When people think about quantum computing, they mostly focus on the qubit’s physical aspects and challenges. At Quantum Machines, we take a wholistic view and spend a lot of time thinking about something else: the classical computing infrastructure needed to control them. The control electronics must act on qubits with nanosecond precision before their quantum states decohere. That demand for ultra-fast, tightly synchronized control is exactly what shapes how we design our systems. 

That is also why we work closely with AMD technologies across multiple layers of our Hybrid Control architecture. From real-time pulse processing inside the quantum runtime to large-scale quantum error correction decoding and optimization workloads, AMD hardware helps us bridge the gap between nanosecond-scale control and millisecond-scale computation. 

One quantum computer, multiple computing layers 

The challenge is straightforward to describe but difficult to solve. A modern quantum computer depends on many kinds of classical computation happening simultaneously. Some operations need to happen in hundreds of nanoseconds. Others can take microseconds. More computationally intensive workloads, such as image processing for atoms readout, optimization algorithms, and quantum error correction decoding, may require milliseconds of processing time and significant compute resources. 

Trying to force all of these workloads onto a single computing architecture simply doesn’t work. That’s why, at Quantum Machines, we’ve built our Hybrid Control architecture around the idea that different workloads belong in different places. 

 When you run an application on a modern classical computer, different parts of the workload may run on different resources, from CPUs and GPUs to accelerators and other specialized hardware. What matters is that each part of the program runs on the resource best suited to it, while the whole application is orchestrated as one system. Quantum computing is moving in the same direction. Some operations require deterministic, real-time execution inside the quantum control loop itself. Others benefit from acceleration on specialized hardware. Still others belong on high-performance servers. The key is ensuring that all these resources operate as a single system. 

That may sound obvious, but achieving it is far from trivial. Quantum systems are uniquely demanding. They combine some of the most stringent timing requirements in computing with increasingly complex classical workloads. The challenge is not simply one of performance; it is one of orchestration. Data must move seamlessly between different layers of the stack while preserving deterministic timing and ensuring that the quantum processor receives the information it needs exactly when it needs it. 

How AMD fits into the Quantum Machines architecture 

We use AMD technologies across multiple layers of the control stack to make that possible. Inside the quantum runtime, AMD adaptive computing devices support real-time pulse processing and feedback operations where latency requirements are measured in nanoseconds. These devices help power the fast classical computations required to support adaptive experiments, active reset protocols, calibration routines, and other low-latency workflows. 

 For larger-scale hybrid workflows, our Open Acceleration Stack and OPNIC create a deterministic, low-latency interconnect between the quantum controller and external compute resources. Unlike a conventional Ethernet connection, OPNIC is a PCIe-based optical interconnect designed for time-deterministic communication, ensuring that computation performed on external accelerators can be synchronized precisely with quantum operations. 

This enables workloads such as quantum error correction decoding, optimization routines, ML-based classical processing, and application-specific acceleration to run on AMD Versal™ devices while maintaining deterministic communication with the quantum controller. AMD EPYC™ processors provide the server-side compute infrastructure needed for more complex classical processing tasks. Together, these technologies support workloads across timescales ranging from hundreds of nanoseconds to milliseconds. 

 What makes this particularly powerful is that users can map workloads to the computing resources best suited to their timing and performance requirements, while relying on a unified programming and execution environment to coordinate them. Or, as I often describe it: hybrid control is about using the right classical computing resource for the job—whether that’s inside the controller, on an accelerator, or on a server—and making those results available to the quantum controller in real time. 

Beyond control: enabling quantum error correction 

 Fault-tolerant quantum computing requires sustaining gate fidelity over trillions of operations while continuously detecting, decoding, and correcting errors in real time. Achieving that level of performance demands an enormous amount of classical computation operating alongside the quantum processor. One of the most important applications of Hybrid Control is making that possible through quantum error correction. 

 Error correction is often discussed in terms of logical qubits and code distances, but underneath those concepts sit a massive classical computing challenge. Every round of error correction generates information that must be processed, decoded, and translated into corrective actions. Those computations need to happen quickly enough that errors can be corrected before they accumulate. 

As systems scale, the volume of data and the amount of processing required increase dramatically. A future fault-tolerant quantum computer may require thousands of physical qubits working together to create a single reliable, logical qubit. Coordinating those systems will require sophisticated classical computing operating alongside the quantum processor. 

This is one reason why, in my perspective, quantum computing should be viewed not as a standalone technology but as a hybrid computational system. The quantum processor and the classical infrastructure are partners in the computation, each contributing what they do best. 

Looking ahead 

I’m excited about the day when we have 10,000 or even 100,000 qubits to control. Getting there will require advances across the entire stack, not just in quantum hardware but also in the classical computing infrastructure that supports it. 

The next generation of quantum computers will depend on powerful heterogeneous computing architectures capable of handling a wide variety of workloads across multiple timescales. Some computations will happen inside the controller. Others will run on external classical compute resources, including accelerators and high-performance servers.  The challenge is making them work together as one coherent system. 

That’s why collaborations such as the one between Quantum Machines and AMD matter. Building scalable quantum computers isn’t only about creating better qubits. It’s also about building the computing architecture needed to control, orchestrate, and scale them.