Partnership · September 5, 2026

High-performance compute for business software.

TETRA CA PBC and HorneSci have entered a cooperative agreement. The partnership aims to bring high-performance computing methods used in national laboratories to business software.

Download the announcement
SAME MATRIX MATHEMATICSTETRA · DATA MATRIXHORNESCI · SPARSE TENSOR

HorneSci

  • U.S. national labs

    The founder's background: years of contracted work in the ecosystem

  • 15–21×

    Tensor-based routing acceleration, in HorneSci benchmarks

  • 2 + 2

    Patents granted, and patents pending

  • 0 new chips

    DeltaRouting runs on today's silicon

G-07 · The announcement

What each side brings.

On 5 September 2026, TETRA CA PBC and HorneSci entered a cooperative agreement. The aim is to bring high-performance compute, which has belonged to national laboratories and research institutions for decades, to ordinary business software.

Storing more data makes software more expensive, and the cost grows faster than the data. Optimizing code, adding hardware and adding caches each buy a fixed amount of headroom and then have to be repeated.

TETRA contributes TETRAbase, a matrix database. The LDBC Social Network Benchmark at scale factor 1 takes 211.5 MiB in TETRAbase and 1.683 GiB in Neo4j, 8.1 times less. TETRAbase uses no indexes and has two interfaces onto the same data: SQL, and openCypher over the Bolt protocol. The announcement states that existing drivers and client libraries for those interfaces work against it unchanged.

HorneSci contributes Hierarchical Sparse Tensors, a method for running matrix operations faster on the same hardware. Most entries in a matrix of business records are empty, and sparse tensor methods compute only on the entries that hold values, which can shorten queries and reduce energy use. HorneSci was founded by Hunter Horne, who has years of experience supporting contracted work in the U.S. National Laboratories ecosystem.

The integration applies HorneSci’s methods to the matrix operations inside TETRA, with the goal of lower query latency and energy use. HorneSci customers get TETRAbase as a general-purpose interface to Hierarchical Sparse Tensors.

TETRA can run alongside existing systems, hosted by TETRA or on the customer’s own infrastructure, so a business can move workloads over in its own time. The agreement is signed by Kyle Mickey and Reid Johnson for TETRA CA PBC and by Hunter Horne for HorneSci.

The announcement · PDF · 6 pages TETRA and HorneSci, September 5, 2026 Download →

G-08 · The documents

Published together.

Both dated September 13, 2026, and both free to download. The research brief examines whether software can reduce data center energy requirements. The datasheet describes the potential energy and cost savings.

See what it saves on your stack.

One workload, your data, a free assessment.

Free. No obligation. Your data stays where it is.