A Mach-Zehnder modulator splits an optical signal into two arms, applies a differential phase shift via the electro-optic effect, then recombines the arms so that constructive or destructive interference encodes the data. A ring modulator instead uses a microring resonator evanescently coupled to a bus waveguide: when the drive voltage shifts the ring's resonant wavelength onto or off the bus wavelength, the signal is modulated. MZMs offer inherently broadband, chirp-manageable operation and are preferred for long-haul and high-baud-rate coherent links, while ring modulators are compact and energy-efficient, making them attractive for short-reach datacom. The trade-off is thermal sensitivity: rings must be actively stabilised, and that stabilisation challenge is one of the remaining open areas in the patent corpus.

The corpus of 116 families is distributed between a small number of recurring assignees and a longer tail of single-filing entrants. The top of the ranking is rendered in the assignee table on this page. Notably, a tight co-assignment cluster — linking a Dutch national research organisation, a technical university, and a commercial photonics firm — accounts for the densest collaboration activity, each pair sharing six co-assigned families. Several large semiconductor and telecommunications firms that filed actively through 2022 have shown zero new families in the most recent data year, which may signal a strategic shift rather than withdrawal from the technology.

Within this corpus, TFLN modulator claims appear as a subset of the broader electro-optic modulator population classified under G02F1/035. Filing density on TFLN-specific integration — particularly thermal compensation methods, electrode geometry optimisation, and heterogeneous bonding to silicon waveguide platforms — is lower than on mature silicon carrier-depletion MZM designs, suggesting meaningful white space remains. However, 'lower density' is not the same as 'clear': several foundational TFLN platform patents from outside this corpus (filed at CNIPA, for example) may still affect freedom to operate. Any entrant should run a platform-specific search before treating TFLN as open territory.

With 69 of 116 families filed at the USPTO and only 2 at CNIPA, the claim landscape is geographically skewed. A company manufacturing or selling primarily in Asia-Pacific faces a different risk profile than a US-market competitor: the dense US prior art constrains claim breadth for new USPTO filings, but the thin CNIPA presence means that a well-drafted Chinese filing on an integration approach not yet claimed there could achieve broad protection. European filers face an intermediate situation — 19 EPO families and 16 PCT families provide meaningful prior art, but EPO examination standards and prosecution practice differ enough from the USPTO that claim scope may vary substantially between jurisdictions. Geographic portfolio planning should treat these three jurisdictions as distinct rather than assuming US allowances will be mirrored elsewhere.

A plateau in family counts at the architecture level is a maturity signal, not a decline signal. It indicates that the primary claim space — broad independent claims on MZM, ring modulator, and TFLN modulator structures — is sufficiently occupied that new filers are working in a narrower window, typically refining manufacturing methods, control electronics, or system-level integration rather than the optical element itself. The G06N co-classifications (AI-assisted modulator control) and the H01L co-classifications (driver–modulator integration) are the visible edges of this shift. Publication lag also means the most recent year in the trend is understated; the true 2025 filing rate will not be fully visible until mid-2026 at the earliest. Read the plateau as a signal to focus on adjacent integration claims rather than as evidence that the field is commercially unimportant.

A closed three-party co-assignment structure — where each pair shares the same count of co-assigned families — strongly suggests a formal joint research or consortium agreement rather than ad-hoc collaboration. Co-owned patents require all owners to consent to exclusive licensing in most jurisdictions, which can complicate exclusive deals but also means the patent holder group is likely to have a coordinated licensing strategy. Academic co-assignees (NWO and Eindhoven) are typically more willing to license than vertically integrated commercial OEMs, but the presence of a commercial co-assignee (POET Technologies / POET Technologies Inc.) means any licence negotiation will involve commercial interests as well. Any company seeking a licence to or freedom to operate against this cluster's families should approach all three co-assignees.