Erik Garcell is all in on quantum computing. He applauds the strides in hardware innovation in recent years that has pulled the compute industry closer to the commercial quantum era, understands the power of generative AI to help smooth the road forward, and is a firm believer in the promise that quantum computing holds for everything from scientific discovery to business profits in most every industry.

None of that should be surprising, given Garcell’s position as director of quantum solutions architecture for Classiq, a six-year-old company headquartered in Tel Aviv that offers a platform designed to make it easier and faster for developers to build quantum algorithms and applications at scale. It is a job that’s both technical and evangelical.

That said, he also believes the scientific community and enterprises need to understand that – as happened with classical computing – commercial quantum computing is something that will start relatively small, with the benefits rippling out over the years as development of the hardware and software allows. Microsoft and others have talked recently about the age of fault-tolerant, commercial quantum computing essentially being right round the corner, but the benefits will take time to spread.

He is hearing the same concerns now that were voiced as traditional systems evolved.

“’They take up whole rooms, they're big, they take a lot power,’” Garcell told The Next Platform during Nvidia’s recent GTC 2026 conference. “Yeah, they do. And the applications for them initially are going to be very small, the same way that regular computers were. It was initially a banking tool. You were just doing calculations. Banks were using them. As the power of classical computers increased, more bits were available, you could do more things. More applications became available and the ROI started to spread to more and more industries, more and more applications. We're going to see the same evolution with quantum computing. It's going to be applicable for a certain set of small applications. Few people will start making ROI on it soon. As the hardware continues to scale, more and more industries and use cases will open up naturally. It's going to follow these same kinds of footsteps, and that's just because the hardware is scaling up little by little.”

The software – and the algorithms used by developers to create them – is where Classiq comes in. The vendor’s goal, through its platform, its Python-like Qmod (Quantum Modeling Language) and GitHub library of open code examples, is to make it easier for developers to more quickly create quantum software that can run across multiple quantum systems that use different modalities, including superconducting circuits, trapped ions, neutral atoms, cat, and spin qubits. With Classiq, developers only have to create their programs once rather than recode them for each system.

Abstracting And Automating

The platform abstracts and automates much of the lower-level work, and the development toolkit lets users building, debug, and visualize their quantum circuits. It also analyzes performance of the simulations or quantum hardware runs. In August 2025, the company made its library and SDK compatible with AI agents, including Anthropic’s Claude Code. Classiq knows that developers rarely read the entire manual for their or anyone else’s programming language, Garcell said.

“We trained Claude on it to help our users get up to speed a little bit faster, build out some code without having to know everything in our documentation,” he said. “You want to find a code example that looks similar to what you're doing, and then you want to build up from there. That's why we have our code library, that's why we have AI agents that help support the ways people actually learn and use quantum, use any programming languages. We don't want quantum computing to be different or novel. We want it to be similar – the same [as developing for classical systems], easy to develop, easy to use – and that's what's going to foster adoption at the end of the day.”

Integrating With Nvidia

At GTC, Classiq added more capabilities by integrating its software development platform with Nvidia’s CUDA-Q, an open an hybrid quantum-classical platform that developers can use to simulate and program QPUs from multiple quantum hardware makers through common languages like Python and C++ and to accelerate their work through Nvidia GPUs.

Quantum computing is going to be part of the larger HPC stack, working alongside CPUs, GPUs, and other specialized processors, Garcell said. Classiq’s integration with CUDA-Q is “quite crucial,” he added.

“We just have a new compute resource, a QPU,” he said. “In this way, developers need to have this resource and access this resource so they can send the right kinds of jobs to the right compute resource. [The QPU] needs to be integrated into this HPC stack and we need workload managers to be able to parse these jobs back and forth easily. Quantum computing by itself can do a lot of things, but it's absolutely more powerful in collection with other algorithms, especially if you're looking for the near-term applications of quantum. There is not going to be quantum by itself. It's going to be a quantum algorithm working in tandem with a classical algorithm to either improve the fidelity or resolution of that result, or load that data in maybe more officially in certain ways. But it's this integration that's needed.”

Classiq has worked with other vendors to build hybrid workload managers, such as modifying Slurm to work on HPE’s Cray supercomputers, Garcell said. However, with CUDA-Q, the vendor can make such work a single line of code, turning a quantum circuit that Classiq generates into a CUDA-Q kernel that can be parsed among different machines. It becomes an object that can be easily transferred and worked in the CUDA-Q workflow, speed up and increasing compatibility between quantum and classical systems.

A Hybrid Future

It goes back to widely understood concept that quantum computing, particularly in the early years, will be part of larger hybrid systems with traditional systems, and that the software will need to be able to run across the hardware.

“’Quantum’ is a fun word, it's a great marketing word, but it tends to be a word that makes it sound far more scientific and almost intimidates people,” Garcell said. “The hardware does get to some very scientific places, but if you're a programmer for this, it's computer science. You have to build the integrations, you have to make it work in the full stack, you have worry about data transfer, you have the worry about how to efficiently code your information into a system. Every problem that classical computing experts have for programming, they need to now consider for quantum.”

Nvidia is the latest in a growing lineup of partners for Classiq that includes Microsoft. The vendor in February announced a demonstration with AMD and Comcast that showed using quantum algorithms could improve the routing resilience in networks by determining independent backup paths for network sites when maintenance or other changes were underway. A month before, Classiq said it was partnering with C12, a European that builds spin qubit quantum processors. Developers using Classiq’s platform could use the Qmod language and synthesis engine to design, compile, and test their quantum algorithms on Callisto, C12’s digital twin of its in-development hardware.

Classiq’s technology also is now available on the Amazon Web Services (AWS) marketplace. The company has raised more than $200 million in funding, with AMD Ventures, Qualcomm Ventures, and quantum hardware maker IonQ.

Garcell is confident in the future of quantum as he talks to Classiq users. A key reason is that he can trust the roadmaps of the hardware makers.

“If you look at IBM's roadmap, they've been keeping their promises since 2016, since they put their first quantum computer on the cloud,” he said. “Year after year, they have actually delivered on the hardware they said they would have by that year. If they say they're going to have a certain hardware in three years with a certain capacity, they've built enough trust that I can trust that they will have that. It lets me, as a developer of this program, working with clients, help them figure out, 'All right, this algorithm you're working on that you want to have in production, I know that these vendors say they are going to this hardware by this year. That hardware will have enough compute capacity to do this in production.’”