Compute-in-Memory Analog Accelerator Patents: Who's Filing and Where the Gaps Sit
46.3% of the field sits with five assignees. 332 of 717 records in scope belong to the top five filers, a concentration that makes freedom-to-operate checks against a short list the first move for anyone entering this space. Annual filings rose from 9 records in 2017 to a peak of 84 in 2022, then eased to 70 by 2024 -- a -14% move across that three-year window. 2025 and 2026 counts are still incomplete because publication lags filing by roughly 18 months; treat the tail as a floor, not a decline.
- 1HEWLETT PACKARD ENTERPRISE DEV LP106
- 2TETRAMEM INC73
- 3INTERNATIONAL BUSINESS MACHINE CORPORATION68
- 4QUALCOMM INC48
- 5PACIFIC BIOSCIENCES OF CALIFORNIA INC37
See the full compute in memory analog accelerators analysis in Eureka
- The complete ranking, not just the top five
- Every IPC branch with its share of the corpus
- The most-cited records, and where claim space is still thin
Common questions about this landscape
Who holds the most patents in compute-in-memory analog accelerators?
The dataset ranks 100 assignees across 717 records, with the leading filer holding 106 records -- well ahead of fifth place at 37. The top five combined hold 46.3% of all records in scope, so a small number of large filers set the competitive baseline. That said, more than half the field sits outside those top five, spread across a long tail of smaller filers and single-patent entrants.
Is patent filing in this field growing or slowing down?
Filing rose steadily from 9 records in 2017 to a peak of 84 in 2022, then eased to 70 by 2024 -- a -14% change over that three-year window, the most recent span the data can treat as complete. Counts for 2025 and 2026 look lower still, but that is an artefact of publication lag rather than a real drop, since publication typically trails filing by around 18 months. Read the recent tail as understated, not as evidence of a slowdown.
What technical problems dominate the patent claims in this space?
The search was built around the practical failure modes that decide whether an analog accelerator works outside a lab: conductance variation between cells, IR drop across crossbar bit lines, the analog-to-digital conversion step, weight-mapping accuracy, and per-operation energy cost. IPC data confirms the split -- G06N (AI-model computing) and G11C (memory) each cover over 40% of the 717 records, with G06F (digital data processing) close behind, showing claims spread across the compute, memory and system-software layers rather than concentrating in one.
Disclaimer. This analysis is based on Patsnap Eureka data drawn from a limited snapshot of global patent records and is provided for general information and reference only. Patent data carries inherent limitations — recent filings are under-counted because of publication lag, counts may be on a record or family basis, classification and applicant-name data may contain errors or duplicates, and the underlying search query defines the scope shown — so the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.
Nothing here is an exhaustive prior-art, novelty, freedom-to-operate or validity search, nor does it constitute legal, financial or professional advice, and it should not be relied upon as such. Verify independently and review with qualified patent and legal professionals before acting on it.
Method: Filing trend and technology composition. Derived from a Patsnap search on Compute in Memory Analog Accelerators covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish. Every share divides by all records in scope. Data: Patsnap Eureka. See the full landscape report.