International Business Machine Corporation is the largest filer in this dataset by a wide margin, with 554 records against a fifth-place figure of 52 and a tenth-place figure of 39. That gap means the field is concentrated at the top: the top 5 assignees together hold 42.4% of all 2,934 records in scope, and the top 10 hold 49.6%. Behind that leading group sits a long tail of assignees with only a handful of filings each, so a single company's dominance does not mean the surrounding claim space is closed.

Filing rose through the late 2010s, peaked at 234 records in 2021, and had fallen to 92 by 2024 — a 61% decline over that three-year span. However, publication lags filing by roughly 18 months, so 2025 and 2026 figures in any dataset are always undercounted and should not be read as evidence of a continuing drop. The more reliable read is that the wave of foundational architecture filing that built up to 2021 has largely settled, with several major historical filers showing near-zero activity in the latest complete year.

G06F, the electric digital data processing class, appears in 74.4% of the 2,934 records in scope, making it by far the dominant classification. H04L (digital information transmission) and H05K (printed circuits and assemblies) follow at 14.3% and 10.1% respectively, while AI-model computing (G06N) sits at 6.4%. Because a single patent record can carry multiple IPC classes, these percentages overlap rather than summing to 100%, and they should be read against the full 2,934-record base rather than against each other.

The thinnest overlap with core supercomputing architecture claims shows up in domain-specific applications rather than in the architecture itself: bioinformatics (G16B) touches only 3.0% of records and business/commerce data processing (G06Q) only 2.3%. That suggests parallel-compute and memory-hierarchy mechanics have been claimed extensively in general-purpose form but far less often when tied to a specific application domain, which is where a first claim is less likely to collide with dense prior art from the leading filers.

US20110219208A1, filed by GLOBALFOUNDRIES INC., describes a multi-petascale parallel supercomputer built from system-on-chip ASICs with four or more processors integrated onto a single die, each with full access to system resources and adaptive partitioning between compute and messaging I/O functions. It is representative of a recurring pattern in this dataset: claims anchored in packaging density, interconnect topology and per-node resource allocation rather than in software algorithms alone. Anyone drafting in adjacent territory should treat multi-processor-per-die packaging and adaptive I/O partitioning as claim elements likely to have prior art coverage.