Dynex Quantum Platform: A cloud-based, qubit-agnostic platform

Dynex provides a qubit-agnostic compute environment that integrates neuromorphic p-qubit hardware, large-scale algorithmic-qubit emulators, and third-party quantum processing units into a single, unified execution layer. The platform abstracts device-specific characteristics and presents a consistent programming and operational model for optimization, simulation, and quantum-inspired workloads.

The system is designed to support heterogeneous quantum and classical devices with differing qubit modalities, topologies, noise characteristics, and control interfaces, while maintaining a coherent computational workflow for the end user.

1. Architecture Overview

1.1 Unified Compute Layer

All supported devices—Dynex Apollo, Dynex GPU/CPU qNodes, and external QPUs (IBM, IonQ, Rigetti, D-Wave, QuEra, IQM, etc.)—are represented as Dynex Quantum Nodes.
Each node operates under the coordination of the Dynex Engine, which manages:

workload routing,
resource scheduling,
cross-node synchronization, and
result aggregation.

This ensures consistent execution semantics irrespective of the backend’s underlying qubit technology or computational paradigm.

1.2 Hybrid Execution Model

The platform supports mixed-modality computation. Workflows may combine:

continuous-time annealing (Apollo),
large-scale algorithmic-qubit simulation (GPU/CPU), and
gate- or annealing-based quantum operations (third-party QPUs).

The execution planner determines backend selection based on:

problem graph density,
latency tolerance,
required connectivity,
device-specific constraints,
and operational availability.

2. Apollo Integration

The Apollo p-qubit chip is a primary compute node in the Dynex architecture. It implements a 10,000-element stochastic computing fabric based on quantum driven probabilistic bits (p-qubits) operating in continuous time. The chip is fabricated in 16 nm mixed-signal CMOS and operates entirely at room temperature.

> Watch the Apollo video
> Download Apollo Datasheet (PDF)
> Scientific Publications

2.1 Fundamental Characteristics

p-Qubit Fabric: 10,000 parallel stochastic units
Connectivity: Δ256 Hyperion topology (up to 256 couplings per node)
State Dynamics: continuous-time stochastic switching governed by local fields
Entropy: one Integrated Quantum Entropy Unit (IQEU) per p-qubit
Performance:
- ~10⁸ flips/s per p-qubit
- ≤10 fJ energy per flip
- ~0.5 W total power consumption
Operating Environment: 0–85°C, no cryogenics or laser control required
Control Interface: tightly coupled Dynex Control Unit (DCU) for bias/coupling scheduling

2.2 Technical Role in the Platform

Apollo provides native support for:

Ising/QUBO minimization via thermodynamic relaxation
Boltzmann sampling
generative probabilistic modeling
analog vector–matrix multiplication for inference tasks

Its Δ256 topology significantly reduces embedding overhead compared to low-degree architectures (e.g., Pegasus Δ=15, Zephyr Δ=20). This enables direct execution of dense industrial problem graphs without decomposition into multiple virtual qubits.

3. Third-Party QPU Integration

Dynex integrates a broad range of quantum devices to provide coverage across several computing models. Examples include:

Provider	Device	Technology
IBM	Eagle, others	Superconducting transmon (gate model)
IonQ	Aria, Forte	Trapped-ion (gate model)
Rigetti	Ankaa series	Superconducting (gate model)
D-Wave	Advantage / Advantage2	Quantum annealing
QuEra	Aquila	Neutral-atom / Rydberg analog simulation
IQM	Garnet, Emerald	Superconducting

The Dynex Engine abstracts away device-specific APIs, differing coherence properties, and topology constraints. This allows external QPUs to serve as complementary resources for circuit-based workflows, verification tasks, or algorithmic benchmarking.

4. Dynex GPU/CPU qNodes (Algorithmic Qubits)

The Dynex platform includes high-performance algorithmic-qubit simulators implemented on GPU and CPU clusters. These nodes provide:

up to 1 million algorithmic qubits,
deterministic reproducibility,
flexible embedding,
and compatibility with the same SDK used for Apollo and QPUs.

GPU qNodes are particularly suited for:

large-scale sweeps,
embedding validation,
debugging of Hamiltonian structures,
and scenarios where massive problem sizes exceed practical physical qubit counts.

> Benchmarks