Cornelis Networks Patent Portfolio Analysis

Two hundred and five million, one pulse

The Series C was announced on 14 September 2026: $205M, with Alumni Ventures, Foresight Group, Global Brain Corporation, Adit Ventures and Intel Capital named. The company was founded in 2020. That is six years from incorporation to a round of this size, which is neither fast nor slow, but it does mean the technical decisions being funded here were made a long time before the cheque cleared.

The useful thing about a single dated event is what it lets you measure backwards from. The most recent filing anybody outside the company can read carries a date of 15 September 2024, almost exactly twenty four months before the announcement. Everything filed since then is still inside the publication delay. The visible record stops two years short of the moment the money arrived.

Eighteen ticks, eight of them solid

The first filing is dated 25 June 2021, roughly a year after founding. Eighteen families follow. Eight have granted and are enforceable today. Ten are published and pending, and the earliest of those has been waiting long enough that a grant could land at any point. Filings sit in two offices, the United States and Europe. Nothing exotic, nothing scattered.

Eighteen is a small number for a company that has just raised nine figures. It is also a deliberate one. A portfolio this size cannot have been assembled by filing everything the engineers wrote down. Somebody chose. The interesting question is not how many ticks are on the arc but where on the arc the solid ones fall, and what their claims actually reach.

Eight granted families are eight enforceable ones.

Digital communication, with circuits underneath it

The classification weight is lopsided. Digital communication codes, H04L, account for 96 hits across the portfolio. Data processing, G06F, accounts for 15. Read on its own, that shape says protocols and routing behaviour. Read alongside the titles, it says something more specific. The recent filings cover as-needed forward error correction, dynamic maximum throughput capacity-based load balancing for adaptive routing in high-performance networks, and telemetry generation circuits for load-balanced fine-grained adaptive routing in high-performance system interconnect.

That third title is the one to sit with. Telemetry generation circuits. Not a method, not a scheme: circuits. Meanwhile the open roles include Principal ASIC Verification Engineer and ASIC Design Engineer, and hardware roles make up half of the hundred positions posted. The claims are classified as communication. The hiring is silicon. When those two disagree, the device claims are usually where the value is hiding.

Ten dots, and one that keeps recurring

Ten inventors are named across the eighteen families. Gary Muntz appears on eleven of them. Aruna Ramanan appears on four, Brent Rothermel on three. The remaining seven names are thinly spread. There is no wide bench here and no sign of one forming between 2021 and 2024. One person's judgement runs through roughly two thirds of everything the company has filed.

Concentration like that is a fact with two faces. It gives the portfolio a coherence that committee-filed estates rarely have: the families read as one continuing argument rather than a collection of departmental outputs. It also means the institutional memory of why each claim was drafted the way it was lives largely in one head. Both things are true at once, and both matter when the scope is tested.

Two thirds of the estate carries the same signature.

Eighteen against fourteen hundred, same classification

In the same CPC groups over the last five years: Huawei with 1,385 families, Samsung with 606, Cisco with 477, Intel with 467, Micron with 432, Suzhou Yuannao Intelligent with 325. Cornelis Networks has eighteen. On a shared scale the dot is barely visible. This is not an empty field that a small team wandered into. It is one of the most heavily worked classification areas in the record.

Volume is not the only signal on the page, though. Organisations whose own documents cite these filings include Huawei with four citations, Microsoft with four, Samsung with three, NVIDIA with two and Mellanox with two. Eighteen families from a company founded in 2020 have entered the reading of examiners and drafters working inside far larger estates. In a crowd this dense, that is what narrowness earns you.

What the two-year gap is holding

So the picture resolves. A small, tightly authored portfolio, started in 2021, concentrated in digital communication claims but drafted in places as circuits, built by a company now hiring ASIC designers and verification engineers at scale. Eight families enforceable, ten waiting. Against filers with two orders of magnitude more volume in the same groups, the only thing that can matter is where the granted claims actually land.

And the record goes quiet in September 2024. Whatever the $205M is funding was conceived well before the round, and if it was filed, it was filed into a window that has not opened yet. The visible eighteen are the argument so far. The next eighteen months of publications will show whether the same hand kept drafting, and whether it kept drafting circuits.