The same architecture that runs the Validiti substrate on silicon today is structurally a better target for light-wave computing than any workload the photonic-computing industry has been chasing. We're publishing the architectural argument and inviting silicon-photonics foundries and academic photonic-compute labs to collaborate on implementation. This is not a product for sale. It is a research direction we are offering to the field.
Validiti owns the substrate architecture. Photonic-computing foundries own the fabrication capability. The match between them — what we believe to be the most natural mapping of a substrate-shape compute primitive onto a physical photonic medium — is published here for the field to evaluate. Architectural license terms favor genuine research collaboration. Validiti does not operate fabs and will not. The shape of this engagement is closer to a Bell Labs preprint than to a commercial roadmap.
Photonic computing has spent two decades targeting neural network matrix multiplication. The fit has always been structurally poor — NN inference needs higher precision than photonic systems can deliver cleanly, weights need to be reprogrammable in ways photonic circuits resist, intermediate state between layers requires storage that photons don't naturally provide, and modern attention-based NN operations don't fit photonic primitives at all.
The Validiti substrate's primitives — the operations it runs at its core — map onto photonic computing's native operations almost one-to-one. Substrate-shape pattern lookup is what holographic media do natively. Substrate-shape joint-distribution match is what optical correlators have done since the 1960s. The fundamental distance computation the substrate uses is exactly what interferometry plus a photodetector array compute physically. And the substrate's multi- scale read pipeline becomes a single-pass parallel operation when expressed as wavelength-division multiplex.
The photonic-computing industry has been targeting the wrong workload. The substrate is the one it natively wants to run.
Each substrate operation has a photonic native equivalent. The mapping is direct, not metaphorical.
| Substrate operation | Photonic native equivalent | Fit | |
|---|---|---|---|
| Library lookup by frequency-ranked code | → | Holographic content-addressable storage with multi-pattern superposition; reference-beam readout returns match intensity per stored pattern | Native |
| Joint-distribution match across multiple dimensions | → | Joint transform correlator computing N-dimensional correlation via Fourier optics in a single lens-transit | Native |
| Bit-level distance between encoded patterns | → | Interferometric phase comparison; photodetector array sums the bit-position differences as photocurrent integration | Native — interference IS the operation |
| Multi-resolution coarse-to-fine descent | → | Wavelength-division multiplexing: separate wavelengths read separate resolution layers in parallel through the same medium | Native — parallel by construction |
| Append-only library writes | → | Holographic add-pattern in phase-change waveguide media; new patterns add to existing diffraction structure without rewriting | Good fit |
| Conditional read with neighbor expansion | → | Holographic readout with reference-beam phase variation reads neighboring patterns in spatial-frequency space | Native extension |
A working photonic substrate chip would compute substrate-shape decisions at the speed of light through a centimeter of medium. Latency budget estimated at 1-5 nanoseconds end-to-end, including the modulator and photodetector electronics that necessarily bracket the optical core. That is two to three orders of magnitude faster than substrate on classical silicon.
Energy per decision drops by three to four orders of magnitude. The library medium itself consumes no power; only the laser source and the electronic I/O draw current. Total chip-level power for a substantial library and channel count sits in the half-watt to two-watt range.
The faster floor turns several physical phenomena from "below the electronic intervention floor" into "catchable" for the first time.
Cognitive electronic warfare, distributed spectrum coordination, and RF jamming defense all face cascades that propagate at nanosecond bandwidths. Photonic substrate is the only architecture under serious consideration that can run substrate-shape recognition at those rates.
Shock waves crossing engine chambers, pressure vessels, and reactor cores in hundreds of microseconds become trivially intercepted at sub-microsecond decision latency.
Tokamak edge-localized modes and disruption precursors operate at microsecond timescales. Photonic substrate enables active intervention rather than passive disruption recovery.
Beam loss in high-energy accelerator chains can be intercepted before beam-dump events that take days to recover from. Microsecond decisions; photonic substrate well inside the budget.
Stepped-leader formation patterns develop over tens of microseconds. Photonic-speed decision plus high-voltage actuators could shift attachment points away from protected assets.
Laser ignition + directed-energy systems experience nanosecond-scale backscatter events that damage components. Photonic substrate sits inside the optical path itself and intercepts in-line.
Three categories of collaborator we are inviting to evaluate and co-develop this direction.
Architectural license + collaborative tape-out programs. We supply the architecture; the foundry supplies the fab. Royalty and license terms negotiated per partnership.
Open research collaboration. Joint preprint publication. Demonstrator builds on lab test benches. No commercial commitments; mutual academic credit.
DARPA, NASA, AFRL, ONR, NSF, DOE Office of Science. The application targets in Section 04 align with multiple existing program lines. Architectural license terms suitable for federal partnership.
Validiti is not raising capital to build a photonic foundry. We do not operate fabs. We are not pursuing this as a commercial product for the Marketplace. The substrate-on-light architecture is research-mode work, published for the field, available for licensing to foundries and research collaboration with labs. The substrate's classical silicon implementation continues to serve customers through the fabric Marketplace at present.
If a photonic substrate chip is built in the next decade, it will be built by the silicon-photonics industry under license, not by Validiti directly. Our role is to have supplied the architecture.
Architectural argument, available on request to qualified silicon- photonics researchers and foundry-affiliated investigators. Preprint paper in preparation; will appear on arXiv and through the relevant silicon-photonics conferences. Patent filings on the substrate-on- photonic-medium architectural mapping in process.
Quantify this direction’s workload · Compute per Solar Watt →