A Better Way to Fail: Quantum Computing and Drug Discovery

Key takeaways

• Using quantum compute to raise the success rates of clinical trials by 10 percentage points could cut a third of the cost of bringing a drug to market, potentially allowing 50% more therapies to find their way to patients.

• The economic viability of this solution is bounded by the throughput and operating cost of the quantum computer, both of which vary radically with qubit modality.

• Diraq’s systems are designed to be economical to run and compact to deploy, even at utility scale, and that comes with the added benefit of information security.

Right now, somewhere in the world, a person with Parkinson's disease is waiting for a therapy that doesn't yet exist. Others are enrolled in clinical trials for new cancer therapies that could eventually become standard of care. Bringing any one of those medicines to market requires years of work and billions of dollars of investment.

The idea that all this time and money could be saved is one of the most popular justifications for building a quantum computer. But where exactly does quantum computing factor into the pipeline process from drug discovery to patient delivery? And will it generate enough value to justify the price tags expected for some quantum computers?

A 2010 analysis published in Nature Reviews Drug Discovery offers an eye-opening perspective on these questions, despite the fact that quantum computers were still merely science experiments at the time of publication. Digging into the data reveals what Diraq has long known: quantum computers need to be economical and deployable to achieve true utility.

But first, a primer on the pharmaceutical pipeline.

Pharma’s giant funnel

The process of drug discovery begins with the target-to-hit stage: the task of finding a protein that’s implicated in a disease like Parkinson's and then searching for a molecule that interacts with that target in exactly the right way.

Classical computers already do the heavy lifting here, sifting through millions of candidate molecules before a single compound is even synthesized in the laboratory. Most of these molecules are discarded almost immediately, and those that survive face increasingly demanding tests.

Next up is the hit-to-lead stage, where researchers attempt to separate genuinely promising drug candidates from molecules that merely looked good on paper. By the end, only a small number of molecules remain worthy of serious attention.

Then comes lead optimization, where researchers modify molecules, synthesize them, test them experimentally, analyze the results, and repeat the process over and over again. The aim is to create a molecule that works, but is also safe, stable, manufacturable, and capable of surviving the long journey through clinical trials.

This is where the economics becomes brutal. Every successful drug stands on a mountain of discarded molecules, abandoned candidates, failed experiments, and unsuccessful clinical trials.

Enter quantum

The potential value of a better computer model is to help researchers make better decisions sooner in the drug discovery pipeline. The difficulty in improving these decisions is that the underlying chemistry is extraordinarily hard to calculate, and this is because of (you guessed it) quantum mechanics.

Many of a molecule’s properties emerge from the interactions between its electrons, which influence everything from its shape and reactivity to how strongly it binds to a protein. In principle, we can calculate these properties from scratch, but the mathematics is so difficult that even the world's largest supercomputers are forced to rely on approximated models.

Most of the time, those approximations work remarkably well, but they are still only approximations. And when researchers are trying to distinguish between a handful of promising candidates, tiny differences in molecular energy can determine which molecule advances and which is discarded.

A slightly more accurate prediction might not sound revolutionary. But in drug discovery, the economics is highly asymmetric. The cost of performing an additional calculation is tiny compared with the cost of selecting the wrong candidate and carrying it through years of experimental work and clinical trials.

The economics of avoiding failure

This is the context of the 2010 analysis, in which researchers at Eli Lilly asked which improvements would have the greatest effect on the total cost of delivering an approved drug. Their model estimated a capitalized development cost of US$1.78 billion per launched medicine and then they systematically varied parameters throughout the development process to determine which mattered most.

The results were striking: the greatest benefit came from preventing weak drug candidates from reaching expensive clinical trials. In fact, improving clinical success rates by just 10% across each of three phases could cumulatively reduce capitalized development costs by more than US$600 million for every medicine that ultimately reaches patients.

That’s a third of the total cost in 2010 terms. Reinvesting that money saved could potentially result in 50% more new medicines finding their way to patients. Much of that value comes from avoiding dead ends earlier, allowing companies to focus their resources and researchers on the therapies most likely to succeed. It also means fewer patients having to participate in trials that are destined to fail.

But it’s not obvious from the 2010 study that a 10% improvement in trial success is possible, nor that it will actually represent a significant return on investment once the cost of the quantum computer is taken into account.

The capitalized savings achievable with a quantum computer is dependent on many factors, including the throughput of the machine (how many logical quantum operations it can perform in a given timeframe) and the cost of operating it.

What a quantum computer should cost

The quantum computers available to buy today are already expensive. Market data from The Quantum Insider put the median for on-prem systems at about US$10 million, across tracked sales of 175 systems. Superconducting systems average US$26 million with a mean qubit count of 74, and trapped-ion systems are around US$19 million for a similar average number of qubits.

These are small machines, a long way from the millions of qubits needed for useful drug discovery. No vendor has published a price for a utility-scale system, but the economics point one way.

Most quantum computers grow by adding modules, and each module needs its own fridge, lasers or vacuum system. Those support systems cost the most, and they cost roughly the same no matter how many qubits are inside. To get from hundreds of qubits to millions, you multiply the modules, and the bill grows with them. A single fridge can cost millions of dollars. At that rate, some utility-scale designs could cost billions, and a drug discovery team simply can't buy a machine like that.

Diraq's qubits are modified transistors made in commercial foundries, and a chip with millions of qubits is roughly the size of the eight-qubit chip that we’ve just installed inside a shared commercial data center. It goes in a single fridge that’s about the same size as a standard server rack. Because the cost of the cryogenics stays flat while qubit count grows, Diraq's system trends toward US$1 per qubit at utility scale.

For pharma, that changes the business case. A billion-dollar machine has to serve dozens of drug programs to pay for itself. A machine with a much lower price tag can likely earn its keep on a few. And Diraq’s early calculations show that serving any more than a handful of drug programs might be too tall an order, even for a utility-scale machine.

Co-location is key

The other crucial thing to consider when integrating quantum computers into the drug discovery pipeline is how deployable they are.

The molecules that pharma companies care most about are often worth hundreds of millions, or even billions, of dollars in future revenue. These programs are built on years of R&D and proprietary biological insights, so companies are understandably cautious about protecting their intellectual property.

This creates a strong incentive to deploy quantum computers within existing enterprise environments. A compact system that can be installed alongside existing infrastructure allows organizations to retain control over their data, security policies, and compliance processes while still benefiting from quantum-enhanced workflows.

It is one thing to run routine calculations on external infrastructure. It is another to entrust your most commercially important molecules to systems beyond your direct control. Compact, deployable quantum computers offer a different model: bringing the computation to the data rather than sending the data to the computation.

A real challenge is making this degree of utilization possible. Quantum calculations represent only a small but highly valuable part of the overall process, so a quantum computer that is isolated from existing infrastructure will struggle to achieve the utilization needed to justify its cost.

By contrast, a quantum processor that can sit directly within the computational workflow can be invoked whenever a calculation reaches a point where quantum advantage exists. Higher utilization translates directly into a lower effective cost per program, improving the economics of the platform.

Diraq's silicon spin-qubit architecture is designed with this deployment model in mind. It is compact and compatible, creating a path toward tightly integrated quantum–classical computing.

The post-COVID correction

Since the 2010 article was published, drug-development costs have continued to rise, in part due to the accelerating effect of COVID. Deloitte's 2025 analysis of major pharmaceutical companies reported an average development cost of approximately US$2.23 billion per asset in 2024, substantially higher than the US$1.78 billion baseline used in the 2010 model.

Other recent analyses point to the same trend. A 2024 JAMA commentary noted that widely cited estimates for the cost of developing a new medicine have increased dramatically over the past two decades, with some studies placing the capitalized cost in excess of US$3 billion.

The central message of the 2010 study therefore remains as relevant today as it was when it was first reported: technologies that improve decision quality early in the pipeline can create disproportionate value downstream. The challenge is to ensure that the cost of those technologies remains lower than their economic benefits.

The hardest question in pharma

For all the excitement surrounding quantum computing, building a business case inside a pharmaceutical company is far from straightforward.

A computational chemist might look at quantum computers and immediately see the potential: more accurate molecular models, better candidate selection. But the business leaders are tasked with allocating finite resources to myriad new technologies, including AI and robotics.

And these investments are ultimately reliant on their potential to create competitive advantage up to 10 years into the future when the drugs actually make it to market. Few decisions are harder than choosing between technologies that are all improving rapidly and whose ultimate impact remains uncertain.

This is why economics matters as much as physics. It is not enough for a quantum computer to perform calculations that classical computers cannot. Pharma companies must be confident that those calculations create sufficient value to justify the investment.

Against that backdrop, the case for quantum computing is ultimately the same as the case for any other enabling technology in drug discovery: can it help researchers make better decisions today that translate into better medicines tomorrow? With a quantum computer built the way the chips in today's servers are built, the answer is a resounding yes.