Probabilistic computing is about more than computing with randomness.
The computing industry has settled on the term probabilistic computing for a collection of new technologies, applicable across every corner of computing, which either use insights from probability theory or harness properties of materials and physics, randomness among them, to solve general-purpose problems more efficiently. The range runs from pricing and risk in quantitative finance through design margins in EDA and semiconductors, state estimation in robotics, trustworthy and efficient AI, and supply chains, to pharmaceutical R&D. This page sizes that market bottom-up, from named institutions and what their workloads cost today. No analyst multiple appears anywhere in the arithmetic. Published by Signaloid.
The baseline is always conventional Monte Carlo on CPUs and GPUs. Every figure on this page is what that incumbent costs the counted institutions today, so deselecting technologies never changes the market, only which approaches address it.
Use caseOne use case at a time. Only use cases with a sourced entity count carry a figure. The rest state what evidence is still missing.
Pills name the technologies with recorded evidence for the selected use case, colored by technology and always labeled. The grades follow the published rubric: bold on a tint is measured, a recorded measurement with a stated baseline. Plain is demonstrated, production or public demonstration with a source. Dashed is claimed, a vendor or press assertion. No pill means no recorded evidence. Each pill names its recorded source in its tooltip.
Vendor capture assumption, %What share of the addressable spend suppliers of these technologies could capture. The assumption is yours to make, and the Vendor Opportunity figure stays empty until you set it.
Buy or rent todayIBM Quantum
Quantum
Gate-model systems, rentable by the shot today through IBM Quantum Platform and the cloud marketplaces.
Buy or rent todayIonQ
Quantum
Trapped-ion gate-model systems, rentable today through the major cloud providers.
Buy or rent todayD-Wave
Quantum
Annealing systems and the Leap quantum cloud, rentable today, with systems also sold for on-premises deployment. Publicly targets scheduling, vehicle routing, resource optimization, cargo loading, materials science, drug discovery, AI, and financial modeling.
Buy todaySignaloid Cloud
Distribution arithmetic
Turnkey UxHw compute by Enterprise tier. Product page.
UxHw compute modules for the edge, shipping today. Product page and Mouser.
Appears availableQuantum Dice QRNG
RNG hardware
Self-certifying quantum entropy sources for Monte Carlo and cryptography.
Open sourceStan and PyMC
Probabilistic software
Open-source probabilistic programming on the hardware you already run.
AnnouncedSqortex p-bit architecture
Probabilistic bits
Stochastic nanodevice hardware for optimization workloads.
AnnouncedLudwig Computing
Probabilistic bits
Probabilistic hardware for energy-efficient AI, named for Boltzmann.
AnnouncedExtropic TSU
Thermodynamic and analog
A thermodynamic sampling unit for probabilistic AI workloads.
AnnouncedNormal Computing CN101
Thermodynamic and analog
A thermodynamic computing chip taped out for AI and HPC data centers.
AnnouncedUnconventional AI
Thermodynamic and analog
Analog and mixed-signal chips storing probability distributions in the physical substrate.
Buy today and buy or rent today mean a purchase or rental route verified against the vendor\u2019s live pages: the Signaloid product pages, Mouser, the AWS Marketplace listing, and the quantum cloud services. Appears available means the vendor says so and Signaloid has not yet verified a route. Greyed cards are announced, not yet purchasable. Named vendors, with no endorsement either way.
Cite this as
How this calculator works: Addressable Annual Spend counts what named institutions pay today for the selected use case, multiplying sourced institution counts by per-entity spend estimates. Vendor Opportunity multiplies that spend by a capture share you enter under Use Cases, so it applies only where a spend figure exists, today Value at Risk and the all-use-cases floor. Use cases without a sourced count show their recorded evidence instead of a figure.