Control-Plane Openness in Near-Term Quantum Computing: A Survey of Vendor Stacks and Field Implications

This paper surveys thirteen commercial quantum computing vendors to document the growing bifurcation in control-plane openness, highlighting how major superconducting platforms are restricting pulse-level access while other modalities remain open, and analyzes the resulting implications for reproducibility, hardware-aware research, and cross-vendor benchmarking.

The Gist

The Big Picture: The "Control Panel" of Quantum Computers

Imagine a quantum computer not as a magical black box, but as a high-tech car.

• The Engine (Hardware): This is the actual quantum machine (the qubits, lasers, or chips).
• The Dashboard (Gate Level): This is the standard interface where most drivers go. You press "Start," "Turn Left," or "Accelerate." In quantum terms, this is where you tell the computer to run a standard "CNOT" or "Hadamard" gate.
• The Control Plane (The Paper's Focus): This is the layer between the dashboard and the engine. It's the mechanic's panel where you can tweak the fuel mixture, adjust the spark timing, or manually override the transmission. In quantum terms, this is pulse-level control—the ability to shape the exact radio waves or laser pulses that make the qubits work.

The Main Problem:
The paper argues that access to this "mechanic's panel" is splitting into two very different groups. Some companies are locking the panel shut, while others are throwing the doors wide open.

---

The Great Split: Who is Locking the Doors?

The author surveyed 13 different quantum computing companies and found a clear divide:

1. The "Do Not Touch" Giants (Closed):

   • The Big Players: The largest companies, like IBM and Google, have decided to close this layer.
   • The IBM Event: The paper highlights a specific moment: In February 2025, IBM removed the ability for the public to control the raw pulses on their machines. Before this, researchers could tweak the engine; now, they can only press the buttons on the dashboard.
   • The Result: If a scientist published a paper in 2024 using IBM's "tweak" features, they can no longer repeat that experiment on IBM's current machines. It's like writing a recipe that requires a specific spice, but the store has decided to stop selling that spice to the public.

2. The "Open Workshop" Midsized Players (Open):

   • The Counter-Group: Smaller or mid-sized companies like Rigetti, IQM, and Pasqal (who use neutral atoms) are doing the opposite. They are keeping their "mechanic's panels" open.
   • The Result: Researchers can still see the raw data, tweak the pulses, and understand exactly how the machine is behaving.

3. The "Stuck in the Middle" Group:

   • Trapped Ions (IonQ, Quantinuum): These companies sit in a stable middle ground. They let you use the dashboard but won't let you touch the engine. They haven't closed the door, but they never opened it fully to begin with.
   • Photonic (Light-based) Computers: This group is a mix. Some (like Xanadu) are very open; others (like PsiQuantum) are completely closed because they are building machines for a future where errors are fixed automatically, and they don't want the public messing with the prototype.

---

Why Does This Matter? (The Three Harms)

The paper argues that closing this "Control Plane" causes three specific problems for science:

1. The "Lost Recipe" Problem (Reproducibility)

• Analogy: Imagine a chef publishes a famous dish. A year later, the restaurant changes the recipe and locks the kitchen. Now, no one can cook that dish again to prove it tastes good.
• Reality: If IBM closes the door, any scientific paper written before 2025 that relied on custom pulse tweaks cannot be repeated. Science relies on being able to repeat experiments; if you can't, the results become unverified.

2. The "Blind Researcher" Problem (Hardware-Aware Research)

• Analogy: Imagine you are trying to fix a car, but you are only allowed to press the gas pedal. You can't see the engine, so you can't figure out why the car is making a weird noise.
• Reality: Scientists want to study how the hardware behaves (like how it drifts over time or how to fix errors). If they can't see the raw pulses, they can't do deep research. They are forced to guess rather than know.

3. The "Apples to Oranges" Problem (Benchmarking)

• Analogy: Imagine trying to compare two cars. One lets you measure the engine temperature; the other doesn't. You can't fairly say which engine is better because you don't have the same data for both.
• Reality: You can't fairly compare a machine that lets you tweak pulses with one that doesn't. The comparison has to happen at a high level (the dashboard), which hides the real differences in performance.

---

The Big Surprise: It's Not About the "Type" of Car

A common assumption might be: "Maybe superconducting computers are hard to open, but atom computers are easy to open."

The paper says: No.

• Superconducting: IBM (Closed) vs. Rigetti/IQM (Open).
• Neutral Atoms: Atom Computing (Closed) vs. Pasqal/QuEra (Open).
• Photonic: PsiQuantum (Closed) vs. Xanadu (Open).

The Real Reason:
The difference isn't the physics (the type of car); it's the business strategy.

• Companies with huge public cloud user bases (like IBM) tend to close the doors to simplify things for the average user.
• Companies that started with a research focus or are smaller tend to keep the doors open to attract scientists.
• It's a choice, not a technical necessity.

---

What Would a "Minimum Open" World Look Like?

The paper doesn't propose a new product. Instead, it sketches a "floor" of basic rights that researchers should expect, similar to how a car owner should always be able to see the engine oil level:

1. A Documented Interface: You should be able to talk to the machine in a stable way (like a manual) so your experiments don't break next year.
2. Published Data: The machine should tell you its "health stats" (calibration data) in a format you can read and save.
3. Transparency: You should know how the software talks to the hardware, even if you don't own the hardware.
4. Proof of History: When you get a result, the machine should tell you exactly which version of the software and hardware produced it, so you can trust the data.

The Bottom Line

The paper is a warning and a map. It warns that the biggest players are locking away the "mechanic's panel," which hurts science's ability to repeat experiments and learn. But it also maps out that many other players are keeping those doors open.

The author concludes that openness is a business choice, not a technical limit. The field needs to decide if the largest platforms are willing to keep the "mechanic's panel" accessible, or if they are willing to accept that science will become harder to verify on their machines.

Technical Summary

Technical Summary: Control-Plane Openness in Near-Term Quantum Computing

Problem Statement
The paper identifies a critical bifurcation in public access to the "control plane" of commercial quantum computing platforms. Defined as the layer between gate-level circuit specification and physical control electronics (encompassing pulse synthesis, calibration loading, scheduling, and feedback dispatch), this layer is essential for reproducibility, hardware-aware research, and cross-vendor benchmarking. The author documents a trend where the largest superconducting cloud platforms (notably IBM, following the February 2025 removal of pulse-level control from all production QPUs) have closed access at this layer. Conversely, mid-tier superconducting vendors and several neutral-atom and photonic platforms have moved toward or maintained open access. This divergence creates three specific harms:

1. Reproducibility: Published results relying on pulse-level interfaces (e.g., pre-2025 IBM Qiskit Pulse literature) can no longer be re-executed on the original hardware.
2. Hardware-Aware Research: Experiments requiring real-time feedback, novel pulse decompositions, or dynamical decoupling are impossible on platforms that do not expose raw waveform synthesis or calibration data.
3. Cross-Vendor Benchmarking: Identical benchmark protocols cannot be executed across platforms due to incompatible or absent calibration metadata and differing pulse-level abstractions.

Methodology
The paper employs a structured survey of thirteen commercial vendors across four hardware modalities: superconducting transmon, trapped ion, neutral atom, and photonic. The analysis is based strictly on public documentation accessed on or before May 1, 2026. The author grades each vendor on a six-axis rubric defining the "control plane":

1. SDK Pulse-Level Access: Ranging from raw waveform synthesis (Open) to native-gate-only or closed.
2. Calibration Data Publication: Ranging from documented, retroactive archives (Open) to no publication (Closed).
3. Control Electronics Interface: Ranging from documented schemas allowing third-party substitution (Open) to no public information (Closed).
4. Queue Model: Ranging from dedicated reservations with real-time guarantees (Open) to shared queues with no guarantees (Closed).
5. Reproducibility Guarantees: Ranging from versioned calibration and provenance metadata (Open) to no commitments (Closed).
6. SDK License: Ranging from permissive open-source licenses (Open) to proprietary or non-existent (Closed).

The resulting data is presented as a machine-readable catalog (available as a separate Zenodo artifact) and a printed table, accompanied by per-row commentary that contextualizes the grades within the broader industry trajectory.

Key Contributions

• The Control-Plane Catalog: A comprehensive, citable artifact grading thirteen vendors on six specific axes of openness. This serves as a baseline for the field to track access landscapes.
• Definition of the Control Plane: The paper explicitly defines the "control plane" as a distinct architectural layer, borrowing terminology from networking to clarify the boundary between gate-level circuits and physical electronics.
• Refutation of Modality Determinism: The survey demonstrates that hardware modality (e.g., superconducting vs. neutral atom) does not predict openness. Instead, openness is determined by commercial position, the size of the cloud user base, and the historical inflection point of the vendor's stack design.
• Counterargument Analysis: The paper systematically addresses eight common defenses for closed access (e.g., safety risks, calibration IP value, premature standardization) and argues that these concerns justify tiered access or safety-aware abstractions, not the total closure of the control plane.

Results
The catalog reveals distinct trajectories across the industry:

• Closing Trajectory: The largest cloud platforms (IBM, Google Quantum AI) and stealth fault-tolerant programs (PsiQuantum) are closed at the pulse level. IBM's removal of Qiskit Pulse is identified as the most consequential event in the review period.
• Open Trajectory: Mid-tier superconducting vendors (IQM, Rigetti, OQC) and specific neutral-atom (Pasqal, QuEra) and photonic (Xanadu) vendors maintain open pulse-level interfaces.
• Stable Equilibrium: Trapped-ion vendors (IonQ, Quantinuum) have settled on native-gate-only access, neither closing further nor opening to pulse-level control.
• Bimodal Neutral Atoms: The neutral-atom modality shows a split between open-by-design SDKs (Pasqal, QuEra) and closed-via-cloud-abstraction models (Atom Computing via Azure).
• Photonic Divergence: The photonic modality exhibits the widest variance, with open (Xanadu), closed (PsiQuantum), and mid-tier/limited (ORCA) postures coexisting.

Significance and Claims
The paper modestly claims to describe the current state of the field rather than proposing a new architecture or reference implementation. Its primary significance lies in documenting the "loss" of access as the landscape shifts and defining what "minimally open access" would look like. The author argues that the field has lost the ability to reproduce a non-trivial fraction of recent literature and that hardware-aware research is increasingly gated by vendor choice.

The paper concludes that the asymmetry of the closing trajectory is driven by commercial strategy and scale, not technical necessity. It posits that a "floor" of minimally open access—consisting of open pulse APIs, published calibration schemas, documented control electronics interfaces, and reproducibility guarantees—is technically feasible and already met by several vendors. The paper invites vendors to justify their interface choices against this catalog, funders to use it as a reference for grant conditions, and researchers to cite it when describing hardware access models. The work is framed as a living document intended to be forked, updated, and corrected by the community.