Swegan Patent Portfolio Analysis

Money arrives, five years after the paperwork

A $14M Series B from Atlantic Bridge, Almi Invest and MediaTek. MediaTek is the name worth pausing on: a chip company putting money into a Swedish epiwafer house founded in 2014. That is not a generalist bet. It is somebody downstream deciding the material layer matters enough to own a piece of.

The interesting thing about this pulse is where it sits on the timeline. SweGaN's last publicly visible patent filing is dated 20 July 2021. That is sixty-one months before this round closed. Whatever the company has filed since then is still inside the eighteen-month publication window, or was never filed at all. The public record stops well before the money starts.

Four solid ticks and three thin ones

Seven families since 2015. Two in 2015, three in 2017, one in 2019, one in 2021. Four have granted. Three are published and pending, and the first of those is old enough that it could issue at any time. On the arc, the solid amber marks are the ones that are enforceable today, which is a real distinction and not a bookkeeping one.

The titles tell you what the company decided to protect. A heterostructure for a high electron mobility transistor and its production method. Then two closely related filings on semiconductor device structures with recessed ohmic contacts. This is not a broad materials-science land grab. It is a narrow, repeated claim on the interface where the epitaxial stack meets the metal, filed in CN, KR, TW, US and WO.

Four granted families is a small portfolio, but four granted families is not nothing.

One name on six of seven

Twelve inventors appear across the portfolio. One dot carries almost all of it. Jr-Tai Chen is named on six of the seven families, and appears again under a variant spelling on five. Erik Janzen appears on three. Everyone else is a single appearance or two, clustered around the middle years.

Portfolios shaped like this are the signature of a founding technologist who never handed the work off. The filings track one person's understanding of epitaxial growth deepening over six years, from the heterostructure itself to the contact geometry built on top of it. That concentration is a strength when you are reading the claims for coherence. It is a question when you are reading the company for depth.

Seven families against nearly five thousand

The CPC groups where SweGaN's claims sit, H10D and H10P, have drawn TSMC with 4,969 families in five years, Samsung with 2,136, Semiconductor Energy Lab with 1,010, Intel with 970, IBM with 827 and Toshiba with 785. SweGaN's seven sit on the same scale. On the diagram the circle is barely visible.

That comparison is less damning than it looks, and more instructive. Nobody expects a Linkoping epiwafer company to out-file TSMC. What it tells you is that count is a useless metric here and scope is the only one that matters. A narrow claim on a recessed ohmic contact in a crowded class is worth more than a hundred filings that duplicate what six giants already cover.

Who reads a small Swedish portfolio

Aixtron cites SweGaN four times. Globalwafers four. National Yang Ming Chiao Tung University three, Infineon three, Huawei three. Read that list by role rather than by name. Aixtron builds the deposition equipment. Globalwafers makes the substrates. Infineon and Huawei put GaN devices into products. A university sits in the middle doing the science.

Citations are not endorsements and they are not permissions. What they are is evidence that the claims are legible to people who work at every layer of this stack, from the reactor to the finished part. For a seven-family portfolio in a class where TSMC files five thousand times over, being visible to the equipment makers and the device makers at once is the more meaningful signal.

Equipment, substrate and device makers all found the same seven families worth citing.

Hardware in the claims, software in the hiring

Eighteen open roles. Five in software and AI, five uncategorised, four in go-to-market, two in hardware, two in bio or clinical. The named ones lean the other way: SiC PVT development engineer, senior R&D epi engineer. So the company is recruiting deep in crystal growth and epitaxy while simultaneously staffing a software function that no filed claim touches.

That is the shape to hold onto. A patent record that is entirely about the physical structure of a semiconductor layer, four families enforceable now, one inventor behind almost all of it. A hiring plan whose largest single bucket is software. Between the 2021 filing and the 2026 money there is a five-year gap in the public record, and the answer to whether the moat grew during it is not yet published.