IBM: Quantum Computing Enters Industry with Hybrid Supercomputing
Quantum computing is moving out of the phase where it was mainly the domain of physicists, laboratories, and proof of concept. It is not yet a technology ready for large-scale production, and it would be wrong to portray it as such. However, it is no longer just a topic to be confined to pure research. The trajectory indicated by IBM is to build a computing infrastructure in which classical computers, GPU accelerators, and quantum processors work together, each on the part of the problem for which it is best suited.
This is the model that IBM calls quantum-centric supercomputing. So, it’s not an isolated quantum computer, nor a replacement for existing architectures. Rather, it is an extension of supercomputing: the CPU will continue to do what it does well, the GPU will remain central for many parallel workloads, and the QPU, the quantum processing unit, will be used in algorithm segments where the quantum nature of the problem makes classical computation inefficient.
There are many applications in the classical world that have taken decades to become reality simply due to a lack of computational power: one example is artificial intelligence, which has become feasible due to the availability of computational power and data. However, others are simply intractable with classical processors and require a new approach: this is where quantum computing comes into play, which, however, requires new algorithms and new ways to tackle and break down problems.
“AI has relied on what preceded it. The techniques we knew for decades, but the processors became fast enough to use them on a large scale. Quantum is a completely new way of addressing things,” explains Adam Hammond, head of business development for quantum computing in IBM’s EMEA region.
Quantum computing: materials, batteries, electric networks, and logistics are already the first fields of work
The point is not to ask whether quantum will be useful in the abstract. The correct question is: where are classical computers already working by approximation, with high times, high costs, or overly simplified models? This is where IBM sees the first industrial spaces.
In manufacturing and automotive, for example, quantum can enter into the development of new materials and batteries for electric vehicles. Not because a QPU can single-handedly replace the entire design process, but because it can improve some steps currently managed with inevitably approximated classical models.
“The type of problems that quantum computing allows us to tackle is highly relevant for industrial companies. If we think of companies like Stellantis, we are talking about new materials, new ways to organize factories, and optimization,” says Hammond.
Another area is energy. The transition to renewables makes grid management more complex: generation is less predictable, consumption changes, and electric cars introduce new distributed loads. IBM is collaborating with E.ON in Germany on optimizing the electric grid, a case that clearly illustrates the logic of quantum: not producing a magic answer, but improving the capacity for prediction and allocation in a system with many variables.
The same goes for logistics. Managing airport gates, scheduling crews, recovering after a railway or air blockage are optimization problems already addressed today with classical techniques. But the number of combinations grows rapidly and makes it difficult to obtain truly efficient answers in a timely manner.
The most concrete promise: better simulating what is still physically tested today
Another area where quantum computing can make a difference is computational fluid dynamics (CFD). Automotive companies (including Dallara) and aerospace companies already use simulations before going to the wind tunnel. According to IBM, quantum will not eliminate this process. However, it could make it more accurate and further reduce the number of physical tests necessary.
The point is not only speed, but simulation quality. Classical models work, but often require approximations. With sufficiently large quantum machines, the promise is to better simulate the behavior of complex physical systems: aerodynamics, computational chemistry, materials, molecular interactions.
“I think it will be more accurate,” says Hammond when talking about fluid dynamics. “It’s not about going from nothing to something, but about taking what we do today and extending it, building greater accuracy, perhaps being able to do it faster or more conveniently.”
The industrial advantage, if it arrives, will be very practical: bringing fewer variants to the wind tunnel, physically testing only the most promising models, reducing development times, and prototyping costs. This same reasoning holds in chemistry and pharmaceuticals: even small improvements in the simulation process can translate into years saved in the discovery of new drugs.
Quantum-centric supercomputing: computing becomes a hybrid chain
IBM's vision is not to build a quantum computer that works on its own. It is to create a hybrid computing chain. A problem is divided into parts: one runs on the CPU, one on the GPU, one on the QPU. The results are then recomposed into a single workflow.
“You take the problem and divide it into the part that runs best on a CPU, the part that requires a GPU, and the part that requires a QPU. The right part of the problem is executed on the right processor, and then the results are combined again,” states Hammond.
For the end user, in perspective, quantum could become almost invisible. A chemist will continue to use their software library. An engineer will continue to work with their simulation tools. The system, if it has access to quantum resources, will decide which parts of the calculation to delegate to the QPU.
Today it is not yet like that. Those working on these cases must understand the workflow, identify bottlenecks, and evaluate where a quantum algorithm can bring improvement. But the goal is clear: to take quantum out of the niche of specialists and bring it into tools already familiar to technical users.
Also, the lock-in issue is central. The quantum market is still fragmented: IBM, Google, Microsoft, specialized startups, and other players work on different approaches. However, Hammond downplays the risk of total fragmentation, at least for gate-based architectures. The reason is that the quantum circuit model already offers a common language. Furthermore, Qiskit, IBM's software framework, is open source and aims to support hardware from multiple manufacturers.
For IBM, the goal is to be a central player in hardware while keeping the software layer open enough to attract developers, algorithms, and applications. It is an ecosystem game, not just about chips.
The year 2029 is a key point in the roadmap, but it doesn’t mean quantum for everyone and everything The recurring date is 2029. IBM identifies it as the horizon for Starling, the first fault-tolerant quantum system in its roadmap. Fault tolerance is the pivotal step: today’s quantum computers are still subject to errors and cannot support complex calculations on a production scale. Error correction serves to make longer circuits and more extensive problems reliable.
Caution: 2029 should not be considered as the year when any company will be able to use quantum for any workload. It should be viewed as a technical turning point if the roadmap is adhered to: more stable machines, larger problems, a greater chance of approaching industrial use cases with measurable returns.
Hammond is cautious:
“Today’s quantum computers allow us to solve very complex problems, but still not on a scale that we can put into production. Fault-tolerant quantum computers will give us the potential to look at real-time.”
Real-time is relevant for dynamic optimization cases: restoring a railway network after a failure, reorganizing crews and passengers after a blockage, optimizing electrical loads in the presence of variable generation. These are problems where it is not enough to find a good solution; it must be found quickly.
For CIOs, the urgent issue is not quantum computing. It’s cryptography
As industrial quantum progresses towards more mature applications, security is already a concrete issue. The reason is well known: a sufficiently powerful quantum computer could compromise widely used cryptographic algorithms today. The exact date is debated, but for companies, banks, governments, and critical infrastructures, the problem cannot be postponed.
The reason is simple: many encrypted data must remain protected for ten, twenty, or fifty years. Health data, financial data, passports, energy networks, information on critical infrastructures. Even if there is no machine capable of decrypting them on a large scale today, the risk is that they are collected now and decrypted in the future. From an algorithmic point of view, the path already exists. Post-quantum standards have been defined and enterprises can begin migration. The real problem is organizational and infrastructural.
“Technologically it is a solved problem and any company could become quantum-safe tomorrow. In reality, it’s a massive business transformation because most organizations do not manage their own cryptography in the same way they manage their software,” states Hammond.
Cryptography is not just in central systems. It’s in routers, wireless access points, applications, devices, credit cards, protocols, legacy systems. Often companies do not even have a complete inventory of where it is used. And without an inventory, there is no migration.
From here comes the concept of crypto-agility: managing cryptography as an upgradable, versioned, monitored component. Not an invisible piece of infrastructure that remains static for years, but a living part of the IT stack.
“Become crypto-agile, but start the process today,” says Hammond. This is probably the most operational message for CIOs: don’t wait for the quantum computer capable of breaking current algorithms. The migration will take years, especially in large organizations.
Europe is very focused on hardware, but economic value will be in applications
Then there is the European issue. Europe is investing in quantum, with a strong focus on technological sovereignty and building a local hardware industry. Relying on a few non-European suppliers for a strategic technology would be an industrial and geopolitical risk. However, according to Hammond, this attention is not enough. The economic value will not come only from the capacity to build quantum machines. It will mainly come from adoption in production sectors: manufacturing, energy, logistics, banking, pharmaceuticals, healthcare, oil & gas.
“The real value of computing is not in the IT industry itself, but in the value it adds across all sectors,” observes Hammond. It is the same reasoning that guided the adoption of cloud and IIoT: the impact is not only on technology providers but on the productivity that technology enables in user enterprises.
There is also a skills issue. Quantum must exit the perimeter of PhD physicists and become a standard subject in computing, engineering, applied mathematics curricula. Today, very few people can program quantum computers. Too few if the goal is to bring these systems into industrial applications.
“We are still far from the day when computer science students will have a module on quantum as if it were normal. And that’s where we need to get,” says Hammond.
Quantum readiness begins before commercial availability
For companies, the point is not to buy a quantum computer today. The point is to understand where quantum could have an impact tomorrow. This means mapping processes with computational bottlenecks: simulations that are too slow, overly approximated models, intractable optimization problems, costly R&D cycles, workflows where even a small improvement can yield significant returns.
This is quantum readiness, that is, being prepared for the arrival of quantum computers: not a theoretical exercise, but industrial preparation: data, models, skills, partnerships, use cases, return metrics. Companies that start earlier will not necessarily have an advantage because they will use the QPU sooner. They will have an advantage because they will already know where to use them.
The second line of work is even more urgent: quantum-safe migration. Here it is not necessary to wait for quantum computing maturity. It is necessary to start with the cryptographic inventory, identify exposed areas, plan the replacement of vulnerable algorithms, and build a model of crypto-agility.
In short, quantum should be viewed on two levels. As a computing technology, it is an industrial bet that will be played over the coming years on integration with classical supercomputing. As a threat to traditional cryptography, it is already an IT governance issue.