On April 23rd, 2026, Riverlane hosted a deep-dive webinar into our Quantum Error Correction (QEC) Technology Roadmap, marking a pivotal moment in our mission to unlock quantum computing’s promise sooner.
Led by Neil Gillespie, our VP of Applied Research, the session explored how a staged engineering path focused on error correction can accelerate the arrival of utility-scale quantum computing by an estimated three to five years.
We’d like to thank our panellists for their time and expert insights: Frank Deppe from IQM, Masoud Mohseni from HPE, Krysta Svore from NVIDIA, Tobias Thiele from Zurich Instruments, and Ben Yuen from Infleqtion.
The launch of our roadmap and its corresponding whitepaper sparked a conversation across the industry, raising fundamental questions about how we transition from small-scale quantum systems to reliable, utility-scale machines.
This post captures the top insights and answers to the most pressing questions raised during the roadmap launch and the subsequent webinar session. These include:
- How can a roadmap for Quantum Error Correction speed up utility-scale quantum computing?
- How and when will quantum computers do something useful?
- What’s a quantum gate? What’s the difference between a Clifford and non-Clifford gate?
- What is a QuOp?
- Why do we need "real time" quantum error correction?
- What is a quantum decoder?
- How does quantum software work?
- How important are qLDPC codes?
- Why do quantum computers need control systems?
- How will AI and HPC work with quantum?
- How do we integrate HPC, AI and quantum systems?
- What is an AI pre-decoder?
- How can AI help quantum computers?
- What challenges are quantum computing companies currently prioritising?
- How much will it cost to build a utility-scale quantum computing?
- What’s the difference between horizontal and vertical integration in quantum computing?
- Does the quantum computing industry need better standards?
- What does the future look like for quantum?
Key takeaways: The future of Quantum Error Correction
- Accelerated timelines: The Riverlane QEC Technology Roadmap identifies an engineering path that could bring utility-scale quantum computing to market 3 to 5 years sooner. Riverlane’s Deltaflow system and Local Clustering Decoder (LCD) reduce physical qubit overheads by 75% to help achieve this.
- The "MegaQuOp" milestone: The industry is converging on 2028 as the pivotal year for the "MegaQuOp", which is the point where a quantum computer can execute one million error-free operations to solve intractable classical problems.
- Real-time QEC is essential: To be useful, quantum computers must process errors in "real-time" during the computation, rather than after. This real-time QEC must be done with speed and accuracy.
- Hybrid supercomputing (HPC + AI + QPU): The future of quantum is hybrid. AI and High-Performance Computing (HPC) will work alongside Quantum Processing Units (QPUs) to manage real-time decoding and workload allocation, for example.
- Industry shift to horizontal integration: Building a utility-scale machine is highly complex and costly. The industry is pivoting toward a horizontal model, using standardised interfaces (like QECi) to combine specialised components from different vendors.
Section 1: The Riverlane Technology Roadmap
1. How can a roadmap for Quantum Error Correction speed up utility-scale quantum computing?
Building a quantum computer is more than just a challenge of physics; it’s a challenge of speed and precision. Physical qubits are noisy, fragile, and prone to errors. Without a robust QEC system, those errors cascade, rendering any computation useless. In other words, utility-scale quantum computing needs QEC. You can read more about this here.
Now, let’s focus on how The Riverlane Technology Roadmap helps with that acceleration.
In the quantum industry, roadmaps are common, but they often focus on qubit counts or physical hardware milestones that feel like a distant horizon. Many roadmaps, for example, track the transition from 10 to 100 to 1,000 physical qubits or look at logical qubit numbers. However, these metrics don't fully consider the critical noise problem that prevents those qubits from doing meaningful work.
Figure 1. Unreliable physical qubits are turned into a useful, logical qubit.
The Riverlane Technology Roadmap is different because it is a clear, staged engineering path designed to solve the most significant barrier in the field: Quantum Error Correction (QEC).
As Neil Gillespie, our VP of Applied Research, noted during our recent webinar, Riverlane doesn't view the future as a vague target. Instead, we have a concrete plan to "make quantum computers useful sooner" by mastering QEC.
The 3–5-year acceleration figure is rooted in Riverlane’s development of the Local Clustering Decoder (LCD), a hardware-based QEC solution introduced in a December 2025 paper in Nature Communications.
In a leakage-dominated noise model, Riverlane demonstrated that our LCD can achieve one million error-free quantum operations while using 75% fewer physical qubits than a non-leakage-aware decoder.
This breakthrough allows quantum computers to achieve the same computational results with significantly less hardware overhead, effectively bypassing several years of traditional physical qubit scaling requirements.
2. How and when will quantum computers do something useful?
The world’s quantum roadmaps, generally, converge around 2028 as the date for when quantum computers will solve a real-world problem that’s intractable on a classical machine. This point is often called the ‘MegaQuOp’, when quantum computers can execute one million (Mega) error-free Quantum (Qu) Operations (Op).
Riverlane is already executing a plan that moves quantum computing out of the “lab science” phase and into reliable, scalable engineering. Our roadmap is defined by specific "eras," each representing a defined capability step:
- 2025: The Foundation of Memory. Our first milestone was the mastery of "auantum memory". This is the fundamental ability to maintain a long-lived logical qubit. This proved we could preserve quantum information. Now, we are learning to compute with it.
- 2026: The Era of Logic. We are currently moving into the era of "logical gates." This marks the transition to genuine quantum computation using Clifford logic.
- 2028–2032: The Leap to Non-Clifford Logic. Looking further out, we focus on the first QEC systems that meaningfully deliver Non-Clifford logic. This is the "holy grail" of the roadmap, as these operations are required for the algorithms that will finally outperform classical computers on problems that truly matter.
You can read more about our roadmap and each of the eras in our whitepaper: The Riverlane QEC Technology Roadmap.
Figure 2: The Riverlane Technology Roadmap.
3. What’s a quantum gate? What’s the difference between a Clifford and non-Clifford gate?
In quantum computing, gates are the elementary units of operation used to manipulate qubits to perform calculations. The roadmap describes them as the building blocks used to perform algorithms. When these gates are protected by error correction, they are referred to as logical gates or error-corrected logical gates.
Clifford gates are the fundamental operations used in early fault-tolerant systems and are crucial for many Quantum Error Correction (QEC) codes.
- Examples: These include the Hadamard (which creates superposition), Phase (S), and CNOT gates.
- Classical simulation: A key characteristic of Clifford gates is that they are a specific set of operations that can be simulated classically.
- Role in QEC: They are indispensable for manipulating qubits and performing error correction. Because their errors can be tracked and reinterpreted at the end of a circuit, it is often sufficient to decode them in "post-processing" after the circuit has been executed.
The transition to non-Clifford gates represents a "Logic Leap," as these gates are required to reach universal quantum computation.
- Examples: A commonly referenced non-Clifford operation is the T-gate.
- Quantum advantage: Unlike Clifford gates, non-Clifford gates introduce the true power of quantum computation. By combining Clifford gates with at least one non-Clifford gate, a quantum computer gains universality, enabling it to execute complex algorithms (like Shor’s) that offer exponential speedups over classical methods.
- Implementation challenge: Non-Clifford gates are much harder to implement fault-tolerantly. Some require techniques such as logical branching, for example, where a logical operation conditional on a corrected measurement.
- Real-time requirement: Because the computer must decide whether to apply a subsequent gate based on a previous measurement, decoding for non-Clifford gates must happen in real-time during circuit execution, rather than after.
Figure 3: Using Clifford gates (left) is like being able to navigate to six specific points on a globe, where your movement is limited to those locations. However, non-Clifford gates (right) allow you to explore the entire globe, giving you the freedom to reach any point.
4. What is a QuOp?
A QuOp is simply one reliable Quantum (Qu) Operation (Op). By reliable, we mean error corrected. In other words, a QuOp is the fundamental unit of useful computational work.
We introduced this new metric to the industry back in 2023, and we still use it today to ensure this roadmap remains grounded.
By focusing our roadmap on increasing "QuOps per second," we shift the conversation away from how many qubits a machine has, and toward how much useful work that machine can actually perform.
You can think of a QuOp like a clock cycle in a classical processor: that fundamental unit of useful computational work. More QuOps per second means a more powerful machine. But the cost of a QuOp isn’t the same for every quantum technology. It depends on the underlying architecture.
For systems based on fixed qubits using the surface code, a single QuOp corresponds to roughly ‘D’ rounds of QEC cycles on a logical qubit, where D is the code distance.
Figure 4: Schematic to demonstrate the notion of a QuOp using a circuit made up of ‘N’ logical qubits with a depth of logical operations ‘D’. The number of QuOps is N.D.
For architectures using transversal logic, it can be as short as a single QEC cycle with a potential speed advantage there. Some systems may combine both: transversal gates where they can and lattice surgery where they must.
So, the QuOp is a nuanced metric, but that’s precisely why we find it useful. It forces an honest comparison across different hardware platforms and architectures. When you see the QuOp numbers on the roadmap, you’re seeing our best assessment of real computational power.
Section 2: General questions on Quantum Error Correction
5. Why do we need "real time" quantum error correction?
A common misconception is that error correction can be done after a calculation is finished. In reality, the only way to run a useful quantum computation is to detect and process errors faster than they accumulate while the computation is still running. This is real-time QEC.
This "real-time" requirement is the primary function of our Deltaflow QEC system: it acts as the heartbeat of the system, identifying and resolving errors in the middle of a live operation. You can read more about why we need real-time QEC here.
6. What is a quantum decoder?
Decoders are a crucial part of the error-correction process, which groups physical qubits to create more stable logical qubits, making quantum computations more reliable.
In other words, quantum decoders are classical algorithms used in quantum computing to infer and correct errors in a quantum computer's qubits. It functions by interpreting data from ‘syndrome measurements’, which are indirect checks on the qubits, to identify and fix errors that may have affected the encoded logical state.
The engineering challenge here is immense: for solid-state qubit systems, the decoder must handle extremely fast cycle times, often on the order of microseconds. This speed is critical because decoder accuracy directly translates into computational power, the more reliably you can correct errors in real time, the more QuOps your system can run. We have spent years engineering a decoding system that can keep up with these demanding hardware requirements.
However, different architectures, such as superconducting qubits or AMO (Atomic, Molecular, and Optical) qubits operate at vastly different cycle speeds and face unique connectivity challenges.
Because of this, our roadmap is designed to consider all major qubit types. By building tools that work across the entire hardware landscape, we ensure that as the hardware evolves, our error correction solutions evolve with it, supporting any leading qubit modality to reach the finish line of fault tolerance.
Figure 5: An overview of QEC in action with a Quantum Processing Unit (QPU).
7. How does quantum software work?
Quantum software enables the development, simulation and execution of algorithms on quantum computers.
In other words, quantum software is needed to make quantum computers accessible to the researchers and partners who are building the algorithms of the future. This is where Deltakit comes in.
Deltakit acts as an interface layer between the complex algorithm that end users want to execute on the quantum device and the underlying error correction machinery. By using Deltakit, researchers can work at the level of logical operations without needing to understand the physical gate-level complexities happening underneath. This allows for faster innovation and exploration of different QEC approaches.
8. How important are qLDPC codes?
qLDPC (Quantum Low-Density Parity-Check) codes are a QEC code that can encode multiple logical qubits. These codes promise to significantly reduce the overhead required for error correction, but they come with a catch: they require long-range connectivity between qubits.
This presents a real challenge for superconducting qubits. Because these are solid-state systems, they are generally limited to "nearest-neighbour" connectivity. This means a qubit can usually only interact with other qubits directly adjacent to it. To overcome this, hardware makers are investing heavily in code engineering to break the planar picture and enable the complex connections these modern codes require.
In contrast, the AMO community has a natural advantage: reconfigurable arrays that make long-range connectivity much easier. However, this flexibility introduces its own set of trade-offs. To create these connections, atoms must be physically "shuttled" or moved around the array.
This physical motion has a cost: specifically, the time required for the acceleration and deceleration of atoms. This introduces overhead into the QEC cycle and potential noise that the decoder must then solve. For companies in this space, the priority is optimising the "degree of motion" to ensure that the benefits of all-to-all connectivity aren't outweighed by the time it takes to move the qubits.
9. Why do quantum computers need control systems?
Control systems are the hardware and software components that interact with quantum devices to manipulate their behaviour and achieve desired states. These systems use classical signals, such as precisely timed microwave or radio-frequency pulses, to control the qubits in a quantum computer. They are essential for operations such as gate calibration, qubit manipulation and QEC, forming the crucial bridge between abstract quantum algorithms and physical quantum hardware.
Control electronics face challenges too because, as we scale from dozens of qubits to thousands (and eventually millions), the system must manage thousands of signal lines simultaneously.
The primary challenge here is synchronisation. We cannot compromise on timing; every pulse must arrive at the exact right moment across the entire system.
To solve this, providers like Zurich Instruments are developing specialised ASICs (an application-specific integrated circuit that’s designed to do one specific task very efficiently).
These ASICs are designed to distribute a global clock and global time across every component. Ensuring that thousands of pulses can be synthesised with high fidelity at scale is the "backbone" that will allow large-scale quantum systems to function as a single, cohesive machine.
Section 3: AI and quantum computing
10. How will AI and HPC work with quantum?
There's growing interest in integrating AI and machine learning with quantum computing and HPC, particularly for error correction and hybrid algorithms.
Our second panel featured experts from NVIDIA and HPE who shared a clear vision: the quantum processing unit (QPU) will serve as a specialised node within a much larger, heterogeneous supercomputing environment.
For example, AI can help deduce error syndromes and apply corrections in QEC, and AI's inference capabilities are well-suited for real-time decoding tasks. AI is already being used in hybrid quantum algorithms and developer tools such as compilers and code-writing assistants.
Quantum computers are also increasingly envisioned as specialised co-processors or accelerators integrated within existing HPC centres. This model is critical because to fully realise the promise of utility-scale quantum computers, seamless integration with HPC infrastructure, efficient data transfer mechanisms, and sophisticated scheduling are essential for managing hybrid classical-quantum workflows.
As such, the computationally intensive classical control and compilation elements necessary for real-time QEC will heavily leverage the capabilities of HPC resources, ensuring a synergistic relationship where quantum computing and classical supercomputing work together.
In other words, the future of quantum computing isn’t a solo act; it’s a hybrid performance. The Quantum Scaling Alliance, for example, is a consortium that’s helping to bring together stakeholders across industry, academia, startups, and national labs to drive progress in quantum.
11. How do we integrate HPC, AI and quantum systems?
To achieve real-time error correction, we cannot rely on distant cloud connections. There is a critical need for a "tighter integration loop" between the QPU and a local, real-time host, typically a GPU or CPU node.
This inner loop allows the system to perform high-speed calibration and error decoding while the computation is in progress. By integrating a real-time host directly with the quantum controller, we can achieve the low-latency feedback required to stabilise qubits during complex operations.
12. What is an AI pre-decoder?
A pre-decoder is a specialised circuit or algorithm used to simplify complex data before it reaches the main decoder, helping to reduce latency and complexity.
NVIDIA recently introduced a tool in this space: the "Ising" family of AI models. These models act as a pre-decoder that sits in front of the global QEC decoder.
By using a convolutional neural net to process initial syndrome data, this pre-decoder can "lighten the load" for the main system, according to Krysta. The results are transformative, potentially offering capabilities that are faster and more accurate than standard global decoders alone.
13. How can AI help quantum computers?
Beyond the task of pre-decoding, AI is driving what experts call "extreme co-design" across the entire stack. We are now seeing AI used for:
- Algorithm discovery: Finding new applications and proving where quantum advantage exists.
- Circuit optimisation: Using Large Language Models (LLMs) to find shorter, lower-depth quantum circuits.
- Cycle compression: Using AI to optimise syndrome extraction and compress the cycles required for quantum logic.
AI is also solving the massive challenge of workload management in a hybrid supercomputer. Not every part of a problem needs to be solved on a quantum machine; in fact, it is often more cost-effective to keep low-entanglement tasks on a classical machine.
AI can monitor "entanglement heat maps" in real time, intelligently allocating sub-problems to either the CPU, GPU, or QPU based on the specific requirements of the data. This ensures that the quantum processor is used only where it offers a truly "advantageous" application.
Section 4: The future of quantum computing and QEC
14. What challenges are quantum computing companies currently prioritising?
As we move toward scaling quantum systems, the conversation is shifting from theoretical physics to practical engineering. During our panel sessions with leaders from IQM, Infleqtion, and Zurich Instruments, it became clear that while different hardware architectures offer unique advantages, they each face specific hurdles in implementing efficient error correction.
For example, achieving a balance between exploration (and getting to utility scale as soon as possible) versus designing cost-efficient solutions was cited as an ongoing challenge.
15. How much will it cost to build a utility-scale quantum computing?
We must confront the economic reality of building a utility-scale quantum computer. As we scale from dozens of qubits to millions, the cost per qubit becomes a make-or-break factor for the industry.
Frank Deppe (IQM) noted that if we are not careful, the cost of a million-qubit system could easily spiral into the billions or even tens of billions of dollars. This is simply not a realistically scalable path.
To reach utility-scale, we must find a "compromise", bringing the cost into a "conceivable and reasonable regime" through smart engineering and shared infrastructure.
By focusing on cost-effective, modular solutions today, we ensure that the quantum computers of tomorrow are not just scientifically possible, but commercially and practically sustainable. This is where horizontal integration can help.
16. What’s the difference between horizontal and vertical integration in quantum computing?
In the early days of the field, many companies followed a "vertical integration" model, attempting to build everything from the physical qubits to the software stack in-house. However, as Masood Mohseni (HPE) highlighted, this approach is becoming "too risky" and difficult to maintain as we scale the systems.
The industry is now pivoting toward horizontal integration. This model allows companies to join forces, combining "best-in-class" components from different specialised vendors.
By moving away from proprietary, vertical silos, the entire ecosystem can accelerate R&D and reduce the risk of doubling down on individual mistakes.
17. Does the quantum computing industry need better standards?
In short: yes. For horizontal integration to work, the industry needs a common language. There is a critical and growing need for agreed-upon interfaces that allow different control systems, decoders, and qubit modalities to work together seamlessly.
Initiatives like the NVIDIA’s NVQLink or Riverlane’s QECi (QEC interface) are essential steps in this direction. By establishing these industry standards, we ensure that components from different companies are modular and interoperable.
This standardisation is especially vital given the rapid explosion of AI; well-defined interfaces allow AI agents to integrate across the stack more effectively, further accelerating innovation.
18. What does the future look like for quantum?
The panellists were in agreement: as we move toward the goal of utility scale, the way the quantum industry operates is undergoing a fundamental shift. Building a utility-scale quantum computer is no longer a task for a single company working in isolation; it is becoming a collaborative effort across a specialised ecosystem where horizontal integration is key.
To quote Neil, “at Riverlane, we’re genuinely excited about what comes next”.
In conclusion
As we move into the era of practical engineering, collaboration across the entire quantum ecosystem is the only way to overcome the final hurdles to utility-scale quantum computing.
By focusing on modularity, real-time error correction, and industry-wide standards, we can ensure that the quantum computers of tomorrow are not just scientifically possible, but commercially sustainable.
Ready to explore the technical details behind our strategy? Download the Riverlane QEC Technology Roadmap whitepaper for a deep dive into how we are accelerating the path to utility-scale quantum computing.
To hear the full discussion from our expert panels and see the roadmap in action, you can also watch the webinar on demand here.