Why a new kind of compute?
The world's hardest problems, routing millions of deliveries, scheduling networks, optimising chips, training energy-based AI, are probabilistic by nature.
Deterministic processing units (such as CPUs, GPUs, TPUs) need to emulate randomness with pseudo-random number generators, incurring latency, energy cost and scaling limits, especially as models and datasets grow.
Purpose-built probabilistic hardware removes that bottleneck.
What is probabilistic computing?
In probabilistic computing, the fundamental unit is the p-bit: a bit that fluctuates between 0 and 1 with a controllable probability. By wiring many p-bits into a network and tuning their interactions, the system naturally explores and settles into low-energy configurations that correspond to high-quality solutions for well-known mathematical problems.
QD's approach
QD's ORBIT architecture combines ultra-fast photonic p-bits with electronic control to deliver a scalable, room-temperature probabilistic computer.
At its core is our patented DISC protocol, which continuously monitors the quantum source of randomness to verify quality and stabilise computation. The result is precise, reproducible sampling—critical for rapid convergence in probabilistic algorithms.
Advantages
What problems does it solve?
Combinatorial optimisation | Faster, lower-energy solutions to NP-hard problems such as MAX-CUT, graph colouring, routing/scheduling, and portfolio selection. |
Probabilistic & energy-based AI | Native support for MCMC and energy-based learning enables training that is difficult or inefficient on deterministic hardware. |
Quantum-inspired simulation | Accelerates sampling-heavy simulators and lattice-based optimisation at room temperature. |
How is QD different?
Photonic p-bits for speed and efficiency
Leveraging the quantum nature of light yields extremely fast, low-energy state flips. This is ideal for large-scale probabilistic sampling.
Quantum-verified randomness (DISC)
Continuous, on-device verification of entropy quality improves accuracy and reproducibility of results versus PRNG-driven approaches.
Deployable today
Standard 2U form factor, Ethernet control, and no special cooling. Built for both lab pilots and data-centre integration.
Credible, science-backed roadmap
White papers, customer presentations and competitive analyses detail performance metrics and a clear path from prototype to scaled products.