The SparXiom IP Portfolio
The emergence of SparXiom is backed by a comprehensive intellectual property fortress - currently comprising five provisional patent applications filed with the USPTO in early 2026. They establish priority dates and foundational claims across proprietary sparse data structures (SpAx), asynchronous scheduling methodologies, and physical silicon micro-architectures (vector and systolic compute units).
At present, our IP portfolio spans three core Technology Pillars:
PILLAR I:
The "Liquid Vector" Paradigm
Secures the methodology for continuous, asynchronous sparse data-streaming.
Bypasses traditional SIMD lock-step execution, allowing independent register-level saturation on vector hardware without destructive data pruning or zero-padding.
PILLAR II:
The Sparse Streaming Coprocessor (SSC & SSC v2)
Secures the algorithmic concepts and micro-architectural blueprints for 1D vector acceleration.
Defines a dedicated silicon block utilizing Deterministic Memory Orchestration. Includes advanced mechanisms for variable-width vectorized sparse data streaming to support highly adaptable Edge AI workloads.
PILLAR III:
The Sparse Tiling Coprocessor (STC)
Secures the algorithmic concepts and micro-architectural blueprints for 2D systolic array acceleration.
Defines hardware-level sparse data co-discovery for "pseudo-dense" tiling, enabling rigid matrix-multiply units (e.g., AMX, TPUs) to process irregular, unstructured SpMM and SpGEMM operations at peak theoretical cadence.
The Long-Term R&D Horizon
Details of patent filings, cycle-accurate gold-models, emulation code and hardware blueprints can be shared and made available to qualified investors and potential partners under NDA.
(SEIS/EIS Advance Assurance has recently been granted by HMRC.)
Our future research and development pipeline will continue to explore and expand the SpAx streaming framework into other domains and beyond 1D and 2D HPC/AI operations. Future patent filings are currently being prepared to bring sparse data orchestration, for example, to:-
Scientific Exascale Computing:
Asynchronous, vectorized implementations of the Thomas Algorithm for Computational Fluid Dynamics (CFD).
Volumetric Spatial Computing:
The extension of SpAx logic into 3D sparse space and voxel tree traversal for next-generation physical simulations and robotics.