The tracked corpus of 24 families includes established chipset and infrastructure vendors as well as a radio-unit specialist, with filing concentrated around a 2022 peak. No single assignee shows continued filing momentum into the most recent tracked year, so current leadership is better read from family counts and citation influence than from recent activity. Anyone assessing this space should weigh both the historical filing volume and the fact that momentum has stalled across the board.

The dataset shows filings rising to 7 in 2022, the highest point in the tracked range, with no year since matching that level. This pattern is consistent with a narrow technical problem where the core claim positions — waveform generation, fronthaul handling, PHY-layer split — were staked early and few new angles have opened since. Publication lag of roughly 18 months means the latest one to two years are understated, so some of the apparent flattening may still resolve upward as more filings publish.

EP4138472A1, filed by Mavenir Systems, covers an O-RAN single frequency network in which multiple radio units transmit the same signal simultaneously and the distributed unit filters uplink or downlink signals based on power or signal-to-noise ratio, with radio units linked directly to the distributed unit. It is one of the more heavily cited records in this dataset, which signals it is a frequently referenced prior-art point rather than automatically an infringement risk for every design. Whether it blocks a specific implementation depends on whether that design uses the same direct-link, power-or-SNR-based filtering approach rather than a different fronthaul aggregation method.

The evidence shows thin coverage in a few adjacent branches relative to core waveform claims: AI-assisted PHY parameter tuning, remote radio head low-power tuning procedures, and control-plane-specific downlink precoding. These sit inside the same IPC classes as the dominant filings but have far fewer records attached to them. A first claim in one of these branches would need to specify the exact mechanism — for example, a learned model tuning a named PHY parameter under O-RAN split architecture — rather than restating the generic waveform processing already claimed by the top-cited families.

The raw numbers suggest caution rather than a rush: 24 total families, a single peak year in 2022, and zero recent-year filings across every tracked assignee. That does not mean the area is closed, since publication lag hides the most recent filings, but it does mean a new entrant should expect to compete against already-established claims from named infrastructure and chipset vendors rather than filing into open ground. The better opportunity, based on this dataset, lies in the specific under-claimed sub-branches rather than the core waveform generation claims themselves.