92 patents currently held
Held is read from USPTO's recorded assignments; a transfer never recorded there does not appear.
The next term ends in Dec 2029. from the record
Free plan: 3 watched companies. Pro: unlimited.
Alerts on its grants, lapses and transfers.
4
Ending by 2030
Dec 2029
Next term end
43
Patent families
When these terms end
Held patents by the year their term ends
notable patents
all others
Updated Nov 14, 2026 · includes USPTO assignment & status data published Nov 14, 2026 and grant & fee data published Nov 10, 2026 (weekly) · how we know
Notable: the highest-ranked patents in this portfolio by citations and family size.
Notable patents
Ranked by citations and family size, against patents of the same kind granted the same year. How we rank
Notable
Ends Apr 2041 · High confidence
Also notable
Ends Oct 2037 · High confidence
Also notable
Ends Jun 2030 · High confidence
Spatial Locality Transform of Matrices
US 11,537,687
Cited by 34 later US patents, top 5% of utility patents granted in 2022 in our data. Protection runs to 2041.
Multi-Pass Compression of Uncompressed Data
US 10,448,054
Cited by 44 later US patents, top 5% of utility patents granted in 2019 in our data. Protection runs to 2037.
A Method of Transferring Data with Offsets
US 8,589,600
The most cited patent in this portfolio, cited by 78 later US patents, top 10% of utility patents granted in 2013 in our data. Protection runs to 2030.
Fee deadlines
Maintenance fees fall due 3.5, 7.5 and 11.5 years after grant. A missed fee ends a patent at the end of the six month grace period that follows, not on the due date.
- 1 fee due Apr 15, 2027The maintenance fee for Multi-Pass Compression of Uncompressed Data (US 10,448,054) comes due Apr 15, 2027. Unpaid by Oct 15, 2027, it lapses early.
- 2 fees due May 7 to 21, 2027 · next May 7
- 2 fees due Jun 24 to 26, 2027 · next Jun 24
- 4 fees due Jul 9 to 16, 2027 · next Jul 9
- 1 fee due Aug 6, 2027The maintenance fee for Power Optimization In An Artificial Intelligence Processor (US 11,892,896) comes due Aug 6, 2027. Unpaid by Feb 7, 2028, it lapses early.
- 1 fee due Sep 5, 2027The maintenance fee for Power Grid Distribution for Tensor Streaming Processors (US 11,921,559) comes due Sep 5, 2027. Unpaid by Mar 6, 2028, it lapses early.
- 1 fee due Oct 16, 2027The maintenance fee for Energy Provisioning (US 11,960,346) comes due Oct 16, 2027. Unpaid by Apr 17, 2028, it lapses early.
Recent changes
- Sep 22Assignment recorded on 113 records: Groq, Inc. to Groq LLC Recorded Sep 22, 2026, executed Jun 22, 2026. Conveyance as recorded: Change of Name. A recorded assignment is a filed document, not a statement of current ownership. Reel 76553, frame 41.
- Sep 11Assignment recorded on 1 record: Ross, Jonathan Alexander to Groq, Inc. Recorded Sep 11, 2026, executed May 26, 2010. Conveyance as recorded: Assignment of Assignor's Interest. A recorded assignment is a filed document, not a statement of current ownership. Reel 75979, frame 231.
- 3 of the patents on this record were granted since Aug 16, 2026
What they patent
The 5 largest of its 43 families, grouped by real patent family and labeled by the earliest filing in each. No technology classification is inferred from titles.
| Tiled Switch Matrix Data Permutation Circuit | 34 in force | to Jun 2041 |
| Power Optimization in an Artificial Intelligence Processor | 3 in force | to Jun 2040 |
| Multi-Pass Compression of Uncompressed Data | 3 in force | to Nov 2038 |
| A Method of Transferring Data with Offsets | 3 in force | to Jun 2030 |
| Deterministic Memory for Tensor Streaming Processors | 2 in force | to Jul 2042 |
Patents held now
Soonest term end first.
| Patent | Title | Status | Expires | Confidence |
|---|---|---|---|---|
| US 8,639,859 | Method of Transferring and Processing Data for Efficient Parallel Computation In force · High confidence | In force | High confidence | |
| US 8,552,740 | Method of Measuring Delay in an Integrated Circuit In force · High confidence | In force | High confidence | |
| US 8,626,964 | Method of Transferring Data, a Method of Processing Data, an Accelerator, a Computer System and a Computer Program In force · High confidence | In force | High confidence | |
| US 8,589,600 | A Method of Transferring Data with Offsets In force · High confidence | In force | High confidence | |
| US 8,464,190 | Method of, and Apparatus for, Stream Scheduling in Parallel Pipelined Hardware In force · High confidence | In force | High confidence | |
| US 8,689,156 | Method of, and Apparatus for, Optimization of Dataflow Hardware In force · High confidence | In force | High confidence | |
| US 8,972,923 | Method and Apparatus and Software Code for Generating a Hardware Stream Processor Design In force · High confidence | In force | High confidence | |
| US 9,367,658 | Method and Apparatus for Designing and Generating a Stream Processor In force · High confidence | In force | High confidence | |
| US 8,930,876 | Method of Debugging Control Flow in a Stream Processor In force · Medium confidence | In force | Medium confidence | |
| US 8,631,380 | Method of, and Apparatus for, Data Path Optimisation in Parallel Pipelined Hardware In force · High confidence | In force | High confidence |
Also on record, not counted in the 92: 49 applications pending · 1 application abandoned. List them
Sources and methods
Compiled from USPTO application file histories, grants and recorded assignments, with the company's own names reviewed by hand. How this record was built
Portfolio grouped automatically from ownership conveyances to the names matched to this company, usually every name that begins with its own, so it counts patents an affiliate holds; each list row names the holder of record. Security interests are excluded and ambiguous records are not reviewed by hand.
Computed patent terms end on the dates shown. Related patents and pending applications may cover overlapping subject matter.