Question 1

What is PRISM?

Accepted Answer

PRISM is a GPU-native portfolio optimization engine that achieves 10–21× speedup vs Gurobi on constrained quadratic programs with real US equity data. It's designed for direct indexing, tax-loss harvesting, and institutional rebalancing at up to 100,000-asset scale.

Question 2

How fast is PRISM compared to other solvers?

Accepted Answer

PRISM achieves 10–21× speedup vs Gurobi 13.0.1 and 4–10× vs OSQP on N=5,000 constrained portfolio optimization with real market data (Feb 2026). Median kernel compute is 4.41ms. At N=100,000 assets, GPU latency is 766ms.

Question 3

What constraints does PRISM support?

Accepted Answer

PRISM enforces wash-sale rules, lot-level tax accounting, turnover limits, tracking error bounds, liquidity constraints, cardinality limits, sector and ESG exposure limits, and multi-account batch optimization.

Question 4

What asset scale can PRISM handle?

Accepted Answer

PRISM has been verified at 100,000-asset scale with real US equity data (Feb 2026). GPU latency remains sub-second at all tested scales: 757ms at N=20K, 671ms at N=50K, and 766ms at N=100K.

Question 5

How is PRISM deployed?

Accepted Answer

PRISM is available as an HTTPS API deployed on dedicated cloud infrastructure or within your VPC. All solves produce deterministic results with replayable audit logs.

PRISM — GPU-Native Portfolio Optimization Engine | Asymmetry Computing