take the chance that a rival nation might get a quantum computer first. As long as factoring remains the killer app of quantum computing (and assuming no one discovers an efficient factoring algorithm for classical computers), it’s hard to envision a scenario where funding for quantum computing dries up, despite the inevitable decline in hype.

Thoughts? Leave a note in the comments and I’d love to discuss.