Delos Data Patent Portfolio Analysis

The money arrived fifteen months late

One pulse, September 2026. A hundred million dollars in a Series A, with Matrix, Playground Global, Matter Venture Partners, Capricorn Investment Group and Socratic Partners on the line. For a company founded in Canada in 2020, that is a long quiet stretch before a loud event.

The paper record does not follow the money. It precedes it. The first filing is dated 4 June 2025. The most recent published one is 19 September 2025. Everything on file was in an examiner's hands more than a year before the round was announced. Whatever conviction the investors bought, it had already been written down and dated.

The filings are older than the round.

Nothing yet, and that is on schedule

Two patent families. Both published, both pending, none granted. The granted ribbon has no width at all. On a chart built to show the ratio between issued and outstanding rights, this company is entirely outstanding, which is the normal condition of a portfolio started in mid 2025.

First grants should land around 2028. Until then there is no enforceable right here, only a claim of priority and a public description of what the team thought was new in the summer of 2025. That makes the reading exercise simpler and harder at once: there is nothing to measure but scope, and scope is not settled.

One subject, restated three times over

The bands are not close. Digital communication, H04L, runs roughly six times the classification weight of data processing, G06F. There is no third band. For a record this small, that is unusually disciplined: no side bets, no defensive filings in an unrelated art unit, no hedging into some second idea the team half believed.

Read the titles and the reason is plain. Configuring network fabric between endpoint processing units. Configuring a network for connecting endpoint processing units that collectively execute a distributed application. A control plane proxy server for managing use of the network connecting endpoint processing units. One problem, approached from the fabric, the configuration and the control plane in turn.

Six names and no centre of gravity

Six inventors are named across two families. Daniel P Daly appears on one of the two. So do Edward V E Doe and Alain J E Gravel. No one is on everything. The highlighted dots are highlighted only because the list has to start somewhere, not because a single author sits behind the whole record.

That is a different shape from the usual young-company portfolio, where one founder's name is stamped on every page. Here the work is distributed across a group from the start. The hiring pattern is consistent with it: twenty open roles, eleven in software and AI, four in hardware, including Principal AMS Design, Physical Design Engineer and SerDes Validation Leader.

Two families against several thousand

Same scale, same classification groups, last five years. Huawei filed 2,801 families. China Mobile Communications 1,686. China Telecom 1,061. China Unicom 879. Cisco 841. Samsung 829. Delos Data has two. Cornelis Networks filed in the same group, H04L45, in the same week as one of these filings.

Nobody wins that comparison on volume, and volume is the wrong test. What a small record can hold is a narrow, early, precisely drafted position. What it cannot hold is breadth. In a class this heavily worked, the question is never how many pages a company owns. It is whether the claims that survive examination sit somewhere the large filers did not reach first.

Count is not the contest here. Scope is.

What 2028 will actually settle

Set the pieces back together. A four-month burst of filing in 2025, all of it on configuring and controlling the network between endpoint processing units. Six inventors, none dominant. A hundred million dollars arriving fifteen months afterwards. A classification space where six organisations have filed thousands of families since 2021.

The moat, if there is one, is not in the count and not in the funding. It is in three titles that describe one layer of a system, filed before the round, in a class that is worked hard by very large filers. That either narrows into something defensible by 2028 or it does not, and the record will say which.