Why Korea Doesn't Buy Software — A Build/Buy/Assemble Reframe of Korea's AI B2B Market
Korea's software spending sits at 0.3% of GDP (rank 51) vs. R&D at 5.0% (rank 1). The series thesis reduces five 'peculiarities' of Korea's market to one root cause.
Exactly Inverted
World’s 10th-largest economy · Top-tier telecom infrastructure · OECD-leading individual ChatGPT usage · Consistently top in UN e-government rankings.
Yet Korean enterprises spend just 0.3% of GDP on packaged software. The U.S. spends 1.3%. This isn’t an impression — it’s a statistic. By the WIPO Global Innovation Index 2025 (139 economies, source: S&P Global Market Intelligence, 2024 data), Korea ranks 51st in Software Spending as % of GDP.
Same report, same page, same Korea — R&D expenditure stands at 5.0% of GDP, ranked #1 in the world. Business-funded R&D is also #1. Combined R&D investment by the top 3 corporate spenders is also #1.
R&D is #1, software is #51. Is this a coincidence?
This is the 0th installment and thesis declaration of a 4-part series. It reduces the five “peculiarities” commonly cited about Korea’s AI B2B market — SI dominance, CSAP, local LLMs, multi-layer approval, Korean tokenization issues — not as five independent facts but as different faces of one root cause: a stuck build market. And it raises the real question: if global is also returning to build through AI, is Korea trapped, or is it a leading indicator global is following?
The Limits of the “Korea Is Conservative” Explanation
The most common reading of this gap is “Korean B2B is conservative and doesn’t buy SaaS.” Intuitive, and partially correct. But too much data conflicts with the conservative hypothesis.
| Area | Korea’s position | Fits “conservative” hypothesis? |
|---|---|---|
| Individual ChatGPT usage | OECD top tier | No |
| Telecom · payments · fintech infrastructure | Global #1–#2 | No |
| Digital public services (UN) | Global top | No |
| AI Pilots (large enterprise · public) | Active | No |
| Production conversion rate | Low | Yes (but not explained by conservatism) |
Conservative markets don’t run pilots. Korea runs pilots actively, then stalls there. This break signals a different mechanism at work — not conservatism.
Another phenomenon the conservative hypothesis can’t explain — Korean SaaS companies can’t sell at home and detour through Japan.
- LINE WORKS (Naver Cloud): Just 12.0% (#2) in Korean groupware and 6.3% (#5) in business messengers, while taking #1 in Japan’s paid business messenger market for 8 consecutive years since 2017 (Fuji Chimera business chat market revenue share 33.6%, #1 for 5 consecutive years).
- Younglimwon Soft Lab SystemEver: Established a separate Tokyo subsidiary after securing 100+ Korean clients. Leveraging the fact that Japanese ERP license prices run 3x Korean prices, focusing growth on Japan.
- RSUPPORT: #1 in Korea’s remote solution market, but Japan has been a critical overseas market since the early years. SaaS demand for Japanese telework exploded after the pandemic.
Same companies, same products — blocked by SI outsourcing and on-prem pressure in Korea, but landing as standard SaaS in Japan. This isn’t Korea being conservative; it’s that a SaaS-buying market structure was never formed here. Detailed case analysis follows in Part 2.
Build / Buy / Assemble — Three Stages of Software Market Evolution
The global software market over the past 30 years moved through three broad stages:
%%{init: {'look': 'handDrawn', 'theme': 'neutral'}}%%
graph LR
A[Build<br/>1990s~] --> B[Buy<br/>2000s–2020]
B --> C[Assemble<br/>2022~]
style A fill:#fef3c7,stroke:#d4a373
style B fill:#d1f0c5,stroke:#7a9e7e
style C fill:#ffd6a5,stroke:#c98a5a- Build (1990s–): In-house development or SI outsourcing. Markets demanded “custom-built.”
- Buy (2000s–2020): Standardization through packaged software and SaaS subscriptions. The golden age of Salesforce, Workday, ServiceNow.
- Assemble (2022–): APIs, composable, agentic. Not one big SaaS chunk but small components and agents stitched together.
Simplified positioning of the U.S., Japan, India, and Korea:
| Country | Primary stage | Mechanism |
|---|---|---|
| United States | Buy → Assemble (transitioning) | In-house engineering + full SaaS stack |
| Japan | Build → Buy (transitioning) | Gradual SaaS adoption, partly conservative |
| India | Buy + own build (parallel) | Global IT services industry + domestic SaaS growth |
| Korea | Build holdout (outsourced build) | SI commissioning, sparse in-house engineering |
To state Korea’s position precisely: not “stuck at the build stage” but “stuck at outsourced build”. This distinction is the spine of the entire series. In-house engineering teams building directly is a different mechanism from commissioning SI to build for you — same word, different game. A ‘build’ where global engineering teams continuously evolve their codebase is different in output, cycle, and pricing from a ‘build’ where Korean conglomerates commission N people for M months from SI. This distinction returns as the key variable in the final installment (AI-era build resurgence).
The Five Peculiarities Are Symptoms, Not Causes
Now look at the results this outsourced-build position produced. The five items typically cited as peculiarities of the Korean market — SI dominance, CSAP, local LLMs, multi-layer approval, Korean tokenization — that’s the result list. Usually treated as five independent variables, but viewed as different faces of one root cause, the picture changes.
SI Dominance — The Mechanism for Outsourcing Build Demand
Samsung SDS, LG CNS, SK C&C — the SI Big 3 dominating Korea’s IT market is a result. The market demands build (custom systems on a company’s own infrastructure), and the mechanism that processes that demand through outsourcing is SI. If Korea were a SaaS-buying market, SI wouldn’t be this big. SI isn’t strong because SI is inherently strong — SI is strong because build demand gets processed through outsourcing.
CSAP · Network Separation · Data Sovereignty — Result of Build Markets Captured by Government Procurement
CSAP (Cloud Security Assurance Program), network separation (망분리), and data sovereignty laws are entry barriers for global SaaS. The common reading is “Korea is closed.” But these regulations form when build markets become deeply dependent on public and financial procurement. Once buyers standardize “deploy directly inside the internal network,” that standard hardens into a security regulation. The entry barrier is a byproduct of build commissioning practices.
Local Big-Tech Strength (HyperCLOVA · EXAONE · KAKAO) — Native LLMs in a Closed Market
Naver HyperCLOVA, LG EXAONE, KAKAO — their LLM strength is also a result. Korean tokenization efficiency is a partial factor, but the bigger one is that while foreign players couldn’t enter, government R&D and public procurement nurtured native LLMs. HyperCLOVA didn’t take “a different path from global”; Korea’s market structure differs in how foreign players enter, and that’s what allowed native models to survive.
Multi-Layer Approval (Planning · Strategy · Operations · IT · Security · Legal · CEO) — Governance for Outsourced Commissioning
The 7-layer approval chain is also a build-market byproduct. In a market where companies build directly, an engineering leader decides and that’s it. In an outsourced commissioning market, who is responsible for the commission is the central question, and that distributed responsibility hardens into multi-layer approval. It’s a different species from SaaS decision-making.
Korean Tokenization · Translation Lag — Residual Cost from Closure
GPT token efficiency, Korean LLM performance, English content translation lag — each is a real cost, but secondary to the series thesis. Had the market been open to buy, foreign SaaS and foreign LLMs would have invested more in Korean adaptation. A closed market doesn’t get that investment, leaving residual costs.
The five peculiarities aren’t to be analyzed individually. They are five shadows of one fact: Korea is not a buy market but a build (specifically, outsourced build) market.
Thesis — Only Half Right
That’s the half the series agrees with. “Korea is trapped in build” is half right. But the other half is missing. Global is also returning to build through AI.
Evidence is everywhere. Explosive adoption of in-house build tools like Cursor, Claude Code, and v0; “SaaS is dead” discourse; in-house sLLMs and agentic workflows; large enterprises internalizing their own LLMs. Investors like a16z and Bessemer publish piece after piece arguing vertical agents replace horizontal SaaS. The SaaS pendulum is swinging back toward build.
Here, a “things come full circle” conclusion is dangerous. Korea’s build and AI-era build are the same word, different mechanisms.
| Aspect | Korea’s build | AI-era build |
|---|---|---|
| Actor | SI outsourcing firm | In-house engineering team |
| Tools | Custom coding, headcount inputs | AI collaboration (Cursor, Claude Code, agentic) |
| Cycle | 6–24 months (waterfall) | Days to weeks (iterative) |
| Output | A system built once | An evolving codebase |
| Pricing | Headcount × duration | Outcome · subscription · hybrid |
Direction converges toward Korea. Method risks Korea falling behind. Korean conglomerates and mid-market firms have sparse in-house engineering teams compared to global peers, and build demand still routes to SI. Whether SI can convert AI-era build into in-house engineering — that’s the real question to test.
So the real thesis of the series is:
The diagnosis “Korea is trapped in build” is only half right. Global is returning to build through AI. The real question is whether Korea can jump from SI-outsourced build to engineering-team build — and that jump determines the fate of Korea’s software market over the next 5 years.
4-Part Series Preview
The thesis is decomposed along a time axis into 4 installments. Each is readable independently, but the order — Part 0 (thesis) → Part 1 (past) → Part 2 (present) → Part 3 (future) — follows the causal chain.
| # | Title | Topic | Anchor evidence |
|---|---|---|---|
| 0 | This article | Series thesis + Build/Buy/Assemble frame | WIPO GII 2025, R&D #1 vs SW #51 |
| 1 | How We Got Trapped | Government SW pricing, Hancom, 30 years of SI outsourcing | Public SW pricing history, SI Big 3 revenue |
| 2 | The Cost of Being Trapped | Korean SaaS detouring through Japan, foreign SaaS blocked | LINE WORKS · Younglimwon · RSUPPORT cases |
| 3 | Will AI Free or Cement | Deregulation × AI Build × SI inertia collision | CSAP 2023 reform, K-PaaS, in-house sLLM trends |
Closing Question
The most important question for Korea’s software market over the next 5 years isn’t whether SI revenue declines. It’s whether Korean enterprise in-house engineering teams can build directly using AI tools.
If they can, Korea moves closer to the conclusion that it was a leading indicator global is following — not a trapped market. If not, it remains a market locked in SI outsourcing where in-house build capacity never grew, even in the AI era.
Is Korea trapped, or is it a leading indicator that global is following? This series tests the answer.
Sources
- WIPO Global Innovation Index 2025 — Indicator 6.2.3 Software spending % GDP, source: S&P Global Market Intelligence, 2024
- LINE WORKS achieves #1 in Japan’s business chat market — Naver Works official PR (Korean)
- LINE WORKS at 10 years — Naver Cloud’s AI Work Platform leap (CNET Korea, Korean)
- Younglimwon Soft Lab Cloud ERP SystemEver targets Japan and Indonesia (Electronic Times, Korean)
- “ERP prices in Japan run 3x Korea” — Younglimwon’s overseas push (THE ELEC, Korean)
- How Korean and Japanese SaaS markets differ — KOTRA cited via Japan Insight (Korean)
Series Index: Korea Build Trap
Related Posts
The Cost of Being Trapped — How LINE WORKS, Younglimwon, and RSUPPORT Detour Through Japan
Korean SaaS companies blocked by SI and on-prem walls at home, then taking #1 positions in Japan — a natural experiment in Korea's build lock-in.
How We Got Trapped — Government SW Pricing, Hancom, and 30 Years of SI Outsourcing
How public SW pricing, Hancom/groupware NIH, and 30 years of SI outsourcing produced Korea's build lock-in.
Will AI Free or Cement Korea — Deregulation × AI Build × SI Inertia
Three forces (deregulation × AI build × SI inertia) producing three scenarios for Korea's software market — Korea Vindicated, Double Loss, Bifurcation. The real branching point.