Skip to content

Press Releases

Seyon NexGen opens SNX-C7 core licensing to foundry and OSAT partners

Multi-layer blueprints for the SNX-C7 compute core are now available to chip foundries and OSAT partners under a tiered licensing framework.

June 18, 2026 · 2 min read

Illustration of a Seyon NexGen SNX-C7 processor die with radiating circuit traces

Seyon NexGen has opened licensing for the SNX-C7, the company's highest-density RISC-V co-processor core, to chip foundries and OSAT partners worldwide. The release moves the core out of internal deployment and into a formal, tiered licensing framework covering everything from synthesisable RTL to reference firmware and the trust stack that runs on top of it.

The SNX-C7 is built for one job: running quantised neural networks at the edge, at milliwatt power budgets, with a cryptographic receipt for every inference it performs. It is not a general-purpose application processor, and that focus is what lets it hit the efficiency numbers it does.

What is in the licence

Licensees receive the full delivery package rather than a black-box hard macro. In practice that means partners can retarget the core for their own process node without waiting on us:

  • Synthesisable RTL for the RV32 and RV64 configurations, with the vector and integer-quantisation extensions enabled
  • The 8-bit integer neural runtime, pre-tuned for the core's memory hierarchy
  • Reference firmware, including the local storage ring-buffer architecture published as open source in April
  • The Inference Verification Protocol implementation — the hardware root of trust that signs each decision
  • Verification collateral: testbenches, coverage reports, and the compliance suite we run internally before every tape-out

Three tiers, matched to how partners actually build

The framework separates evaluation from production so a design team can prove the core out before committing to a volume agreement.

Foundry tier

For partners integrating the core into a process-specific IP catalogue. Includes physical-design collateral and characterisation support across corners, plus the right to sublicense within an agreed customer set.

OSAT tier

For assembly and test partners packaging the core alongside sensor front-ends. Covers the test-mode interfaces, the fingerprint-provisioning flow used to burn a unique identity into each die, and the attestation keys that flow depends on.

ODM tier

For device makers building on a reference platform. Includes the board designs, the runtime, and annual activation tokens for fielded units.

Why the fingerprint matters commercially

A licensable core that produces verifiable output changes what a partner can sell. When a detection carries a signature proving which model version processed which input, the resulting record survives an audit — which is the difference between a monitoring product and a compliance product. Regulated sectors, from civil aviation to remote patient monitoring, buy the second one.

That property has to be built into the silicon. Bolting attestation on in software leaves the signing key somewhere an attacker can reach; burning the fingerprint into firmware at compile time, and signing in a hardware root of trust, does not.

Availability

Evaluation packages are shipping now. Production licensing runs on the standard annual activation cycle, with workload review included in every tier — partners send us a target model and deployment profile, and we size the configuration before anyone signs anything.

Foundry and OSAT partners can start the process through the licensing channel; developer teams evaluating the runtime should use the developer ecosystem channel instead.

Leave a Reply

Your email address will not be published. Required fields are marked *