SIer commissioning runs at tens of millions to hundreds of millions of yen. AI-native development runs at the order of hundreds of thousands of yen per month. The 10× to 100× gap is no longer competition — it is market displacement.
Chapter 6 showed that the SIer commission model has reached the level where "with the same effort, you can build it yourself." This chapter takes the next question — not effort, but money — and lines the two up directly.
The conclusion up front. The gap between SIer commissioning and AI-native development runs at 10× to 100×. This chapter walks through the basis for that number and what it means.
Quote the same scope, both ways
Use a concrete quote. A mid-size business system — customer master, order processing, invoicing, a dashboard, SaaS-shaped, scope is defined, users in the hundreds to a few thousand.
SIer commissioning, market range:
- Requirements + basic design: a few million to ten-plus million yen
- Development (6–12 months, several coders): tens of millions to over a hundred million yen
- Operations and maintenance (annual): a few million to ten-plus million yen
- Multi-year contract total: on the order of hundreds of millions of yen
"Tens of millions for a mid-size system" is the standard range in Japan's SIer industry. This is not a special case.
AI-native build, the same scope:
- Tooling: Claude Max (¥30,000/mo) plus other tools, ¥50–100K/mo
- Builder labor (1 person × a few months): a few million yen
- Operations and maintenance (1 person, ongoing): hundreds of thousands per month
- Total at the same scope: a few million to ten million yen
The gap is 10× to 100×. Depending on scope it falls at the narrower or wider end, but it is always an order of magnitude or more.
(¥10M–¥100M+)"] direction TB S1["Coder person-months
(the body)"] S2["Prime contractor margin"] S3["Project management"] S4["Proprietary frameworks,
ops & maintenance"] end subgraph New["AI-native build
(order of ¥hundreds of K/month)"] direction TB A1["Claude Max etc.
subscriptions"] A2["Builder labor"] end Old ==>|"10× to 100× gap
= not competition, but displacement"| New classDef good fill:#e8f5e9,stroke:#7a9a6d,color:#3a4d34 classDef bad fill:#fef3e7,stroke:#c89559,color:#5a3f1a class New good class Old bad
10× to 100× is not competition; it is market displacement
The meaning of a price gap differs qualitatively with order of magnitude.
- 1.2× — price competition. Customers choose by service, track record, and relationship. Both providers coexist in the market.
- 2–3× — hard competition. The cheaper side picks up customers, but the more expensive side still has reasons to keep some (trust, relationship, expertise).
- 10× — structural advantage. Customers flow to the cheaper side unless there is a strong reason not to. The expensive side retreats to limited territory (the one-tenth of specialist work).
- 100× — market displacement. Calling it the same market becomes meaningless. A different supply curve.
A 10×–100× gap is not a competition story. When pocket calculators arrived at one-tenth the price of an abacus, abacus makers did not "lose on price" — the market itself moved (Chapter 3). The same structure is happening now, between the SIer industry and the software development market.
What matters here is that many customers will take time to notice the order-of-magnitude difference. Two reasons:
- The established sense of what things cost is deeply held — "no way this gets built for a few million yen"
- The option of "building it yourself" is not known, or not seriously evaluated
But once a customer has seen it, they cannot go back. "What was supposed to cost hundreds of millions of yen ran for a few million" — after that experience, SIer commissioning drops out of the option set.
1.2× is competition. 2–3× is hard competition. 10× is structural advantage. 100× is market displacement. Between SIer commissioning and AI-native development, the gap runs at the latter end of that scale.
Japan has the widest price gap in the world
In Japan, the gap is even wider, for specific reasons.
- The SIer industry is large — a substantial share of IT spend flows through SIer commissioning. The in-house build ratio is low compared with other markets (especially the US).
- Multi-tier subcontracting — work flows from prime to tier-1, tier-2, tier-3 subcontractors, with margin stacked at each layer. By the time code is written, multiple intermediate layers have taken their cut (the structural details are covered in Chapter 10).
- Weak yen against USD pricing — AI tools are priced in USD; SIer labor is priced in JPY. Exchange-rate dynamics push the SIer rate up relatively, year after year.
- Industry standard is the person-month — pricing is negotiated as "unit rate × person-months," so productivity gains do not translate into price drops.
These stack up. SIer commissioning prices in Japan run higher than in the West. AI-native development cost, on the other hand, is common worldwide (same Claude, same GPT, same Cursor). The result: the price gap between SIer commissioning and AI-native development in Japan is among the largest in the world.
This is a threat and an opportunity at once. The larger the gap, the larger the saving when a customer migrates to AI-native. For builders or organizations who can offer AI-native development services in Japan, the opportunity is greater than in Western markets.
The wider the price gap, the larger the saving on migration. The opportunity in Japan is larger than in the West.
Why SIers cannot follow on price
Restate the structural reasons from Chapter 6, in price terms.
The floor on SIer pricing is set by their own payroll. Coder salaries, social insurance, office costs, management overhead — these fix the bottom. However many times AI multiplies productivity, while they still have to keep paying the people, prices cannot drop by an order of magnitude.
On top of that, in the Japanese SIer model:
- The prime contractor adds margin and takes the order
- The subcontractor adds another margin on top
- For some engagements, four to five layers stack up
Each layer takes margin, so the cost of one coder swells to several times that amount by the time the customer is invoiced. In an AI-native build, none of these intermediate layers exist. The cost is one builder's labor plus tooling.
The SIer's price floor is determined by how many payroll layers must keep being paid. AI getting cheaper does not erase those layers.
Customers with lock-in cannot move yet
Even a 10×–100× gap will not move every customer immediately.
- Customers tied into existing operations contracts
- Codebases dependent on proprietary frameworks and abstraction layers
- Organizations where long-standing personal relationships anchor decisions
- Regulated industries that include SIer track record as a requirement
These all function as lock-in. Even with an order-of-magnitude price gap, migration cost is hard to see and the decision gets deferred.
Lock-in strength differs case by case:
- New projects — no lock-in. The price gap acts directly.
- Extensions to existing systems — partial lock-in. Gradual migration.
- Full replacement of core systems — strong lock-in. Moves last.
What moves first is new projects for new customers. Next, shallow lock-in extensions. Core systems come last. That ordering sets the pace of the industry shift.
The structure of lock-in itself — where it comes from, how it operates, why it is so durable — is covered in the next chapter. Palantir's FDE model is dissected as the archetype.
Where the next chapter goes
A 10× to 100× price gap does not, by itself, move the whole market overnight. Lock-in is the inertia that holds the pace down. Where does lock-in come from, and why is it strong?
The next chapter takes up the lock-in problem.