The Vertical SaaS Moat That Might Be Empty
The Thesis
A recent Chinese article argued that AI startups should "hide in places big companies don't care about." The author used a construction diary AI company as the key example: embedded in approval workflows, private data, high switching costs -- big tech can't touch it, so it thrives.
The first half is reasonable. Wrapper apps are fragile. But "hide in a vertical niche and you're safe" skips several steps. I fact-checked the construction diary case and found that the problem isn't whether big tech shows up -- it's that the land itself doesn't grow crops.
The Example Falls Apart
When a commenter pressed for specifics, the author named three companies: Justwin, Modou Tech, and Huajiantong. He added that "these companies don't show up in funding news, so it does sound made up."
I looked into all three. Justwin was founded in 2003 as a traditional project management platform -- construction diaries are a minor feature. Modou Tech does construction labour identity verification, processing over 10 billion yuan in worker wages. Huajiantong is backed by a state-owned enterprise with CMMI5 certification, offering its platform free to government agencies.
Two things stand out: none of them are primarily in the construction diary business, and none of them are AI companies. Using these three to argue that "AI startups can thrive on construction diaries" is building on a premise that doesn't exist.
The Industry Ceiling
Glodon (002410.SZ) is China's undisputed leader in construction software. Nearly 30 years in the market, 390,000+ enterprise clients, 80+ subsidiaries, 10,000+ employees.
In 2024, Glodon posted revenue of 6.2 billion yuan with a net margin of 4%. Here's the comparison that matters: Conch Cement, which literally makes cement for construction sites, posted an 8.5% net margin. Sany Heavy Industry, which builds excavators, posted 7.6%. Both are capital-intensive businesses -- factories, equipment, logistics, inventory.
Glodon sells software. Near-zero marginal cost. In theory, it should have far higher margins than its hardware counterparts in the same value chain. In practice, selling code to the construction industry earns less than selling cement. That's not a Glodon problem. That's a structural ceiling: the industry's willingness to pay for software is so low it completely erases the scalability advantage.
McKinsey's data confirms it: construction ranks second-to-last in digitization across all industries, ahead of only agriculture.
A Reference Point from Stockholm
A friend of mine, Ludvig, co-founded Teamhub AB in Stockholm in 2015 -- a construction site management SaaS covering scheduling, quoting, work orders, time tracking, materials, invoicing, and subcontractor management. Far broader than just construction diaries.
Sweden has mature software-buying culture, strong compliance norms, and relatively gentle competition. Even so, Teamhub has raised only $550K over ten years. Annual revenue: 4 million SEK (roughly $380K). Team size: three people. Growth: essentially flat.
Sweden's construction SaaS space has about 51 startups, all stuck in the same rut: fragmented customers, low willingness to pay, long sales cycles, field workers who resist using software.
Construction tech VC firm Foundamental put it plainly: "Pure SaaS construction tech companies rarely scale within a decade." Globally, 53% of construction tech startups fail, and 73.4% don't survive ten years.
If construction SaaS struggles in Sweden -- one of the best markets for software adoption -- it's not going to be easier in China.
The Symmetry of Lock-In
The original article's moat logic: once embedded in approval workflows, switching costs are high, so customers can't leave, and startups are safe.
The first half is correct. But it misses a symmetry: the friction that prevents customers from leaving also prevents new customers from arriving. If switching systems requires retraining staff and migrating historical data, that barrier doesn't just block competitors -- it blocks your own expansion.
Lock-in economics research has long established that high-switching-cost markets grow slowly and have high customer acquisition costs. You're hard to replace, but you're also hard to scale. This explains why Glodon needed nearly 30 years to reach its current size, and why Teamhub's growth flatlined.
One more thing: construction diaries are already a built-in, free feature in platforms like Procore, PlanGrid (acquired by Autodesk), and Glodon. A company built around this feature isn't competing against general-purpose AI -- it's competing against industry giants that give this functionality away for free.
Moats Without Water
Should AI startups hide in places big companies don't care about? The direction is right, but the conclusion is too optimistic. Places big companies "don't care about" are often places the market itself can't sustain a business.
Vertical moats are real. But moats don't equal profits, and certainly don't equal "thriving."
Glodon traded 30 years and 10,000 employees for a 4% net margin. Teamhub spent a decade in ideal conditions and barely grew. 73.4% of construction tech companies don't make it past ten years. The pattern suggests that in some verticals, the question isn't "will big tech come?" but "is this land fertile enough to farm?"
When choosing a vertical, instead of asking "will big companies build this?", perhaps ask: can this market's structural conditions -- willingness to pay, digitization maturity, customer concentration -- support healthy growth?
Originally written in Chinese. Translated by the author.