Japan's multi-tier subcontracting structure in the SIer industry is usually framed as the obstacle that will block transition. Dissect the structure and the conclusion reverses — precisely because the structure is multi-tier, the industry shift can proceed without internal lay-offs.
Chapter 9 confirmed that the original driver of IT outsourcing was securing large quantities of coder person-months. This chapter takes the other side — when AI removes coder demand, how the structure built to source those person-months actually moves.
The focus is on Japan-specific dynamics, but the conclusion is simple: multi-tier subcontracting, paradoxically, makes the transition easier.
Multi-tier subcontracting, paradoxically, makes the transition easier
Sketch the typical Japanese SIer structure.
- Prime contractors — large SIers contracting directly with customers. Hundreds to thousands of employees.
- Tier-1 subcontractors — mid-size SIers taking commissioned work from primes.
- Tier-2 and tier-3 subcontractors — smaller vendors and independent contractors below them, supplying head-count by the person-month.
- For some engagements, four to five layers stack.
The reason this hierarchy is disliked is easy to state — margins pile up at each layer, profit does not reach the working coder, accountability gets blurred. Chapters 6 and 7 read this as the structure that inflates price.
But at a transition, the structure's property reverses. Multi-tier subcontracting externalizes coder demand into contracts. Workers are not held inside the prime contractor as employees; they sit in contractual relationships.
What does that mean? When demand disappears, shrinkage happens by not renewing contracts. The prime's own employees do not need to be laid off. Legally, politically, and in public-opinion terms, non-renewing a contract is far easier than adjusting employment.
(employees)"] S1["Tier-1 subcontractor
(service contract)"] S2["Tier-2 subcontractor
(service contract)"] S3["Tier-3 subcontractor
(service contract)"] P --> S1 S1 --> S2 S2 --> S3 end AI["AI takes execution
↓
coder demand disappears"] AI -.->|"non-renewal
= shrinkage without lay-offs"| S3 AI -.->|same| S2 AI -.->|gradually| S1 classDef bad fill:#fef3e7,stroke:#c89559,color:#5a3f1a classDef good fill:#e8f5e9,stroke:#7a9a6d,color:#3a4d34 class P,S1,S2,S3 bad class AI good
Multi-tier subcontracting externalizes coder demand into contracts. When demand disappears, shrinkage happens by not renewing contracts — multi-tier subcontracting acts as a shock-absorber during the transition.
Primes can transition without internal lay-offs by ending subcontractor contracts
Look at the concrete motion.
Say a prime SIer runs 100,000 person-months a year of engagements. The breakdown:
- Own employees: 10,000 person-months
- Sourced from tier-1 subcontractors: 30,000 person-months
- Tier-2 and tier-3: 60,000 person-months
As AI-native engagements grow and coder demand falls, what does the prime do?
- Employees stay — dismissal is legally and culturally difficult. Reassign for builder training or for customer-facing judgment work.
- Shrink subcontractor contracts from the bottom up — do not renew tier-3 contracts first, tighten tier-2 orders.
- New engagements go AI-native — the prime's own employees stand on the judgment side; AI handles execution.
- Staged shift — keep existing long-term maintenance contracts for their duration; evaluate AI-native replacement at renewal.
All of this can be executed by central management decision at the prime alone. No internal employment adjustment is required. There are external effects (subcontractor downsizing), but that is the contracted allowance.
For that reason, the decision barrier to AI-native transition is low for the prime SIer's management. Conversely, not choosing transition means losing engagements to AI-native competitors — "standing still" becomes the higher business risk, and at that point, transition accelerates.
Talented subcontractor coders flow to primes or independence
What happens to the side being shrunk — the subcontractors, especially the coders inside tier-2 and tier-3?
This is hard. When demand disappears, contracts are not renewed. Mid-size subcontractor vendors that stop getting engagements have to scale down or close.
But, for talented coders, there are several exits:
- Move to a prime — primes need people who can stand on the judgment side (= builder candidates). Coders with judgment ability can be absorbed into the prime's full-time employee slots.
- Move to a customer — hired directly as in-house builders by customer companies that used to commission SIers (Chapter 9).
- Go independent — individual contractor or small firm — contract directly with customers as a builder. The lawyer/doctor-style professional model from Chapter 9.
- Move to a different industry — leaving software development is also a path (the same kind of redistribution as "human computers" and typesetters moving to adjacent fields, from Chapter 3).
Not every mid- or lower-tier coder will move successfully. The layer with judgment ability — or the willingness to develop it — flows first. This is harsh, but the industry as a whole moves toward talent flowing upward and outward, not stagnating at the bottom.
What shrinks is the bottom; what flows is judgment ability. The more capable coders, the more likely they flow to primes, customers, or independence.
Labor mobility as the precondition
Everything above rests on the precondition of labor mobility.
Old-style Japanese employment — lifetime employment, seniority-based advancement, internal reassignment — supported the multi-tier structure. The prime was expected to hold a new graduate as a coder for 40 years; the shortfall was filled with subcontractors. Mobility was low.
That premise has been gradually weakening for two decades. Mid-career hiring has become normal, the recruiting market has matured, freelancer and independent-contractor contracts have proliferated. The AI-driven industry transition tests the upper bound of that mobility.
If mobility is high:
- Mid-career moves between primes increase
- Prime → customer-company moves grow (builder hiring per Chapter 9)
- Prime → independence (individual contractor, small firm)
- Subcontractor layers → all of the above
If mobility stays low:
- Talent stagnates in shrinking subcontractors
- New employment forms (the professional model) do not develop institutionally
- Transition pace is held back by social friction
Fortunately, mobility is trending upward. The post-COVID spread of remote work, the wider acceptance of side jobs, the shift toward job-type employment — all of these increase mobility. Society is moving the mobility lever in parallel with the AI shift.
Transitional forms — long-term service contracts, secondments, internal ventures
The transition does not happen overnight. It passes through intermediate forms.
- Long-term service contracts — a former coder contracts as an independent with the original employer or a customer company on a long-term basis. Not an employee, but with stable work.
- Secondment and transfer — an SIer employee is seconded to a customer company and becomes a builder there. If results land, transfer becomes an option.
- Internal ventures / spin-offs — an SIer stands up an AI-native business inside, or — to avoid conflict with existing engagements — spins it out as a separate company.
- Reverse-commission — a customer's in-house builder sells advice to other companies in their domain of expertise (the same shape as Chapter 5's "remaining one-tenth of specialists").
These are not permanent structures; they are shock absorbers for the transition. Japanese society has a cultural habit of moving through rapid change by absorbing it into intermediate forms.
The multi-tier structure allows rapid shrinkage. The intermediate forms absorb rapid change. Both effects act in parallel, so the transition is neither abrupt nor stalled.
Physical goods become scarcer than software
The SIer industry's shrinkage is not an isolated labor problem. Over the same few years, several forces are simultaneously pushing total labor demand upward across the economy. They share a single root — the era's scarce resource is flipping from software to physical things.
AI data-center construction, as the loudest visible example — the AI boom itself is generating massive demand for physical infrastructure. GPUs, fab equipment, electric power, cooling, buildings, land, networks — every one of them is about physical things, not about code. AI data-center construction is backlogged worldwide, with power supply as the bottleneck. The cheaper AI becomes, the more scarce the physical things that run AI — this is the most visible sign of "the era when physical goods get scarcer than software."
Reshoring of manufacturing — Middle-East instability, geopolitical energy-price rises, and global logistics-cost increases are eroding the economics of offshore production. Domestic manufacturing in Japan — especially high-value, low-volume, fast-response work — gains relative competitiveness. As reshoring proceeds, demand for shop-floor labor, production design, and manufacturing engineering rises in a measurable way.
Forced shift to natural farming — the supply of the major chemical-fertilizer inputs (ammonia synthesized from natural gas, potash from Russia and Belarus, imported phosphate rock) is destabilizing and prices are rising. Once the input cost of chemical agriculture crosses the breakeven line, natural farming is no longer a choice but a necessity. Natural farming requires more labor than chemical agriculture across soil preparation, weeding, and harvest — so agricultural labor demand also moves upward (the structural details are covered in the separate aiseed.dev series Phosphorus Depletion and Natural Farming).
The three forces — physical-infrastructure demand from AI itself, reshoring of manufacturing, the shift to natural farming — run on the same time scale as the SIer-industry shrinkage. As a result, the options open to displaced coders broaden significantly: alongside intra-industry flow (primes, customers, independence), the out-of-industry physical labor demand becomes a major channel.
The historical parallel from Chapter 3 — human computers and typesetters moving to adjacent fields — was viable only because labor demand happened to exist where they landed. The same applies here. The side where labor demand disappears (code production) and the sides where it grows (physical production, agriculture, AI physical infrastructure) are moving in parallel inside the same society.
The scarce resource of the era is flipping from software to physical things. It is not that SIer coders are in surplus — it is that the side that makes things does not have enough hands. That is the actual shape of labor demand.
Mobility rises over time
Finally, the direction of the transition period.
Labor mobility, job-type employment, professional models, social acceptance of business commissions — all of these have measurably risen over the past decade. Multiple forces will keep them rising:
- Demographics — a shrinking working-age population pushes mobility up on pure economic grounds.
- International comparison pressure — comparison with overseas (especially US) professional markets pushes Japanese companies to revise compensation.
- AI itself as pressure — talent that does not fit old-style employment models (builders, AI specialists) accumulates, creating pressure for institutional reform.
- Policy direction — government policy on "job-type employment," legalized side jobs, advanced-professional regimes all point the same way.
Mobility will keep rising. As time passes, the friction of the transition into the AI-native industry structure shrinks. The friction maximum is in the first few years; after that the transition tends to accelerate.
Education and hiring axes move at the same time
Alongside labor mobility, another axis has to move — the foundational discipline of the technical profession. Japan's science-and-engineering education has long centered on programming languages, frameworks, and design patterns — the core of software engineering. Once AI has taken that core, the human side has to shift its weight onto the liberal arts (Chapter 4) — logic, verbalization, ethics, systems thinking, history. The shift runs through everything from university curricula to corporate hiring criteria. The question "can you write code?" gives way to "can you judge?" on both sides at once.
Where the next chapter goes
We have seen how the AI-native structural change moves Japan's SIer industry and how labor mobility absorbs that motion. One question remains — over what time scale does this transition complete?
The next chapter takes up the case that the transition completes in a few years. The final chapter.