Within the 33 records in this scope, one assignee holds the largest share at 16 records, well ahead of the four other ranked assignees, each of which files in the single digits. This is a five-company ranking, the full set returned for this topic, not a trimmed top-50 or top-100 list, so the concentration is real rather than an artifact of cutting off a longer list. That said, thirty-three records is a modest corpus, and a reader doing freedom-to-operate work should treat this as a starting map rather than an exhaustive count of every relevant filing worldwide.

Filing rose from effectively zero in the earliest years of this scope to a peak of 13 records in 2022, then declined. A formal growth rate is not calculable here because fewer than four complete years remain once the roughly 18-month publication lag is accounted for, so treat any apparent decline in 2024-2026 with caution: those counts will rise as later filings publish. The honest read is that the field passed its most active filing period around 2022 and has not yet shown a clear return to that pace.

US20230305993A1, assigned to Intel, describes a modular parallel processor built from multiple chiplets populating slots on a base die, where chiplets are tested for characteristics like functional-unit count or power consumption and then grouped into chunks with a pre-determined collective value, such as total execution cores or aggregate power. It is one of the most-cited records in this scope. Anyone designing a chiplet product that tests and bins dies by a collective performance or power metric before assembly should read this claim set closely rather than assume the mechanism is generic.