Tensor network models

Quantum-inspired AI
for the optimal decision under uncertainty.

Today's AI predicts the future. Intrica computes the best move within it — in the live markets of finance and energy, orders of magnitude faster than simulation, on classical hardware.

Mission

We're building the engine that makes the world's hardest problems computable.

Every system that runs the world is governed by uncertainty too vast for any computer to resolve — so for a century, we approximated, simulated, and hoped. Intrica changes what is computable: a single engine, derived from physics, that doesn't just predict the future but computes the optimal decision within it. Starting in the markets that move the world's capital and energy — pricing derivatives, hedging books, dispatching power — problems once beyond reach become ones we can finally compute.

01The frontier

Two waves of prediction. A third that decides.

The two waves that remade AI share one engine — prediction — first of language, then of the physical world. The next is different in kind: computing the optimal action under uncertainty.

Wave · language

LLMs

Predict the next token, trained on the world's text.

Wave · perception

World models

Predict the next state of the world, from video and simulation.

Open frontier · decision

Intrica

Decide the optimal action — weighing the full distribution of futures, not a sample of it.

The space of possible futures explodes exponentially. Predicting it is not the same as knowing what to do — that is the gap.

02Use Cases

The first markets.

One engine for wherever high-dimensional uncertainty meets real money — where the gap between a defensible answer and the optimal one is measured in capital, risk, and P&L. Benchmarked in finance.

Energy & commodities

Dispatch, battery energy storage, and trading across volatile markets — where the optimal call, made in time, is the difference between capturing a price swing and being caught by it.

Pricing derivatives under correlated uncertainty is the hardest decision problem in finance — and the canonical stress test for high-dimensional compute, where Monte Carlo has set the standard for twenty years.

Monte Carlo · incumbent
Sampling

Draws millions of random scenarios and averages them into an estimate — accuracy improving only as 1/√N, every run seed-dependent.

Intrica
Compression

Holds the full range of outcomes in one structured object — solved once, deterministically: the same inputs always return the same answer.

Our results
~1,000×
Faster than
Monte Carlo
≤ 0.1%
Pricing
error
100%
Deterministic,
auditable
10⁻² 10⁻³ 10⁻⁴ 10⁻⁵ 10⁻⁶ 10⁻⁷ 10⁻⁸ 10⁻⁹ 10⁻³ 10⁻² 10⁻¹ 1 10¹ 10² 10³ RELATIVE ACCURACY RUNTIME (SECONDS) ×1,000 FASTER Intrica Monte Carlo
fig. 01 — runtime vs relative accuracy, log–log · 5 correlated underlyings · ~1,000× faster than Monte Carlo
03The technology

A new primitive, built on two revolutions.

Intrica’s engine fuses two breakthroughs: modern AI, which learns the structure of messy real-world systems no one could specify by hand, and the mathematics of quantum physics, which computes over state spaces too vast to enumerate. The result is tensor network models.

Modern AI
Neural nets · Transformers · LLMs

Compute probabilistic tokens. Trained on data.

Quantum algorithms
Optimization · Circuits

Compute exponentially large states. Derived from equations.

Tensor network models

Agile as neural networks. Rigorous and powerful as quantum physics.

04Traction

Already real. At the frontier.

Paid pilots underway in finance and energy, and direct lines into the institutions at the frontier of European quantum technologies.

Commercial
Finance · Energy

Paid proofs-of-concept underway with Tier-1 customers in both sectors.

Network
  • University of Geneva
  • Innosuisse
  • Open Quantum InstituteCERN
  • CEA GrenobleHub Quantique
05Why now

Three forces. One moment.

The transformer scaling era is ending, tensor networks have crossed from research into production, and an entire class of high-stakes decision problems is still out of reach for today's tools.

The wall

Scaling plateaued.

Frontier training runs approach $5–10B. The next 10× of AI won't come from bigger transformers — it needs new primitives.

The primitive

Tensor networks are ready.

Thirty years of physics matured the method. Turning it into a production engine for live markets is the hard part — and takes algorithms and expertise almost no one has.

The category

Decision is underbuilt.

Prediction has a hundred companies; computing the optimal decision — accurately, auditably — has almost none. Under the Fundamental Review of the Trading Book (FRTB), market-risk capital at top US banks rises $100B → $138B.

Compute. Predict. Decide.

Work with us

The decision layer of the AI stack.

PilotsFinance and energy teams with a hard pricing, risk, or dispatch problem.
ResearchGroups working on tensor networks, or optimal decision under uncertainty.
InvestorsEarly-stage inquiries.
hello@intricacomputing.com

Built in Geneva, Switzerland.