Cohesity Patent Portfolio Analysis

One pulse, dated after the work

The pulse is 1 April 2020: a $250 million Series E. Now look at the dates on either side of it. Cohesity's first filing carries a priority date of 3 May 2011, nine years earlier. Add up the families begun before the money landed and the count is 64. The single busiest year anywhere on the record is 2019, with 23 new families started, twelve months ahead of the announcement.

Filing peaks usually trail funding. Hiring happens, budgets appear, drafting follows. Here the sequence runs the other way. The heaviest year of invention sits in front of the round, and the years after it never matched that rate: 16 families in 2020, 21 in 2021, then 14, then 12, then 12. The capital did not buy the positions in this file. It showed up once they were already taken.

The capital arrived after the positions were taken.

Almost nothing is still in the queue

Of 144 families, 125 are granted and 19 are published and pending. Eighty-seven percent has been through examination and come out the other side. That is an unusual shape. Most files of this age carry a thick layer of applications whose scope nobody can yet describe, because an examiner has not finished with them. This one does not. What Cohesity holds is largely fixed, tested and readable.

The thin unexamined remainder is where the newest thinking sits. The most recent published filing is dated 30 October 2025. Its neighbours are titled Unified metadata catalog for data management platforms, filed twice, alongside Dataset clustering and ai-assisted theme extraction and Artificial intelligence-assisted data management for diverse source systems. A settled body of granted work, and a narrow live edge pointed at metadata and at machine assistance.

One band carries nearly everything

G06F, data processing, appears on 140 of the 144 families. That is not a centre of gravity, it is the whole mass. H04L, digital communication, touches 22 families, and families can sit in more than one place at once, so the communication tag is mostly an overlay on data processing rather than a second business. H05K appears once. The file has one subject and has had it since 2011.

G06N, the AI and machine learning code, sits on six families. Six, against 140. Read that next to the recent titles about AI-assisted data management and theme extraction and the picture is specific: the machine learning claims are new and few, and they are being drafted onto a foundation of data-processing positions that were filed and granted years earlier. The novelty is recent. The ground under it is not.

Small circle, large neighbours, same codes

Over the last five years, other assignees filing in these same CPC groups include SAMSUNG ELECTRONICS CO LTD with 2,157 families, MICRON TECHNOLOGY INC with 2,114, DELL PRODUCTS LP with 1,181, SK HYNIX INC with 954, HUAWEI TECH CO LTD with 793 and SUZHOU YUANNAO INTELLIGENT TECH CO LTD with 668. Cohesity's entire file, from 2011 to 2025, is 144 families. On a volume scale it is a dot beside those circles.

Volume in a code area tells you who files a lot there. It does not tell you whose filings matter. These groups are among the most heavily worked in the patent system, which means a file of 144 families here was either drafted narrowly enough to survive or it was not. Eighty-seven percent granted is the answer to that question. The next stage is the answer to the other one.

Four organisations, four hundred and eighty-seven citations

The lines running back into this file come from EMC IP HOLDING CO LLC, which cites these filings 175 times, NUTANIX INC with 126, DELL PRODUCTS LP with 122 and IBM with 64. That is 487 citations from four organisations pointing at 144 families. DELL PRODUCTS LP appears twice in this piece: it files in the same CPC groups, at 1,181 families, and it also cites this file. The record shows both. Nothing more.

This is the thread. A company one twentieth the filing size of its code-area neighbours built a body of data-processing claims starting in 2011, drove it hardest in 2019, got 87 percent of it granted, and is now cited hundreds of times by organisations far larger than itself. Size decided who filed most in G06F. Timing and examination decided whose file the rest of the field has to read.

Volume decides who files most. Citation decides whose file others must read.

What the next nineteen filings decide

The granted 125 describe a decade-old conviction about data processing, laid down before the Series E and largely unchanged in direction since. The pending 19 describe something newer: unified metadata catalogs, dataset clustering, AI-assisted management across diverse source systems. Six G06N families is a small bet. It is also the only part of this file whose scope is not yet settled, which makes it the only part still capable of surprising anyone.

So the question worth holding is not whether Cohesity can out-file SAMSUNG ELECTRONICS CO LTD or MICRON TECHNOLOGY INC in G06F. It never has and the record says it never tried. The question is whether the citation weight now attached to the old data-processing claims extends to the new metadata and machine-assistance claims, or whether those arrive on ground that other filers reached first.