The Alphabet Came With a Design Doc. The Company Named After It Spent 35 Years Refusing to Die.

Make a hard “g” sound and pay attention to your tongue. The back of it humps up and seals the throat for a fraction of a second before it releases. Now look at the Korean letter for that sound, ㄱ. It is a side-on diagram of your tongue doing precisely that. The letter for “n,” ㄴ, is the tongue tipped up against the ridge behind your front teeth. ㅁ for “m” is a closed mouth seen straight on. ㅅ is the edge of a tooth, ㅇ is the open round throat. I find this faintly unbelievable every time it crosses my mind, because no other writing system humans use was built this way.

Hangul is the only major script with a named inventor and a surviving rationale for why every letter looks the way it does. King Sejong, fourth king of Joseon, finished the design in late 1443 and had it promulgated in 1446 as the Hunminjeongeum, “the proper sounds for instructing the people.” He did not just hand over the letters. He published a companion volume, the Hunminjeongeum Haerye, where his scholars laid out the articulatory logic: the basic consonants are stylized pictures of the speech organs, and you add strokes to a base shape to mark a related sound. Modern phonetics checked his homework centuries later and found it essentially right. Geoffrey Sampson classified Hangul as a “featural” system, a category it more or less invented. Sejong’s pitch was that a clever person could learn it in a morning and a slow one in ten days. The Haerye is now National Treasure No. 70, was lost for centuries until a copy surfaced in Andong in 1940, and sits on UNESCO’s Memory of the World register.

Sit with that for a second. In the 1440s, Korea shipped what is basically an open technical specification for a writing system, published with commentary so anyone could implement it. That is the cultural DNA I want you to hold onto, because five and a half centuries later a software company took the alphabet’s literal name and spent its entire corporate life fighting Sejong’s exact battle: who gets to control how Koreans write.

A word processor as national infrastructure

The company is Hancom, short for Hangul and Computer (한글과컴퓨터). It started in 1989 when members of a Seoul National University computer club shipped Hangul 1.0, a word processor built from the ground up for Korean, and incorporated in 1990. The founding crew is a startup-trivia goldmine: one of them, Kim Taek-jin, left to start NCsoft and turn Lineage into a gaming empire. The product became known by its file extension, HWP, for Hangul Word Processor, and it did something almost nothing else on the planet managed. It kept Microsoft Word from owning a national market. After Windows ate the global word-processor business, Korea stayed a country where Word did not win, because HWP simply handled Korean typesetting better and everyone in government and school already used it.

Then 1998 happened. Piracy was rampant, the company was bleeding, and Hancom agreed to sell the HWP source code to Microsoft for around twenty million dollars. The condition attached to that money was the part that still makes my jaw tighten: Microsoft wanted HWP development stopped. Korea would write its documents in Word, like everyone else. What followed was one of the strangest episodes in software history, a national “Save Hangul” movement where citizens and companies raised money and bought legal copies specifically to keep a word processor alive. It worked. The deal collapsed, HWP lived, and Hancom learned that its real product was never the software. It was the idea that a country should own the tool it writes itself with. If that sounds dramatic for a .hwp file, remember we just spent a year watching a US administration pull the plug on frontier AI access overnight. A foreign giant trying to switch off a national tool is not a 1998 problem.

The part where it gets messy

Surviving and thriving are different things, and Hancom’s cap table reads like a hostage being passed between kidnappers. Prime Group bought it in 2003, TG Sambo in 2009, and by 2010 the company had changed hands roughly eight times. It finally landed with Kim Sang-cheol, who folded it into what is now the Hancom Group and still chairs it. The thing that kept HWP breathing through all of that was uncomfortable: the Korean government treated the format as a de facto standard, so every public office, court, and university ran on it, which meant Hancom had a captive market regardless of who owned the logo that week.

That dependence cut both ways, and this is the part of the story I care about most. The HWP binary format was closed, which is a bad property for the permanent records of a democracy to live inside a single private company’s file structure. Critics hammered this for years. In 2017 the government started moving official documents to ODT, the international open standard, and Hancom, reading the room, made HWPX its default in 2021, an open XML format built on something called OWPML so the document structure is finally inspectable. The irony nobody in Korea enjoys: in that same 2017, Hancom lost a US federal case for violating the GNU GPL by shipping Ghostscript code without honoring the license. The national champion of software sovereignty got caught stepping on someone else’s open-source rights. Sovereignty for me, not for thee.

I keep coming back to this because it is the cleanest real-world test of an argument I’ve been making all year. Owning your governance is not the same as owning your infrastructure. Europe writes sovereignty rules without the sovereignty hardware to back them, and the whole point of the shift from open-source-as-preference to open-source-as-imperative is that the file format, the model weights, the thing on disk, is what you actually control or don’t. Korea is the rare case that has owned the infrastructure for thirty years and is only now being dragged toward owning it in the open. HWPX is Hancom finally writing its own Haerye, a format whose insides anyone can read.

What they actually sell

The core is still Hancom Office: the Hanword word processor (HWP), Hancell for spreadsheets, Hanshow for presentations, a PDF tool, and Hancom Sign for e-signatures. Around that sits the cloud push, Hancom Docs, launched September 2022 to replace the older Hancom Space, with an HTML5 Web HWP editor so you can open the country’s documents in a browser without installing anything. There is a cross-platform cloud suite, ThinkFree, underneath a lot of it. The user base is north of 20 million, and the moat is the same one from 1998: Korean government, education, legal, and academic workflows run on HWP, full stop. On Samsung Galaxy devices the suite has shipped free through the Galaxy Store, which is how a generation of Korean students first met it.

The money is healthier than the chaotic history suggests. Hancom Inc on a standalone basis did about 175 billion won in 2025 at roughly a 30% operating margin, and per the Seoul Economic Daily it set a 2026 target of 210 billion won, which would be its first time clearing 200 billion standalone since 1990. The company is family-run now: chairman Kim Sang-cheol’s eldest daughter, Kim Yeon-soo, is CEO of Hancom Inc, Boston-educated, and she is the one steering the pivot. Her framing is blunt, that hitting the target with half of revenue coming from outside office software would prove “Hancom is no longer an office company.” The strategy has a name, “AI Orchestrator,” and a product wave behind it, Hancom Assistant and Hancompedia and a Data Loader aimed squarely at public-sector AI projects, plus an eKYC identity-verification business they are pushing into Japan and a December 2025 cloud alliance with Tencent to get into China. Whether “AI Orchestrator” is a real architecture or a slide title, I cannot tell you yet, and I am suspicious of any legacy software company that discovers it was an AI company all along.

The sprawl, and where my thesis breaks

Here is where Hancom stops being a word-processor story and becomes something I find hard to summarize. The Hancom Group is around 26 affiliates with four listed companies, and the way it funded the reinvention was by selling Hancom MDS, its embedded-systems arm, in 2022 to raise the biggest cash pile in its history. Then it went shopping.

The strangest and best of the affiliates is Hancom InSpace, a spin-off of the Korea Aerospace Research Institute. In May 2022 it put Sejong-1 into low Earth orbit, the first commercial satellite for a private South Korean company, built by Spire and launched on a SpaceX rideshare. Sejong-2 followed in 2024, a hyperspectral Sejong-3 is on the roadmap, and the plan is a constellation of up to 50 satellites feeding an AI image-analysis platform called InStation that watches farms, forests, ships, and cities. Hancom raised its stake to a controlling 31.4% in 2025 and is steering InSpace toward its own IPO. A company named after a 15th-century alphabet now names its satellites after the king who invented it. I did not expect to write that sentence today, and I love it.

The rest of the portfolio is where my tidy narrative falls apart, so I am not going to pretend otherwise. There is Hancom GMD in digital forensics, Hancom Lifecare in safety equipment with an AI “smart safety net,” Hancom WITH in biometrics and security, a five-million-dollar stake in the Spanish facial-recognition firm FacePhi, an investment in Taiwan’s KDAN. And then there is a gold exchange and an auction house, plus a Web3 affiliate that once announced an NFT alliance, and at that point any honest writer has to admit the “coherent sovereignty strategy” frame stops explaining the company. A word-processor firm running a gold exchange is a rabbit hole I am leaving for someone with more patience than I have tonight. Some of this is vision and some of it is a holding company throwing darts, and I cannot always tell which from the outside.

So where does that leave it. The bull case is that Hancom is the rare national-champion software company that actually controls its stack, from the file format to the satellite, in a year when owning the box no government can switch off stopped being a hobbyist concern and became a procurement requirement. The bear case is that the moat is a government mandate the same government is slowly dismantling toward open formats, and the AI-plus-satellites-plus-biometrics sprawl looks less like a plan than a search for the next thing to be when HWP finally stops being load-bearing. I land closer to the bull case than I expected to when I started reading, mostly because Sejong was right the first time: the thing that lasts is the open specification, not the empire around it. I would bet on HWPX outliving the AI Orchestrator. Ask me again when Sejong-3 is in orbit.