Skip to content

OPAQUE

San Francisco confidential-AI company, spun out of UC Berkeley research, that lets enterprises run AI training and inference on sensitive data inside hardware-isolated enclaves.

Visit Website ↗ + Add to Compare Claim This Company
67/100Incremental Innovator

Overview

OPAQUE (Opaque Systems) commercializes confidential-computing research originally developed at UC Berkeley’s RISELab, letting enterprises train, fine-tune, run inference on, and execute AI agents against sensitive or regulated data while keeping that data cryptographically isolated inside hardware-backed enclaves. The platform provides attestation that proves how data and models were handled at runtime, rather than relying on policy documents alone.

Founded by Ion Stoica and Raluca Ada Popa — both established distributed-systems and security researchers, with Stoica also a co-founder of Databricks — the company closed a $24 million Series B in February 2026 led by Walden Catalyst at an approximately $300 million valuation, with returning investors Intel Capital, Race Capital, Storm Ventures, and Thomvest, bringing total funding to $55.5 million. In October 2025 it launched OPAQUE Studio, a tool for building confidential AI agents on LangGraph.

Named customers include ServiceNow, Accenture, and Encore Capital. A reported licensing arrangement in which Anthropic obtained some of OPAQUE’s confidential-AI technology and hired several of its engineers surfaced in October 2026 coverage, though that report rests on limited secondary sourcing and OPAQUE continues to operate as an independent company.

Innovation Matrix Assessment

Innovation Velocity 7/10

Building on peer-reviewed Berkeley research, the company shipped OPAQUE Studio for confidential AI agents and expanded into post-quantum and sovereign-cloud use cases alongside its latest raise.

Operational Value 7/10

Lets regulated enterprises run AI on sensitive data with hardware-backed proof of data handling, directly addressing a compliance blocker that otherwise stalls AI adoption.

Market Momentum 7/10

A Series B at a $300M valuation, named enterprise customers (ServiceNow, Accenture, Encore Capital), and a reported Anthropic technology relationship are independently reported signals of real traction.

Category Disruption 6/10

Confidential computing for AI workloads is technically distinct from typical AI security tooling, though the broader confidential-computing market includes other credible players such as Anjuna and Fortanix.

Real-World Efficacy 5/10

Named enterprise customers lend some credibility, but no independent benchmark or incident-prevention evidence was found.

Enduring Relevance 8/10

Confidential computing for AI addresses a structural, growing need as regulated industries expand AI use on sensitive data.

Why CISOs Should Care

Lets regulated enterprises use AI on sensitive data with cryptographic proof of how that data was handled, addressing a compliance blocker that otherwise stalls AI adoption outright.

What Makes It Different

Built on peer-reviewed confidential-computing research and hardware-backed attestation, rather than policy-layer controls or contractual promises about data handling.

The Matrix Verdict

67/100 — INCREMENTAL INNOVATOR

Incremental Innovator, approaching Meaningful. Credible technical pedigree, real funding, and named enterprise customers in a still-maturing confidential-AI niche.

Editorial Note: Claims vs. Verified Findings

The reported Anthropic technology-licensing and engineer-hiring arrangement comes from limited secondary sourcing and is unconfirmed by a primary announcement; named customers and funding figures are independently reported by multiple outlets.

Sources